BENCHMARK - SyberLabs/RISE GitHub Wiki

RISE affect benchmark

Runnable models were executed on passages held in the RISE archive. pert-emopair and reward-emopair are cached CPU scores from roberta-large checkpoints. minilm-l6-probe is a frozen all-MiniLM-L6-v2 embedding plus a linear head; the cell is the leave-one-out prediction, not the in-sample fit. A teacher column appears only when AFFECT_TEACHER_URL returned axes. None of these models run inside the player. A model listed under "Models not executed" was not scored, and its published numbers are not copied here.

meditations-temper

Origin: rise-archive. Cases: philosophy, prose. Archive: src/content/archive/works/literary-meditations.js Β· Meditations Β· Marcus Aurelius.

From my grandfather Verus I learned good morals and the government of my temper.

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.18 (0.17 inferred) 0.15 (0.46 inferred) 0.04 (0.55 inferred) 0.10 (0.55 inferred) -0.00 (0.55 inferred) 0.50 (0.35 inferred)
arousal 0.26 (0.17 inferred) 0.14 (0.46 inferred) 0.20 (0.55 inferred) 0.40 (0.55 inferred) 0.44 (0.55 inferred) 0.70 (0.35 inferred)
dominance 0.15 (0.17 inferred) 0.14 (0.46 inferred) 0.07 (0.55 inferred) 0.36 (0.55 inferred) -0.19 (0.55 inferred) 0.30 (0.35 inferred)
tension β€” 0.10 (0.46 derived) β€” β€” β€” β€”
intimacy β€” 0.07 (0.46 inferred) β€” β€” β€” β€”
warmth β€” 0.04 (0.46 inferred) β€” β€” β€” β€”
uncertainty β€” β€” β€” β€” β€” β€”
expansiveness β€” 0.03 (0.46 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.47 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.04 (0.46 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.41 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.907 ms, mean confidence 0.17, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.749 ms, mean confidence 0.45, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 237.931 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 407.225 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.086 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 6385.439 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.35 (0.15, 0.50).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.46 (0.04, 0.50).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.40 (0.10, 0.50).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.50 (-0.00, 0.50).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.44 (0.26, 0.70).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.56 (0.14, 0.70).
  • Disagreement on arousal: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.50 (0.20, 0.70).
  • Disagreement on dominance: reward-emopair vs minilm-l6-probe differ by 0.55 (0.36, -0.19).
  • Disagreement on dominance: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.49 (-0.19, 0.30).
  • Teacher hosted-teacher: ok.

dickinson-death

Origin: rise-archive. Cases: poetry, irony, difficult. Archive: src/content/archive/works/literary-poems-dickinson.js Β· Poems Β· Emily Dickinson.

Because I could not stop for Death, He kindly stopped for me; The carriage held but just ourselves And Immortality.

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence -0.02 (0.16 inferred) 0.43 (0.45 inferred) 0.16 (0.55 inferred) 0.15 (0.55 inferred) -0.10 (0.55 inferred) 0.99 (0.35 inferred)
arousal 0.21 (0.16 inferred) 0.13 (0.45 inferred) 0.39 (0.55 inferred) 0.56 (0.55 inferred) 0.52 (0.55 inferred) 0.96 (0.35 inferred)
dominance -0.02 (0.16 inferred) 0.13 (0.45 inferred) -0.53 (0.55 inferred) -0.51 (0.55 inferred) -0.18 (0.55 inferred) 0.93 (0.35 inferred)
tension β€” 0.12 (0.45 derived) β€” β€” β€” β€”
intimacy β€” 0.16 (0.45 inferred) β€” β€” β€” β€”
warmth β€” 0.60 (0.45 inferred) β€” β€” β€” β€”
uncertainty β€” 0.05 (0.45 inferred) β€” β€” β€” β€”
expansiveness β€” 0.08 (0.45 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.53 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.03 (0.45 inferred) β€” β€” β€” β€”
solemnity β€” 0.09 (0.45 inferred) β€” β€” β€” β€”
novelty β€” 0.47 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.251 ms, mean confidence 0.16, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.296 ms, mean confidence 0.44, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped, contrast-marker-present.
  • pert-emopair: 133.693 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 301.814 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.029 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 5474.096 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs contextual-window-v1 differ by 0.46 (-0.02, 0.43).
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.01 (-0.02, 0.99).
  • Disagreement on valence: contextual-window-v1 vs minilm-l6-probe differ by 0.53 (0.43, -0.10).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.55 (0.43, 0.99).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.82 (0.16, 0.99).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.84 (0.15, 0.99).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.08 (-0.10, 0.99).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.75 (0.21, 0.96).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.43 (0.13, 0.56).
  • Disagreement on arousal: contextual-window-v1 vs minilm-l6-probe differ by 0.39 (0.13, 0.52).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.83 (0.13, 0.96).
  • Disagreement on arousal: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.57 (0.39, 0.96).
  • Disagreement on arousal: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.40 (0.56, 0.96).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.44 (0.52, 0.96).
  • Disagreement on dominance: lexical-vad-v1 vs pert-emopair differ by 0.51 (-0.02, -0.53).
  • Disagreement on dominance: lexical-vad-v1 vs reward-emopair differ by 0.49 (-0.02, -0.51).
  • Disagreement on dominance: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.95 (-0.02, 0.93).
  • Disagreement on dominance: contextual-window-v1 vs pert-emopair differ by 0.66 (0.13, -0.53).
  • Disagreement on dominance: contextual-window-v1 vs reward-emopair differ by 0.64 (0.13, -0.51).
  • Disagreement on dominance: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.79 (0.13, 0.93).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.46 (-0.53, 0.93).
  • Disagreement on dominance: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.44 (-0.51, 0.93).
  • Disagreement on dominance: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.11 (-0.18, 0.93).
  • Teacher hosted-teacher: ok.

austen-opening

Origin: rise-archive. Cases: prose, irony, difficult. Archive: src/content/archive/works/pride-and-prejudice.js Β· Pride and Prejudice Β· Jane Austen.

It is a truth universally acknowledged, that a single man in possession of a good fortune must be in want of a wife.

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.11 (0.28 inferred) 0.09 (0.70 inferred) 0.05 (0.55 inferred) 0.06 (0.55 inferred) 0.02 (0.55 inferred) 0.58 (0.35 inferred)
arousal 0.21 (0.28 inferred) 0.12 (0.70 inferred) 0.19 (0.55 inferred) 0.36 (0.55 inferred) 0.15 (0.55 inferred) 0.96 (0.35 inferred)
dominance 0.15 (0.28 inferred) 0.11 (0.70 inferred) 0.26 (0.55 inferred) 0.52 (0.55 inferred) -0.03 (0.55 inferred) 0.27 (0.35 inferred)
tension β€” 0.09 (0.70 derived) β€” β€” β€” β€”
intimacy β€” 0.01 (0.70 inferred) β€” β€” β€” β€”
warmth β€” 0.02 (0.70 inferred) β€” β€” β€” β€”
uncertainty β€” β€” β€” β€” β€” β€”
expansiveness β€” 0.09 (0.70 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.61 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.03 (0.70 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.45 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.119 ms, mean confidence 0.28, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.135 ms, mean confidence 0.64, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 138.795 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 468.658 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.069 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 5042.354 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.47 (0.11, 0.58).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.49 (0.09, 0.58).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.54 (0.05, 0.58).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.52 (0.06, 0.58).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.56 (0.02, 0.58).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.75 (0.21, 0.96).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.84 (0.12, 0.96).
  • Disagreement on arousal: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.78 (0.19, 0.96).
  • Disagreement on arousal: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.60 (0.36, 0.96).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.81 (0.15, 0.96).
  • Disagreement on dominance: lexical-vad-v1 vs reward-emopair differ by 0.37 (0.15, 0.52).
  • Disagreement on dominance: contextual-window-v1 vs reward-emopair differ by 0.41 (0.11, 0.52).
  • Disagreement on dominance: reward-emopair vs minilm-l6-probe differ by 0.55 (0.52, -0.03).
  • Teacher hosted-teacher: ok.

ishmael-water

Origin: rise-archive. Cases: prose, descriptive. Archive: src/content/archive/works/moby-dick-or-the-whale.js Β· Moby-Dick Β· Herman Melville.

