Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection
Asai et al., 2024, Self-RAG (The Twelfth International Conference on Learning Representations (ICLR 2024))
Subscriber deck — free during early access
1 cards in “3.2.1 Training the Critic Model” · whole deck
3.2.1 Training the Critic Model
- How is the critic model trained?GPT-4 was prompted with type-specific instructions and few-shot examples to label sampled input–output pairs with reflection tokens (4K–20K examples per type, agreeing well with humans); a Llama 2-7B critic was then trained on these labels and agreed with GPT-4 over 90% of the time on most token types.
- Asai et al., 2024, Self-RAG (The Twelfth International Conference on Learning Representations (ICLR 2024)), §3.2.1
Source
- Akari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil and Hannaneh Hajishirzi. 2024. Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection. In The Twelfth International Conference on Learning Representations (ICLR 2024). https://arxiv.org/abs/2310.11511