Better Accuracies, Worse Reasoning: A Step-Level Audit of Medical Chain-of-Thought Distillation
For practitioners deploying distilled models in high-stakes medical QA, this work highlights that standard answer metrics can mask degraded reasoning quality, which is critical when traces are reused or audited.
The paper shows that chain-of-thought distillation in medical QA improves final answer accuracy (MedQA-USMLE SC@64 from 74.7% to 84.4%) but increases step-level error rate in reasoning traces (from 30.6% to 50.3%), revealing a trade-off between answer quality and trace factuality.
Chain-of-thought (CoT) distillation trains a smaller model to imitate a teacher's reasoning trace, but it is typically evaluated by final-answer metrics including accuracy. We ask whether gains in answer quality are accompanied by improvements in the trace. In medical QA, where short answer options can leave a richer clinical justification under-specified, a Qwen3-8B student distilled from a DeepSeek-V3-family teacher improves on MedQA-USMLE answer metrics (SC@64 74.7% to 84.4%; expected calibration error (ECE) 0.096 to 0.034). Yet under a Kimi-K2.6 style-blind LLM-judge audit, its error rate over non-abstained steps rises from 30.6% to 50.3%. In this primary medical setting, answer quality and trace factuality move in opposite directions. This before--after pattern persists across evaluators, teacher strengths, student scales and families, medical benchmarks, and style, segmentation, and answer-correctness controls. A 150-step blinded audit by a clinical expert reproduces the same ordering. Boundary checks narrow the scope of the claim: the risk appears when a compact answer under-constrains the rationale and a capable student can imitate expert-like form without reliably grounding each local claim. Standard answer metrics and aggregate hedging rates do not reveal the shift. When such traces are released or reused, answer-level metrics alone are insufficient.