Call me Ishmael. Some years agoβ€”never mind how long preciselyβ€”having little or no money in my purse, and nothing particular to interest me on shore, I thought I would sail about a little and see the watery part of the world.

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.05 (0.14 inferred) 0.01 (0.58 inferred) 0.02 (0.55 inferred) 0.04 (0.55 inferred) -0.05 (0.55 inferred) 0.58 (0.35 inferred)
arousal 0.23 (0.14 inferred) 0.14 (0.58 inferred) 0.18 (0.55 inferred) 0.42 (0.55 inferred) 0.41 (0.55 inferred) 0.96 (0.35 inferred)
dominance 0.12 (0.14 inferred) 0.00 (0.58 inferred) -0.37 (0.55 inferred) -0.30 (0.55 inferred) -0.11 (0.55 inferred) 0.27 (0.35 inferred)
tension β€” 0.15 (0.58 derived) β€” β€” β€” β€”
intimacy β€” 0.04 (0.58 inferred) β€” β€” β€” β€”
warmth β€” 0.00 (0.58 inferred) β€” β€” β€” β€”
uncertainty β€” 0.04 (0.58 inferred) β€” β€” β€” β€”
expansiveness β€” 0.07 (0.58 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.45 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.05 (0.58 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.40 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.269 ms, mean confidence 0.14, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.184 ms, mean confidence 0.55, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped, low-lexicon-coverage.
  • pert-emopair: 158.188 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 290.153 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.023 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 5090.894 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.53 (0.05, 0.58).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.57 (0.01, 0.58).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.56 (0.02, 0.58).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.55 (0.04, 0.58).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.63 (-0.05, 0.58).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.73 (0.23, 0.96).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.83 (0.14, 0.96).
  • Disagreement on arousal: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.78 (0.18, 0.96).
  • Disagreement on arousal: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.54 (0.42, 0.96).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.55 (0.41, 0.96).
  • Disagreement on dominance: lexical-vad-v1 vs pert-emopair differ by 0.49 (0.12, -0.37).
  • Disagreement on dominance: lexical-vad-v1 vs reward-emopair differ by 0.42 (0.12, -0.30).
  • Disagreement on dominance: contextual-window-v1 vs pert-emopair differ by 0.37 (0.00, -0.37).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.64 (-0.37, 0.27).
  • Disagreement on dominance: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.57 (-0.30, 0.27).
  • Disagreement on dominance: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.38 (-0.11, 0.27).
  • Teacher hosted-teacher: ok.

eyre-will

Origin: rise-archive. Cases: prose, dialogue. Archive: src/content/archive/works/jane-eyre.js Β· Jane Eyre Β· Charlotte BrontΓ«.

I am no bird; and no net ensnares me; I am a free human being with an independent will

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.07 (0.25 inferred) 0.18 (0.70 inferred) 0.06 (0.55 inferred) 0.06 (0.55 inferred) 0.01 (0.55 inferred) 0.98 (0.35 inferred)
arousal 0.31 (0.25 inferred) 0.18 (0.70 inferred) 0.50 (0.55 inferred) 0.76 (0.55 inferred) 0.38 (0.55 inferred) 0.97 (0.35 inferred)
dominance 0.16 (0.25 inferred) 0.31 (0.70 inferred) 0.46 (0.55 inferred) 0.97 (0.55 inferred) 0.19 (0.55 inferred) 0.96 (0.35 inferred)
tension β€” 0.20 (0.70 derived) β€” β€” β€” β€”
intimacy β€” 0.06 (0.70 inferred) β€” β€” β€” β€”
warmth β€” 0.04 (0.70 inferred) β€” β€” β€” β€”
uncertainty β€” 0.05 (0.70 inferred) β€” β€” β€” β€”
expansiveness β€” 0.05 (0.70 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.55 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.05 (0.70 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.41 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.069 ms, mean confidence 0.25, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.085 ms, mean confidence 0.65, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 129.650 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 318.680 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.022 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 4729.496 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.91 (0.07, 0.98).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.80 (0.18, 0.98).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.92 (0.06, 0.98).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.92 (0.06, 0.98).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.97 (0.01, 0.98).
  • Disagreement on arousal: lexical-vad-v1 vs reward-emopair differ by 0.45 (0.31, 0.76).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.66 (0.31, 0.97).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.58 (0.18, 0.76).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.79 (0.18, 0.97).
  • Disagreement on arousal: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.47 (0.50, 0.97).
  • Disagreement on arousal: reward-emopair vs minilm-l6-probe differ by 0.38 (0.76, 0.38).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.59 (0.38, 0.97).
  • Disagreement on dominance: lexical-vad-v1 vs reward-emopair differ by 0.81 (0.16, 0.97).
  • Disagreement on dominance: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.80 (0.16, 0.96).
  • Disagreement on dominance: contextual-window-v1 vs reward-emopair differ by 0.66 (0.31, 0.97).
  • Disagreement on dominance: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.65 (0.31, 0.96).
  • Disagreement on dominance: pert-emopair vs reward-emopair differ by 0.51 (0.46, 0.97).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.50 (0.46, 0.96).
  • Disagreement on dominance: reward-emopair vs minilm-l6-probe differ by 0.78 (0.97, 0.19).
  • Disagreement on dominance: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.77 (0.19, 0.96).
  • Teacher hosted-teacher: ok.

dalloway-morning

Origin: rise-archive. Cases: prose, descriptive. Archive: src/content/archive/works/mrs-dalloway.js Β· Mrs Dalloway Β· Virginia Woolf.

Mrs. Dalloway said she would buy the flowers herself.

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.45 (0.14 inferred) 0.41 (0.26 inferred) 0.03 (0.55 inferred) 0.05 (0.55 inferred) 0.03 (0.55 inferred) 0.50 (0.35 inferred)
arousal 0.25 (0.14 inferred) 0.14 (0.26 inferred) 0.04 (0.55 inferred) 0.18 (0.55 inferred) 0.36 (0.55 inferred) 0.75 (0.35 inferred)
dominance 0.00 (0.14 inferred) β€” -0.29 (0.55 inferred) -0.28 (0.55 inferred) 0.01 (0.55 inferred) 0.25 (0.35 inferred)
tension β€” 0.07 (0.26 derived) β€” β€” β€” β€”
intimacy β€” 0.05 (0.26 inferred) β€” β€” β€” β€”
warmth β€” 0.35 (0.26 inferred) β€” β€” β€” β€”
uncertainty β€” β€” β€” β€” β€” β€”
expansiveness β€” 0.01 (0.26 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.48 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.04 (0.26 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.45 (0.26 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.068 ms, mean confidence 0.14, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.089 ms, mean confidence 0.28, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped, low-lexicon-coverage.
  • pert-emopair: 115.439 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 332.639 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.025 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 4512.518 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs pert-emopair differ by 0.42 (0.45, 0.03).
  • Disagreement on valence: lexical-vad-v1 vs reward-emopair differ by 0.40 (0.45, 0.05).
  • Disagreement on valence: lexical-vad-v1 vs minilm-l6-probe differ by 0.42 (0.45, 0.03).
  • Disagreement on valence: contextual-window-v1 vs pert-emopair differ by 0.38 (0.41, 0.03).
  • Disagreement on valence: contextual-window-v1 vs reward-emopair differ by 0.36 (0.41, 0.05).
  • Disagreement on valence: contextual-window-v1 vs minilm-l6-probe differ by 0.39 (0.41, 0.03).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.47 (0.03, 0.50).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.45 (0.05, 0.50).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.47 (0.03, 0.50).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.50 (0.25, 0.75).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.61 (0.14, 0.75).
  • Disagreement on arousal: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.71 (0.04, 0.75).
  • Disagreement on arousal: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.57 (0.18, 0.75).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.39 (0.36, 0.75).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.54 (-0.29, 0.25).
  • Disagreement on dominance: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.53 (-0.28, 0.25).
  • Teacher hosted-teacher: ok.

paradise-woe

Origin: rise-archive. Cases: poetry, solemn. Archive: src/content/archive/works/paradise-lost.js Β· Paradise Lost Β· John Milton.

