Role‑specialized QA pipelines often pass rationales generated by a reasoner to a verifier, yet the actual benefit of this message remains unclear. We introduce a message‑intervention diagnostic that keeps the evidence and candidate answer fixed while varying only the rationale transmitted across the reasoner‑to‑verifier boundary. Experiments on 400 examples from MuSiQue, HotpotQA, and 2WikiMultiHopQA use DeepSeek as both generator and verifier. The findings show that faithful rationales add almost no answer accuracy compared to providing no rationale, whereas corrupted rationales strongly alter support judgments.
Under a blind verifier prompt, harmless paraphrases shift support by only 0‑2.5%, while corrupted rationales shift it by 10‑22%; an explicit rationale‑checking prompt amplifies this pattern to 34‑55%. Final answer changes are smaller (2‑30%), and only 2.9‑35.3% of corrupted support flips co‑occur with answer changes. Human audits reveal that 16 out of 42 valid corruptions are "corruption‑overtrust" cases, and blind humans reject or mark unclear 9/10 audited corrupted rationales that the model accepts. Cross‑model and task‑boundary checks demonstrate that the communication channel can be active, amplified, inert, or folded into the task label.
Consequently, rationale sharing should be evaluated as a verification‑message mechanism rather than merely a route to higher answer accuracy.
Review: This work clarifies the true role of rationales in QA systems through careful intervention experiments, urging designers to consider the reliability and potential failure modes of rationale transmission.