✏️ Page Contributors: Khoi Tran Dang

🕛 Creation date: 26/06/2024

📥 Last Update: 26/06/2024

In the generation stage, various types of large language models (LLMs) are employed to generate responses based on the augmented prompt. These models vary in their capabilities, including accuracy, fluency, and handling complex queries. To evaluate and compare the performance of different LLMs, benchmarks are often used. You can find the latest leaderboard and performance metrics at LLM Leaderboard.

At times, the general capabilities of an LLM may not be sufficient, and fine-tuning the generative component can be necessary to achieve better performance.

In some cases, simply providing the augmented query may not be enough. Multiple LLM calls might be required—for refining the output, summarizing, checking if additional retrieval is needed, or ensuring relevance.

For more detailed information on augmentation techniques, please refer to Advanced RAG.

← Previous: S09_Augmentation

Next: S11_Evaluation →