Of that forbidden Tree, whose mortal taste Brought death into the world, and all our woe, With loss of Eden

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence -0.48 (0.18 inferred) -0.44 (0.56 inferred) -0.31 (0.55 inferred) -0.27 (0.55 inferred) -0.12 (0.55 inferred) -0.98 (0.35 inferred)
arousal 0.28 (0.18 inferred) 0.17 (0.56 inferred) 0.58 (0.55 inferred) 0.79 (0.55 inferred) 0.49 (0.55 inferred) 0.75 (0.35 inferred)
dominance -0.17 (0.18 inferred) -0.12 (0.56 inferred) -0.36 (0.55 inferred) -0.48 (0.55 inferred) -0.36 (0.55 inferred) -0.26 (0.35 inferred)
tension β€” 0.15 (0.56 derived) β€” β€” β€” β€”
intimacy β€” 0.02 (0.56 inferred) β€” β€” β€” β€”
warmth β€” -0.13 (0.56 inferred) β€” β€” β€” β€”
uncertainty β€” β€” β€” β€” β€” β€”
expansiveness β€” 0.08 (0.56 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.55 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.04 (0.56 inferred) β€” β€” β€” β€”
solemnity β€” 0.12 (0.56 inferred) β€” β€” β€” β€”
novelty β€” 0.45 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.082 ms, mean confidence 0.18, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.121 ms, mean confidence 0.53, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 128.600 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 306.644 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.024 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 4091.281 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs minilm-l6-probe differ by 0.36 (-0.48, -0.12).
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.49 (-0.48, -0.98).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.54 (-0.44, -0.98).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.67 (-0.31, -0.98).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.71 (-0.27, -0.98).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.86 (-0.12, -0.98).
  • Disagreement on arousal: lexical-vad-v1 vs reward-emopair differ by 0.51 (0.28, 0.79).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.47 (0.28, 0.75).
  • Disagreement on arousal: contextual-window-v1 vs pert-emopair differ by 0.42 (0.17, 0.58).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.62 (0.17, 0.79).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.58 (0.17, 0.75).
  • Disagreement on dominance: contextual-window-v1 vs reward-emopair differ by 0.36 (-0.12, -0.48).
  • Teacher hosted-teacher: ok.

comedy-dark

Origin: rise-archive. Cases: poetry, solemn. Archive: src/content/archive/works/the-divine-comedy.js Β· The Divine Comedy Β· Dante Alighieri.

Midway upon the journey of our life I found myself within a forest dark, For the straightforward pathway had been lost.

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence -0.21 (0.17 inferred) -0.18 (0.54 inferred) -0.10 (0.55 inferred) -0.11 (0.55 inferred) 0.02 (0.55 inferred) 0.50 (0.35 inferred)
arousal 0.23 (0.17 inferred) 0.13 (0.54 inferred) 0.43 (0.55 inferred) 0.59 (0.55 inferred) 0.40 (0.55 inferred) 0.70 (0.35 inferred)
dominance -0.11 (0.17 inferred) -0.06 (0.54 inferred) -0.59 (0.55 inferred) -0.79 (0.55 inferred) -0.17 (0.55 inferred) 0.30 (0.35 inferred)
tension β€” 0.10 (0.54 derived) β€” β€” β€” β€”
intimacy β€” 0.05 (0.54 inferred) β€” β€” β€” β€”
warmth β€” -0.05 (0.54 inferred) β€” β€” β€” β€”
uncertainty β€” β€” β€” β€” β€” β€”
expansiveness β€” 0.07 (0.54 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.52 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.03 (0.54 inferred) β€” β€” β€” β€”
solemnity β€” 0.02 (0.54 inferred) β€” β€” β€” β€”
novelty β€” 0.45 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.088 ms, mean confidence 0.17, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.142 ms, mean confidence 0.51, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 129.815 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 298.279 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.024 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 4023.499 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.71 (-0.21, 0.50).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.68 (-0.18, 0.50).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.60 (-0.10, 0.50).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.61 (-0.11, 0.50).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.48 (0.02, 0.50).
  • Disagreement on arousal: lexical-vad-v1 vs reward-emopair differ by 0.36 (0.23, 0.59).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.47 (0.23, 0.70).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.46 (0.13, 0.59).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.57 (0.13, 0.70).
  • Disagreement on dominance: lexical-vad-v1 vs pert-emopair differ by 0.48 (-0.11, -0.59).
  • Disagreement on dominance: lexical-vad-v1 vs reward-emopair differ by 0.68 (-0.11, -0.79).
  • Disagreement on dominance: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.41 (-0.11, 0.30).
  • Disagreement on dominance: contextual-window-v1 vs pert-emopair differ by 0.53 (-0.06, -0.59).
  • Disagreement on dominance: contextual-window-v1 vs reward-emopair differ by 0.73 (-0.06, -0.79).
  • Disagreement on dominance: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.36 (-0.06, 0.30).
  • Disagreement on dominance: pert-emopair vs minilm-l6-probe differ by 0.42 (-0.59, -0.17).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.89 (-0.59, 0.30).
  • Disagreement on dominance: reward-emopair vs minilm-l6-probe differ by 0.62 (-0.79, -0.17).
  • Disagreement on dominance: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.09 (-0.79, 0.30).
  • Disagreement on dominance: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.47 (-0.17, 0.30).
  • Teacher hosted-teacher: ok.

blake-tyger

Origin: rise-archive. Cases: poetry, difficult. Archive: src/content/archive/works/literary-poems-blake.js Β· Songs of Innocence and of Experience Β· William Blake.

burning bright In the forests of the night, What immortal hand or eye Could frame thy fearful symmetry?

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence -0.09 (0.21 inferred) -0.07 (0.68 inferred) -0.02 (0.55 inferred) -0.07 (0.55 inferred) -0.11 (0.55 inferred) 0.50 (0.35 inferred)
arousal 0.40 (0.21 inferred) 0.24 (0.68 inferred) 0.57 (0.55 inferred) 0.81 (0.55 inferred) 0.41 (0.55 inferred) 0.90 (0.35 inferred)
dominance 0.02 (0.21 inferred) 0.01 (0.68 inferred) -0.47 (0.55 inferred) -0.30 (0.55 inferred) -0.32 (0.55 inferred) -0.20 (0.35 inferred)
tension β€” 0.16 (0.68 derived) β€” β€” β€” β€”
intimacy β€” β€” β€” β€” β€” β€”
warmth β€” -0.02 (0.68 inferred) β€” β€” β€” β€”
uncertainty β€” 0.40 (0.68 inferred) β€” β€” β€” β€”
expansiveness β€” 0.11 (0.68 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.57 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.09 (0.68 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.45 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.135 ms, mean confidence 0.21, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.131 ms, mean confidence 0.62, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 123.114 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 304.250 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.025 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 4115.356 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.59 (-0.09, 0.50).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.57 (-0.07, 0.50).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.52 (-0.02, 0.50).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.57 (-0.07, 0.50).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.61 (-0.11, 0.50).
  • Disagreement on arousal: lexical-vad-v1 vs reward-emopair differ by 0.41 (0.40, 0.81).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.50 (0.40, 0.90).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.57 (0.24, 0.81).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.66 (0.24, 0.90).
  • Disagreement on arousal: reward-emopair vs minilm-l6-probe differ by 0.40 (0.81, 0.41).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.49 (0.41, 0.90).
  • Disagreement on dominance: lexical-vad-v1 vs pert-emopair differ by 0.49 (0.02, -0.47).
  • Disagreement on dominance: contextual-window-v1 vs pert-emopair differ by 0.49 (0.01, -0.47).
  • Teacher hosted-teacher: ok.

whitman-celebrate

Origin: rise-archive. Cases: poetry, intimacy. Archive: src/content/archive/works/literary-leaves-of-grass.js Β· Leaves of Grass Β· Walt Whitman.

I celebrate myself, and sing myself, And what I assume you shall assume, For every atom belonging to me as good belongs to you.

