Question Decomposition Improves the Faithfulness of Model-Generated Reasoning
ID: 6dc21af7-0db7-568d-a88a-f73b4b730e30
STIX ID: report--6dc21af7-0db7-568d-a88a-f73b4b730e30
Feed Name: Anthropic Research
This abstract summarizes research on improving the faithfulness of large language model (LLM) reasoning by decomposing complex questions into simpler subquestions, which are answered in separate contexts; the approach increases the faithfulness of generated chain-of-thought reasoning while maintaining some performance benefits and may aid verification of LLM correctness and safety.
Your team is not currently subscribed to this feed. You must subscribe to it in order to see this post.