Public demo

Reasoning

A source-grounded theme page on chain-of-thought prompting: step-by-step reasoning, model scale, benchmark gains, and verification limits.

Theme overview

This theme focuses on reasoning as an elicited behavior in large language models. chain-of-thought-prompting shows that the way a task is framed can change whether a model exposes useful intermediate steps.

Takeaways

• Reasoning traces are useful because they create inspectable structure.
• The demo should connect reasoning to evidence: steps still need support from sources.
• This theme bridges writing support and verification, because a draft often needs explicit argumentative structure.

Included papers

• Chain-of-Thought Prompting Elicits Reasoning in Large Language Models

Evidence examples

• Chain-of-Thought Prompting Elicits Reasoning in Large Language Models: Combining rationale generation with few-shot prompting, this work enhances LLM reasoning by eliciting emergent, sequential thought processes, similar to Chain-of-Thought.

  This work explores how the reasoning ability of large language models can be unlocked by a simple method motivated by two ideas.

• Chain-of-Thought Prompting Elicits Reasoning in Large Language Models: Errors are categorized based on the changes needed to correct the chain of thought, aiming for future model improvement.

  We decided to categorize errors into what changes are needed to make the chain of thought correct, with the goal of elucidating how the model can be improved in the future.