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.54 (0.15 inferred) 0.49 (0.43 inferred) 0.34 (0.55 inferred) 0.31 (0.55 inferred) -0.04 (0.55 inferred) 0.99 (0.35 inferred)
arousal 0.39 (0.15 inferred) 0.23 (0.43 inferred) 0.63 (0.55 inferred) 0.86 (0.55 inferred) 0.38 (0.55 inferred) 0.96 (0.35 inferred)
dominance 0.25 (0.15 inferred) 0.22 (0.43 inferred) 0.31 (0.55 inferred) 0.68 (0.55 inferred) 0.14 (0.55 inferred) 0.93 (0.35 inferred)
tension β€” 0.12 (0.43 derived) β€” β€” β€” β€”
intimacy β€” 0.13 (0.43 inferred) β€” β€” β€” β€”
warmth β€” 0.39 (0.43 inferred) β€” β€” β€” β€”
uncertainty β€” β€” β€” β€” β€” β€”
expansiveness β€” 0.06 (0.43 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.43 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.06 (0.43 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.36 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.082 ms, mean confidence 0.15, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.099 ms, mean confidence 0.42, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 166.464 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 369.357 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.027 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 5266.762 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs minilm-l6-probe differ by 0.58 (0.54, -0.04).
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.45 (0.54, 0.99).
  • Disagreement on valence: contextual-window-v1 vs minilm-l6-probe differ by 0.53 (0.49, -0.04).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.50 (0.49, 0.99).
  • Disagreement on valence: pert-emopair vs minilm-l6-probe differ by 0.38 (0.34, -0.04).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.64 (0.34, 0.99).
  • Disagreement on valence: reward-emopair vs minilm-l6-probe differ by 0.35 (0.31, -0.04).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.67 (0.31, 0.99).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.03 (-0.04, 0.99).
  • Disagreement on arousal: lexical-vad-v1 vs reward-emopair differ by 0.46 (0.39, 0.86).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.57 (0.39, 0.96).
  • Disagreement on arousal: contextual-window-v1 vs pert-emopair differ by 0.40 (0.23, 0.63).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.63 (0.23, 0.86).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.73 (0.23, 0.96).
  • Disagreement on arousal: reward-emopair vs minilm-l6-probe differ by 0.48 (0.86, 0.38).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.59 (0.38, 0.96).
  • Disagreement on dominance: lexical-vad-v1 vs reward-emopair differ by 0.43 (0.25, 0.68).
  • Disagreement on dominance: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.68 (0.25, 0.93).
  • Disagreement on dominance: contextual-window-v1 vs reward-emopair differ by 0.46 (0.22, 0.68).
  • Disagreement on dominance: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.71 (0.22, 0.93).
  • Disagreement on dominance: pert-emopair vs reward-emopair differ by 0.37 (0.31, 0.68).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.62 (0.31, 0.93).
  • Disagreement on dominance: reward-emopair vs minilm-l6-probe differ by 0.54 (0.68, 0.14).
  • Disagreement on dominance: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.79 (0.14, 0.93).
  • Teacher hosted-teacher: ok.

heights-souls

Origin: rise-archive. Cases: prose, dialogue, intimacy. Archive: src/content/archive/works/wuthering-heights.js Β· Wuthering Heights Β· Emily BrontΓ«.

Whatever our souls are made of, his and mine are the same

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.15 (0.14 inferred) 0.12 (0.26 inferred) 0.09 (0.55 inferred) 0.13 (0.55 inferred) -0.00 (0.55 inferred) 0.50 (0.35 inferred)
arousal 0.25 (0.14 inferred) 0.15 (0.26 inferred) 0.33 (0.55 inferred) 0.50 (0.55 inferred) 0.53 (0.55 inferred) 0.75 (0.35 inferred)
dominance 0.05 (0.14 inferred) 0.06 (0.26 inferred) 0.26 (0.55 inferred) 0.07 (0.55 inferred) 0.08 (0.55 inferred) 0.25 (0.35 inferred)
tension β€” 0.10 (0.26 derived) β€” β€” β€” β€”
intimacy β€” 0.10 (0.26 inferred) β€” β€” β€” β€”
warmth β€” 0.03 (0.26 inferred) β€” β€” β€” β€”
uncertainty β€” β€” β€” β€” β€” β€”
expansiveness β€” 0.03 (0.26 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.50 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.04 (0.26 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.45 (0.26 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.055 ms, mean confidence 0.14, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.073 ms, mean confidence 0.28, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped, low-lexicon-coverage.
  • pert-emopair: 152.083 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 302.058 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.019 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 4074.121 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.35 (0.15, 0.50).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.38 (0.12, 0.50).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.41 (0.09, 0.50).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.37 (0.13, 0.50).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.50 (-0.00, 0.50).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.50 (0.25, 0.75).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.36 (0.15, 0.50).
  • Disagreement on arousal: contextual-window-v1 vs minilm-l6-probe differ by 0.38 (0.15, 0.53).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.60 (0.15, 0.75).
  • Disagreement on arousal: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.42 (0.33, 0.75).
  • Teacher hosted-teacher: ok.

spoon-hill

Origin: rise-archive. Cases: poetry, solemn. Archive: src/content/archive/works/spoon-river-anthology.js Β· Spoon River Anthology Β· Edgar Lee Masters.

Where are Elmer, Herman, Bert, Tom and Charley, The weak of will, the strong of arm, the clown, the boozer, the fighter? All, all are sleeping on the hill.

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence -0.04 (0.18 inferred) -0.04 (0.64 inferred) -0.02 (0.55 inferred) -0.06 (0.55 inferred) -0.15 (0.55 inferred) 0.58 (0.35 inferred)
arousal 0.31 (0.18 inferred) 0.21 (0.64 inferred) 0.43 (0.55 inferred) 0.83 (0.55 inferred) 0.46 (0.55 inferred) 0.96 (0.35 inferred)
dominance 0.07 (0.18 inferred) 0.05 (0.64 inferred) 0.08 (0.55 inferred) 0.01 (0.55 inferred) -0.11 (0.55 inferred) 0.73 (0.35 inferred)
tension β€” 0.13 (0.64 derived) β€” β€” β€” β€”
intimacy β€” β€” β€” β€” β€” β€”
warmth β€” -0.01 (0.64 inferred) β€” β€” β€” β€”
uncertainty β€” 0.20 (0.64 inferred) β€” β€” β€” β€”
expansiveness β€” 0.04 (0.64 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.60 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.06 (0.64 inferred) β€” β€” β€” β€”
solemnity β€” 0.02 (0.64 inferred) β€” β€” β€” β€”
novelty β€” 0.45 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.088 ms, mean confidence 0.18, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.117 ms, mean confidence 0.59, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 184.379 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 300.626 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.018 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 4655.335 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.62 (-0.04, 0.58).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.62 (-0.04, 0.58).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.60 (-0.02, 0.58).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.64 (-0.06, 0.58).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.73 (-0.15, 0.58).
  • Disagreement on arousal: lexical-vad-v1 vs reward-emopair differ by 0.52 (0.31, 0.83).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.66 (0.31, 0.96).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.62 (0.21, 0.83).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.76 (0.21, 0.96).
  • Disagreement on arousal: pert-emopair vs reward-emopair differ by 0.39 (0.43, 0.83).
  • Disagreement on arousal: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.53 (0.43, 0.96).
  • Disagreement on arousal: reward-emopair vs minilm-l6-probe differ by 0.36 (0.83, 0.46).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.50 (0.46, 0.96).
  • Disagreement on dominance: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.66 (0.07, 0.73).
  • Disagreement on dominance: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.67 (0.05, 0.73).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.65 (0.08, 0.73).
  • Disagreement on dominance: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.72 (0.01, 0.73).
  • Disagreement on dominance: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.84 (-0.11, 0.73).
  • Teacher hosted-teacher: ok.

tao-unnamed

Origin: rise-archive. Cases: scripture, ambiguity, difficult. Archive: src/content/archive/works/sacred-tao-te-ching.js Β· Tao Te Ching Β· Laozi.

The Tao that can be trodden is not the enduring and unchanging Tao. The name that can be named is not the enduring and unchanging name.

