Evaluating Chain-of-Thought Reasoning through Reusability and Verifiability
This addresses a blind spot in current evaluation methods for reasoning quality in AI systems, though it is incremental as it builds on existing CoT frameworks.
The paper tackled the problem of evaluating Chain-of-Thought reasoning in multi-agent IR pipelines by introducing reusability and verifiability as novel measures, revealing that these do not correlate with standard accuracy and that specialized models do not consistently outperform general-purpose LLMs in these metrics.
In multi-agent IR pipelines for tasks such as search and ranking, LLM-based agents exchange intermediate reasoning in terms of Chain-of-Thought (CoT) with each other. Current CoT evaluation narrowly focuses on target task accuracy. However, this metric fails to assess the quality or utility of the reasoning process itself. To address this limitation, we introduce two novel measures: reusability and verifiability. We decouple CoT generation from execution using a Thinker-Executor framework. Reusability measures how easily an Executor can reuse the Thinker's CoT. Verifiability measures how frequently an Executor can match the Thinker's answer using the CoT. We evaluated four Thinker models against a committee of ten Executor models across five benchmarks. Our results reveal that reusability and verifiability do not correlate with standard accuracy, exposing a blind spot in current accuracy-based leaderboards for reasoning capability. Surprisingly, we find that CoTs from specialized reasoning models are not consistently more reusable or verifiable than those from general-purpose LLMs like Llama and Gemma.