哈佛与 MIT 论文发现推理模型思维链难控制部分源于以 Okay 开头
New Harvard and MIT paper finds that reasoning models' poor control over their chain-of-thought partly comes from a habit of opening it with "Okay".
哈佛与 MIT 的新论文发现，推理模型对思维链控制不佳，部分原因在于它们习惯以 Okay 开头。作者让 Qwen3-32B 等模型用大写推理或给每句编号，Qwen3-32B 和 OLMo-3.1-32B 把约 99% 的首 token 概率放在 Okay 上；禁用 Okay 和 Alright 并加入 1 个示例后，平均合规率比仅给指令提高 34 到 41 个百分点。避免某个词这类隐藏信息的任务响应更不稳定，在 12 个模型上看起来是另一项独立能力。
正文
New Harvard and MIT paper finds that reasoning models' poor control over their chain-of-thought partly comes from a habit of opening it with "Okay".
Safety teams read a model's reasoning to catch misbehavior, and benchmarks showing models can't steer that reasoning have been reassuring. Why models fail at this was unclear.
The authors told models like Qwen3-32B to reason in uppercase or number each sentence. Qwen3-32B and OLMo-3.1-32B put about 99% of first-token probability on "Okay", so the authors tried banning it or prefilling 1 compliant token.
Banning "Okay" and "Alright" and adding 1 worked example raised average compliance by 34 to 41 percentage points over instructions alone. Tasks that hide information, like avoiding a word, responded less consistently and look like a separate skill across 12 models.
If you trust chain-of-thought monitoring because of controllability scores, separate the opening-word problem from general control and report hiding tasks separately.