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.03 (0.29 inferred) -0.01 (0.70 inferred) -0.03 (0.55 inferred) -0.02 (0.55 inferred) -0.02 (0.55 inferred) 0.00 (0.35 inferred)
arousal 0.09 (0.29 inferred) 0.05 (0.70 inferred) 0.27 (0.55 inferred) 0.34 (0.55 inferred) 0.36 (0.55 inferred) 0.00 (0.35 inferred)
dominance 0.09 (0.29 inferred) -0.01 (0.70 inferred) 0.44 (0.55 inferred) 0.54 (0.55 inferred) -0.11 (0.55 inferred) 0.00 (0.35 inferred)
tension β€” 0.08 (0.70 derived) β€” β€” β€” β€”
intimacy β€” β€” β€” β€” β€” β€”
warmth β€” -0.00 (0.70 inferred) β€” β€” β€” β€”
uncertainty β€” 0.05 (0.70 inferred) β€” β€” β€” β€”
expansiveness β€” 0.09 (0.70 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.41 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.01 (0.70 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.26 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.113 ms, mean confidence 0.29, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.094 ms, mean confidence 0.64, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 170.478 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 296.255 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.016 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 3294.693 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.36 (0.36, 0.00).
  • Disagreement on dominance: lexical-vad-v1 vs reward-emopair differ by 0.45 (0.09, 0.54).
  • Disagreement on dominance: contextual-window-v1 vs pert-emopair differ by 0.44 (-0.01, 0.44).
  • Disagreement on dominance: contextual-window-v1 vs reward-emopair differ by 0.55 (-0.01, 0.54).
  • Disagreement on dominance: pert-emopair vs minilm-l6-probe differ by 0.55 (0.44, -0.11).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.44 (0.44, 0.00).
  • Disagreement on dominance: reward-emopair vs minilm-l6-probe differ by 0.65 (0.54, -0.11).
  • Disagreement on dominance: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.54 (0.54, 0.00).
  • Teacher hosted-teacher: ok.

hamlet-question

Origin: rise-archive. Cases: dialogue, difficult, uncertainty. Archive: src/content/archive/works/hamlet.js Β· Hamlet Β· William Shakespeare.

nobler in the mind to suffer The slings and arrows of outrageous fortune, Or to take arms against a sea of troubles, And by opposing end them

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence -0.15 (0.26 inferred) -0.12 (0.70 inferred) -0.09 (0.55 inferred) -0.06 (0.55 inferred) 0.08 (0.55 inferred) 0.58 (0.35 inferred)
arousal 0.39 (0.26 inferred) 0.22 (0.70 inferred) 0.54 (0.55 inferred) 0.78 (0.55 inferred) 0.52 (0.55 inferred) 0.96 (0.35 inferred)
dominance 0.06 (0.26 inferred) 0.05 (0.70 inferred) -0.16 (0.55 inferred) -0.01 (0.55 inferred) 0.03 (0.55 inferred) 0.73 (0.35 inferred)
tension β€” 0.16 (0.70 derived) β€” β€” β€” β€”
intimacy β€” β€” β€” β€” β€” β€”
warmth β€” -0.03 (0.70 inferred) β€” β€” β€” β€”
uncertainty β€” β€” β€” β€” β€” β€”
expansiveness β€” 0.10 (0.70 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.60 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.06 (0.70 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.45 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.062 ms, mean confidence 0.26, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.111 ms, mean confidence 0.63, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 163.368 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 293.276 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.013 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 4678.041 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.73 (-0.15, 0.58).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.70 (-0.12, 0.58).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.67 (-0.09, 0.58).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.64 (-0.06, 0.58).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.50 (0.08, 0.58).
  • Disagreement on arousal: lexical-vad-v1 vs reward-emopair differ by 0.38 (0.39, 0.78).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.57 (0.39, 0.96).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.55 (0.22, 0.78).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.74 (0.22, 0.96).
  • Disagreement on arousal: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.42 (0.54, 0.96).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.45 (0.52, 0.96).
  • Disagreement on dominance: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.67 (0.06, 0.73).
  • Disagreement on dominance: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.68 (0.05, 0.73).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.89 (-0.16, 0.73).
  • Disagreement on dominance: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.73 (-0.01, 0.73).
  • Disagreement on dominance: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.69 (0.03, 0.73).
  • Teacher hosted-teacher: ok.

crime-question

Origin: rise-archive. Cases: prose, tension. Archive: src/content/archive/works/crime-and-punishment.js Β· Crime and Punishment Β· Fyodor Dostoevsky.

fearful, frenzied and fantastic question, which tortured his heart and mind, clamouring insistently for an answer.

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence -0.21 (0.33 inferred) -0.11 (0.70 inferred) -0.14 (0.55 inferred) -0.22 (0.55 inferred) -0.04 (0.55 inferred) -0.67 (0.35 inferred)
arousal 0.51 (0.33 inferred) 0.31 (0.70 inferred) 0.82 (0.55 inferred) 0.93 (0.55 inferred) 0.42 (0.55 inferred) 0.83 (0.35 inferred)
dominance -0.07 (0.33 inferred) -0.07 (0.70 inferred) -0.56 (0.55 inferred) -0.92 (0.55 inferred) -0.06 (0.55 inferred) -0.50 (0.35 inferred)
tension β€” 0.18 (0.70 derived) β€” β€” β€” β€”
intimacy β€” 0.07 (0.70 inferred) β€” β€” β€” β€”
warmth β€” 0.22 (0.70 inferred) β€” β€” β€” β€”
uncertainty β€” 0.03 (0.70 inferred) β€” β€” β€” β€”
expansiveness β€” 0.04 (0.70 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.68 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.10 (0.70 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.45 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.057 ms, mean confidence 0.33, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.082 ms, mean confidence 0.65, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 154.455 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 287.970 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.012 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 4229.099 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.46 (-0.21, -0.67).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.56 (-0.11, -0.67).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.53 (-0.14, -0.67).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.45 (-0.22, -0.67).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.63 (-0.04, -0.67).
  • Disagreement on arousal: lexical-vad-v1 vs reward-emopair differ by 0.41 (0.51, 0.93).
  • Disagreement on arousal: contextual-window-v1 vs pert-emopair differ by 0.51 (0.31, 0.82).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.62 (0.31, 0.93).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.52 (0.31, 0.83).
  • Disagreement on arousal: pert-emopair vs minilm-l6-probe differ by 0.40 (0.82, 0.42).
  • Disagreement on arousal: reward-emopair vs minilm-l6-probe differ by 0.51 (0.93, 0.42).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.41 (0.42, 0.83).
  • Disagreement on dominance: lexical-vad-v1 vs pert-emopair differ by 0.48 (-0.07, -0.56).
  • Disagreement on dominance: lexical-vad-v1 vs reward-emopair differ by 0.84 (-0.07, -0.92).
  • Disagreement on dominance: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.43 (-0.07, -0.50).
  • Disagreement on dominance: contextual-window-v1 vs pert-emopair differ by 0.49 (-0.07, -0.56).
  • Disagreement on dominance: contextual-window-v1 vs reward-emopair differ by 0.85 (-0.07, -0.92).
  • Disagreement on dominance: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.43 (-0.07, -0.50).
  • Disagreement on dominance: pert-emopair vs reward-emopair differ by 0.36 (-0.56, -0.92).
  • Disagreement on dominance: pert-emopair vs minilm-l6-probe differ by 0.50 (-0.56, -0.06).
  • Disagreement on dominance: reward-emopair vs minilm-l6-probe differ by 0.86 (-0.92, -0.06).
  • Disagreement on dominance: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.42 (-0.92, -0.50).
  • Disagreement on dominance: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.44 (-0.06, -0.50).
  • Teacher hosted-teacher: ok.

doll-tricks

Origin: rise-archive. Cases: dialogue. Archive: src/content/archive/works/a-doll-s-house.js Β· A Doll's House Β· Henrik Ibsen.

I have existed merely to perform tricks for you, Torvald. But you would have it so.

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence -0.07 (0.17 inferred) -0.06 (0.38 inferred) 0.03 (0.55 inferred) -0.12 (0.55 inferred) 0.05 (0.55 inferred) -0.50 (0.35 inferred)
arousal 0.32 (0.17 inferred) 0.20 (0.38 inferred) 0.48 (0.55 inferred) 0.69 (0.55 inferred) 0.43 (0.55 inferred) 0.70 (0.35 inferred)
dominance -0.05 (0.17 inferred) -0.03 (0.38 inferred) 0.07 (0.55 inferred) -0.02 (0.55 inferred) -0.09 (0.55 inferred) -0.30 (0.35 inferred)
tension β€” 0.17 (0.38 derived) β€” β€” β€” β€”
intimacy β€” 0.02 (0.38 inferred) β€” β€” β€” β€”
warmth β€” -0.01 (0.38 inferred) β€” β€” β€” β€”
uncertainty β€” 0.03 (0.38 inferred) β€” β€” β€” β€”
expansiveness β€” 0.02 (0.38 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.52 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.05 (0.38 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.49 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.050 ms, mean confidence 0.17, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.068 ms, mean confidence 0.38, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped, contrast-marker-present.
  • pert-emopair: 152.365 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 289.856 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.017 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 3462.307 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.42 (-0.07, -0.50).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.44 (-0.06, -0.50).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.53 (0.03, -0.50).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.38 (-0.12, -0.50).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.55 (0.05, -0.50).
  • Disagreement on arousal: lexical-vad-v1 vs reward-emopair differ by 0.36 (0.32, 0.69).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.38 (0.32, 0.70).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.49 (0.20, 0.69).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.50 (0.20, 0.70).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.37 (0.07, -0.30).
  • Teacher hosted-teacher: ok.

epictetus-power

Origin: rise-archive. Cases: philosophy. Archive: src/content/archive/works/epictetus-encheiridion.js Β· Encheiridion Β· Epictetus.

you will find not one which is capable of contemplating itself

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.08 (0.20 inferred) 0.06 (0.42 inferred) -0.04 (0.55 inferred) -0.07 (0.55 inferred) -0.06 (0.55 inferred) -1.00 (0.35 inferred)
arousal 0.20 (0.20 inferred) 0.11 (0.42 inferred) 0.16 (0.55 inferred) 0.50 (0.55 inferred) 0.40 (0.55 inferred) 0.50 (0.35 inferred)
dominance 0.28 (0.20 inferred) 0.21 (0.42 inferred) 0.28 (0.55 inferred) 0.58 (0.55 inferred) -0.12 (0.55 inferred) -1.00 (0.35 inferred)
tension β€” 0.08 (0.42 derived) β€” β€” β€” β€”
intimacy β€” β€” β€” β€” β€” β€”
warmth β€” 0.02 (0.42 inferred) β€” β€” β€” β€”
uncertainty β€” β€” β€” β€” β€” β€”
expansiveness β€” 0.03 (0.42 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.53 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.03 (0.42 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.45 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.050 ms, mean confidence 0.20, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.064 ms, mean confidence 0.42, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 120.997 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 292.927 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.013 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 2907.553 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.07 (0.08, -1.00).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.06 (0.06, -1.00).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.96 (-0.04, -1.00).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.93 (-0.07, -1.00).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.94 (-0.06, -1.00).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.39 (0.11, 0.50).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.39 (0.11, 0.50).
  • Disagreement on dominance: lexical-vad-v1 vs minilm-l6-probe differ by 0.39 (0.28, -0.12).
  • Disagreement on dominance: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.27 (0.28, -1.00).
  • Disagreement on dominance: contextual-window-v1 vs reward-emopair differ by 0.37 (0.21, 0.58).
  • Disagreement on dominance: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.21 (0.21, -1.00).
  • Disagreement on dominance: pert-emopair vs minilm-l6-probe differ by 0.40 (0.28, -0.12).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.28 (0.28, -1.00).
  • Disagreement on dominance: reward-emopair vs minilm-l6-probe differ by 0.70 (0.58, -0.12).
  • Disagreement on dominance: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.58 (0.58, -1.00).
  • Disagreement on dominance: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.88 (-0.12, -1.00).
  • Teacher hosted-teacher: ok.

analects-learn

Origin: rise-archive. Cases: philosophy. Archive: src/content/archive/works/confucius-analects.js Β· Analects Β· Confucius.

Is it not pleasant to learn with a constant perseverance and application?

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.33 (0.30 inferred) -0.18 (0.70 inferred) 0.23 (0.55 inferred) 0.22 (0.55 inferred) -0.08 (0.55 inferred) 0.90 (0.35 inferred)
arousal 0.26 (0.30 inferred) 0.15 (0.70 inferred) 0.40 (0.55 inferred) 0.54 (0.55 inferred) 0.28 (0.55 inferred) 0.85 (0.35 inferred)
dominance 0.21 (0.30 inferred) 0.05 (0.70 inferred) -0.04 (0.55 inferred) 0.01 (0.55 inferred) 0.01 (0.55 inferred) 0.70 (0.35 inferred)
tension β€” 0.22 (0.70 derived) β€” β€” β€” β€”
intimacy β€” 0.00 (0.70 inferred) β€” β€” β€” β€”
warmth β€” -0.33 (0.70 inferred) β€” β€” β€” β€”
uncertainty β€” 0.45 (0.70 inferred) β€” β€” β€” β€”
expansiveness β€” 0.03 (0.70 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.63 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.04 (0.70 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.45 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.070 ms, mean confidence 0.30, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.069 ms, mean confidence 0.65, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 118.306 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 292.897 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.013 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 3928.177 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs contextual-window-v1 differ by 0.51 (0.33, -0.18).
  • Disagreement on valence: lexical-vad-v1 vs minilm-l6-probe differ by 0.41 (0.33, -0.08).
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.57 (0.33, 0.90).
  • Disagreement on valence: contextual-window-v1 vs pert-emopair differ by 0.41 (-0.18, 0.23).
  • Disagreement on valence: contextual-window-v1 vs reward-emopair differ by 0.40 (-0.18, 0.22).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.08 (-0.18, 0.90).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.67 (0.23, 0.90).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.68 (0.22, 0.90).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.98 (-0.08, 0.90).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.59 (0.26, 0.85).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.39 (0.15, 0.54).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.70 (0.15, 0.85).
  • Disagreement on arousal: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.45 (0.40, 0.85).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.57 (0.28, 0.85).
  • Disagreement on dominance: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.49 (0.21, 0.70).
  • Disagreement on dominance: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.65 (0.05, 0.70).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.74 (-0.04, 0.70).
  • Disagreement on dominance: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.69 (0.01, 0.70).
  • Disagreement on dominance: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.69 (0.01, 0.70).
  • Teacher hosted-teacher: ok.

vitruvius-order

Origin: rise-archive. Cases: neutral, descriptive. Archive: src/content/archive/works/vitruvius-architecture.js Β· On Architecture Β· Vitruvius.

Architecture depends on Order (in Greek [Greek: taxis]), Arrangement (in Greek [Greek: diathesis]), Eurythmy, Symmetry, Propriety, and Economy

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.06 (0.21 inferred) 0.05 (0.67 inferred) 0.02 (0.55 inferred) 0.05 (0.55 inferred) 0.05 (0.55 inferred) 0.50 (0.35 inferred)
arousal 0.10 (0.21 inferred) 0.10 (0.67 inferred) 0.08 (0.55 inferred) 0.22 (0.55 inferred) 0.16 (0.55 inferred) 0.70 (0.35 inferred)
dominance 0.17 (0.21 inferred) 0.13 (0.67 inferred) 0.09 (0.55 inferred) 0.26 (0.55 inferred) 0.02 (0.55 inferred) 0.30 (0.35 inferred)
tension β€” 0.04 (0.67 derived) β€” β€” β€” β€”
intimacy β€” β€” β€” β€” β€” β€”
warmth β€” 0.01 (0.67 inferred) β€” β€” β€” β€”
uncertainty β€” β€” β€” β€” β€” β€”
expansiveness β€” 0.05 (0.67 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.54 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.02 (0.67 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.35 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.056 ms, mean confidence 0.21, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.077 ms, mean confidence 0.61, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 148.283 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 299.830 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.012 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 3806.834 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.43 (0.06, 0.50).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.45 (0.05, 0.50).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.48 (0.02, 0.50).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.45 (0.05, 0.50).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.45 (0.05, 0.50).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.60 (0.10, 0.70).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.60 (0.10, 0.70).
  • Disagreement on arousal: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.62 (0.08, 0.70).
  • Disagreement on arousal: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.48 (0.22, 0.70).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.54 (0.16, 0.70).
  • Teacher hosted-teacher: ok.

dow-balance

Origin: rise-archive. Cases: neutral, descriptive. Archive: src/content/archive/works/dow-composition.js Β· Composition Β· Arthur Wesley Dow.

balance of proportions, tone and color. A change in one member changes the whole.

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.10 (0.23 inferred) 0.08 (0.61 inferred) 0.03 (0.55 inferred) 0.03 (0.55 inferred) 0.09 (0.55 inferred) 0.50 (0.35 inferred)
arousal 0.14 (0.23 inferred) 0.09 (0.61 inferred) 0.11 (0.55 inferred) 0.21 (0.55 inferred) 0.33 (0.55 inferred) 0.75 (0.35 inferred)
dominance 0.05 (0.23 inferred) 0.04 (0.61 inferred) 0.13 (0.55 inferred) 0.17 (0.55 inferred) 0.07 (0.55 inferred) 0.25 (0.35 inferred)
tension β€” 0.06 (0.61 derived) β€” β€” β€” β€”
intimacy β€” β€” β€” β€” β€” β€”
warmth β€” 0.02 (0.61 inferred) β€” β€” β€” β€”
uncertainty β€” β€” β€” β€” β€” β€”
expansiveness β€” 0.02 (0.61 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.57 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.02 (0.61 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.45 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.049 ms, mean confidence 0.23, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.066 ms, mean confidence 0.57, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 124.909 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 287.799 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.012 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 4221.982 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.40 (0.10, 0.50).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.42 (0.08, 0.50).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.47 (0.03, 0.50).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.47 (0.03, 0.50).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.41 (0.09, 0.50).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.61 (0.14, 0.75).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.66 (0.09, 0.75).
  • Disagreement on arousal: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.64 (0.11, 0.75).
  • Disagreement on arousal: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.54 (0.21, 0.75).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.42 (0.33, 0.75).
  • Teacher hosted-teacher: ok.

gita-universe

Origin: rise-archive. Cases: scripture, scale. Archive: src/content/archive/works/extended-bhagavad-gita-full.js Β· Bhagavad Gita Β· attributed to Vyasa.

By Me the whole vast Universe of things Is spread abroad;--by Me, the Unmanifest!

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.08 (0.18 inferred) 0.07 (0.48 inferred) 0.15 (0.55 inferred) 0.11 (0.55 inferred) -0.01 (0.55 inferred) 0.98 (0.35 inferred)
arousal 0.28 (0.18 inferred) 0.42 (0.48 inferred) 0.66 (0.55 inferred) 0.80 (0.55 inferred) 0.45 (0.55 inferred) 0.98 (0.35 inferred)
dominance 0.08 (0.18 inferred) 0.07 (0.48 inferred) 0.34 (0.55 inferred) 0.69 (0.55 inferred) 0.04 (0.55 inferred) 0.98 (0.35 inferred)
tension β€” 0.12 (0.48 derived) β€” β€” β€” β€”
intimacy β€” 0.04 (0.48 inferred) β€” β€” β€” β€”
warmth β€” 0.02 (0.48 inferred) β€” β€” β€” β€”
uncertainty β€” 0.02 (0.48 inferred) β€” β€” β€” β€”
expansiveness β€” 0.26 (0.48 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.51 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.24 (0.48 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.40 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.053 ms, mean confidence 0.18, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.070 ms, mean confidence 0.46, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 132.621 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 288.536 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.011 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 4577.984 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.90 (0.08, 0.98).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.92 (0.07, 0.98).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.84 (0.15, 0.98).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.87 (0.11, 0.98).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.99 (-0.01, 0.98).
  • Disagreement on arousal: lexical-vad-v1 vs pert-emopair differ by 0.38 (0.28, 0.66).
  • Disagreement on arousal: lexical-vad-v1 vs reward-emopair differ by 0.52 (0.28, 0.80).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.69 (0.28, 0.98).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.38 (0.42, 0.80).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.55 (0.42, 0.98).
  • Disagreement on arousal: reward-emopair vs minilm-l6-probe differ by 0.36 (0.80, 0.45).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.53 (0.45, 0.98).
  • Disagreement on dominance: lexical-vad-v1 vs reward-emopair differ by 0.61 (0.08, 0.69).
  • Disagreement on dominance: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.90 (0.08, 0.98).
  • Disagreement on dominance: contextual-window-v1 vs reward-emopair differ by 0.62 (0.07, 0.69).
  • Disagreement on dominance: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.91 (0.07, 0.98).
  • Disagreement on dominance: pert-emopair vs reward-emopair differ by 0.35 (0.34, 0.69).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.65 (0.34, 0.98).
  • Disagreement on dominance: reward-emopair vs minilm-l6-probe differ by 0.65 (0.69, 0.04).
  • Disagreement on dominance: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.94 (0.04, 0.98).
  • Teacher hosted-teacher: ok.

faustus-wrath

Origin: rise-archive. Cases: dialogue, tension. Archive: src/content/archive/works/the-tragical-history-of-doctor-faustus.js Β· Doctor Faustus Β· Christopher Marlowe.

heap God's heavy wrath upon thy head!

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence -0.70 (0.14 inferred) -0.56 (0.25 inferred) -0.20 (0.55 inferred) -0.29 (0.55 inferred) -0.03 (0.55 inferred) -1.00 (0.35 inferred)
arousal 0.86 (0.14 inferred) 0.74 (0.25 inferred) 0.78 (0.55 inferred) 1.00 (0.55 inferred) 0.53 (0.55 inferred) 1.00 (0.35 inferred)
dominance 0.35 (0.14 inferred) 0.26 (0.25 inferred) 0.45 (0.55 inferred) 0.88 (0.55 inferred) -0.05 (0.55 inferred) -1.00 (0.35 inferred)
tension β€” 0.34 (0.25 derived) β€” β€” β€” β€”
intimacy β€” β€” β€” β€” β€” β€”
warmth β€” -0.14 (0.25 inferred) β€” β€” β€” β€”
uncertainty β€” β€” β€” β€” β€” β€”
expansiveness β€” 0.02 (0.25 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.51 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.33 (0.25 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.45 (0.25 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.041 ms, mean confidence 0.14, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.062 ms, mean confidence 0.27, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped, low-lexicon-coverage.
  • pert-emopair: 117.970 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 291.032 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.014 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 2599.970 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs pert-emopair differ by 0.50 (-0.70, -0.20).
  • Disagreement on valence: lexical-vad-v1 vs reward-emopair differ by 0.41 (-0.70, -0.29).
  • Disagreement on valence: lexical-vad-v1 vs minilm-l6-probe differ by 0.67 (-0.70, -0.03).
  • Disagreement on valence: contextual-window-v1 vs pert-emopair differ by 0.36 (-0.56, -0.20).
  • Disagreement on valence: contextual-window-v1 vs minilm-l6-probe differ by 0.53 (-0.56, -0.03).
  • Disagreement on valence: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.44 (-0.56, -1.00).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.80 (-0.20, -1.00).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.71 (-0.29, -1.00).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.97 (-0.03, -1.00).
  • Disagreement on arousal: reward-emopair vs minilm-l6-probe differ by 0.47 (1.00, 0.53).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.47 (0.53, 1.00).
  • Disagreement on dominance: lexical-vad-v1 vs reward-emopair differ by 0.53 (0.35, 0.88).
  • Disagreement on dominance: lexical-vad-v1 vs minilm-l6-probe differ by 0.40 (0.35, -0.05).
  • Disagreement on dominance: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.35 (0.35, -1.00).
  • Disagreement on dominance: contextual-window-v1 vs reward-emopair differ by 0.62 (0.26, 0.88).
  • Disagreement on dominance: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.26 (0.26, -1.00).
  • Disagreement on dominance: pert-emopair vs reward-emopair differ by 0.43 (0.45, 0.88).
  • Disagreement on dominance: pert-emopair vs minilm-l6-probe differ by 0.50 (0.45, -0.05).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.45 (0.45, -1.00).
  • Disagreement on dominance: reward-emopair vs minilm-l6-probe differ by 0.93 (0.88, -0.05).
  • Disagreement on dominance: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.88 (0.88, -1.00).
  • Disagreement on dominance: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.95 (-0.05, -1.00).
  • Teacher hosted-teacher: ok.

koan-difference

Origin: rise-archive. Cases: scripture, ambiguity, difficult. Archive: src/content/archive/works/sacred-zen-koans.js Β· Zen koans Β· collected.

Is there any difference between the teaching of the Patriarch and that of the Sutras, or not?

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.05 (0.18 inferred) 0.04 (0.40 inferred) 0.01 (0.55 inferred) 0.04 (0.55 inferred) 0.06 (0.55 inferred) 0.25 (0.35 inferred)
arousal 0.17 (0.18 inferred) 0.11 (0.40 inferred) 0.16 (0.55 inferred) 0.34 (0.55 inferred) 0.18 (0.55 inferred) 0.75 (0.35 inferred)
dominance 0.07 (0.18 inferred) 0.05 (0.40 inferred) -0.30 (0.55 inferred) -0.18 (0.55 inferred) 0.12 (0.55 inferred) 0.25 (0.35 inferred)
tension β€” 0.08 (0.40 derived) β€” β€” β€” β€”
intimacy β€” β€” β€” β€” β€” β€”
warmth β€” 0.01 (0.40 inferred) β€” β€” β€” β€”
uncertainty β€” 0.42 (0.40 inferred) β€” β€” β€” β€”
expansiveness β€” 0.04 (0.40 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.54 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.03 (0.40 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.45 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.053 ms, mean confidence 0.18, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.067 ms, mean confidence 0.40, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 125.873 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 282.717 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.012 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 4260.005 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.57 (0.17, 0.75).
  • Disagreement on arousal: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.64 (0.11, 0.75).
  • Disagreement on arousal: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.59 (0.16, 0.75).
  • Disagreement on arousal: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.41 (0.34, 0.75).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.57 (0.18, 0.75).
  • Disagreement on dominance: lexical-vad-v1 vs pert-emopair differ by 0.38 (0.07, -0.30).
  • Disagreement on dominance: pert-emopair vs minilm-l6-probe differ by 0.42 (-0.30, 0.12).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.55 (-0.30, 0.25).
  • Disagreement on dominance: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.43 (-0.18, 0.25).
  • Teacher hosted-teacher: ok.

probe-negation

Origin: constructed-probe. Cases: negation.

I am not happy.

dimension lexical-vad-v1 contextual-window-v1 pert-emopair reward-emopair minilm-l6-probe Qwen/Qwen2.5-1.5B-Instruct
valence 0.82 (0.23 inferred) -0.76 (0.38 inferred) -0.62 (0.55 inferred) -0.31 (0.55 inferred) 0.09 (0.55 inferred) -1.00 (0.35 inferred)
arousal 0.42 (0.23 inferred) 0.23 (0.38 inferred) 0.43 (0.55 inferred) 0.64 (0.55 inferred) 0.42 (0.55 inferred) 0.00 (0.35 inferred)
dominance 0.25 (0.23 inferred) -0.15 (0.38 inferred) -0.33 (0.55 inferred) -0.53 (0.55 inferred) 0.23 (0.55 inferred) -1.00 (0.35 inferred)
tension β€” 0.39 (0.38 derived) β€” β€” β€” β€”
intimacy β€” 0.00 (0.38 inferred) β€” β€” β€” β€”
warmth β€” -0.69 (0.38 inferred) β€” β€” β€” β€”
uncertainty β€” 0.10 (0.38 inferred) β€” β€” β€” β€”
expansiveness β€” 0.01 (0.38 inferred) β€” β€” β€” β€”
perceptualDensity β€” 0.55 (0.45 derived) β€” β€” β€” β€”
motionEnergy β€” 0.06 (0.38 inferred) β€” β€” β€” β€”
solemnity β€” β€” β€” β€” β€” β€”
novelty β€” 0.45 (0.35 derived) β€” β€” β€” β€”
  • lexical-vad-v1: 0.036 ms, mean confidence 0.23, parameters 0, trained false. Caveats: weak-baseline, negation-ignored, bag-of-words.
  • contextual-window-v1: 0.055 ms, mean confidence 0.38, parameters 40, trained false. Caveats: unfitted-linear-prior, confidence-capped.
  • pert-emopair: 110.237 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, pert-valence-emobank-arousal-operating-range, not-a-rise-human-norm.
  • reward-emopair: 280.070 ms, mean confidence 0.55, parameters 355000000, trained true. Caveats: offline-checkpoint, vad-only, reward-valence-emobank-ad-published-norm, not-a-rise-human-norm.
  • minilm-l6-probe: 0.012 ms, mean confidence 0.55, parameters 22700000, trained true. Caveats: leave-one-out-prediction, distilled-from-pert-emopair, vad-only, twenty-four-passages-leave-one-out.
  • Qwen/Qwen2.5-1.5B-Instruct: 3114.247 ms, mean confidence 0.35, parameters 1500000000, trained false. Caveats: language-model-teacher, local-cpu-no-hosted-credential.
  • Disagreement on valence: lexical-vad-v1 vs contextual-window-v1 differ by 1.58 (0.82, -0.76).
  • Disagreement on valence: lexical-vad-v1 vs pert-emopair differ by 1.44 (0.82, -0.62).
  • Disagreement on valence: lexical-vad-v1 vs reward-emopair differ by 1.13 (0.82, -0.31).
  • Disagreement on valence: lexical-vad-v1 vs minilm-l6-probe differ by 0.73 (0.82, 0.09).
  • Disagreement on valence: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.82 (0.82, -1.00).
  • Disagreement on valence: contextual-window-v1 vs reward-emopair differ by 0.45 (-0.76, -0.31).
  • Disagreement on valence: contextual-window-v1 vs minilm-l6-probe differ by 0.85 (-0.76, 0.09).
  • Disagreement on valence: pert-emopair vs minilm-l6-probe differ by 0.71 (-0.62, 0.09).
  • Disagreement on valence: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.38 (-0.62, -1.00).
  • Disagreement on valence: reward-emopair vs minilm-l6-probe differ by 0.41 (-0.31, 0.09).
  • Disagreement on valence: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.69 (-0.31, -1.00).
  • Disagreement on valence: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.09 (0.09, -1.00).
  • Disagreement on arousal: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.42 (0.42, 0.00).
  • Disagreement on arousal: contextual-window-v1 vs reward-emopair differ by 0.40 (0.23, 0.64).
  • Disagreement on arousal: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.43 (0.43, 0.00).
  • Disagreement on arousal: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.64 (0.64, 0.00).
  • Disagreement on arousal: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.42 (0.42, 0.00).
  • Disagreement on dominance: lexical-vad-v1 vs contextual-window-v1 differ by 0.40 (0.25, -0.15).
  • Disagreement on dominance: lexical-vad-v1 vs pert-emopair differ by 0.58 (0.25, -0.33).
  • Disagreement on dominance: lexical-vad-v1 vs reward-emopair differ by 0.78 (0.25, -0.53).
  • Disagreement on dominance: lexical-vad-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.25 (0.25, -1.00).
  • Disagreement on dominance: contextual-window-v1 vs reward-emopair differ by 0.38 (-0.15, -0.53).
  • Disagreement on dominance: contextual-window-v1 vs minilm-l6-probe differ by 0.38 (-0.15, 0.23).
  • Disagreement on dominance: contextual-window-v1 vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.85 (-0.15, -1.00).
  • Disagreement on dominance: pert-emopair vs minilm-l6-probe differ by 0.56 (-0.33, 0.23).
  • Disagreement on dominance: pert-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.67 (-0.33, -1.00).
  • Disagreement on dominance: reward-emopair vs minilm-l6-probe differ by 0.76 (-0.53, 0.23).
  • Disagreement on dominance: reward-emopair vs Qwen/Qwen2.5-1.5B-Instruct differ by 0.47 (-0.53, -1.00).
  • Disagreement on dominance: minilm-l6-probe vs Qwen/Qwen2.5-1.5B-Instruct differ by 1.23 (0.23, -1.00).
  • Teacher hosted-teacher: ok.

Models not executed

Watched failure classes

Irony, ambiguity, and difficult literary cases are listed with their outputs above. A high confidence on those passages would be a claim this encoder is not entitled to make. The contextual model caps confidence at 0.7 because its readout is an unfitted prior.

Footer

Passages: 24. Recorded disagreements: 386.

Published from docs/affect/BENCHMARK.md. Edits made here are overwritten on the next push to main.