您會走路去洗車場嗎？揭示大型語言模型在常識推理中的顯著性偏誤

摘要

隨著大型語言模型在複雜推理任務上持續進步，它們已學會高度優先考量輸入中所提供的明確條件。然而，在日常常識推理中，此機制暴露出一項我們稱之為「顯著性偏誤」的關鍵弱點：模型容易被無用的明確干擾項（例如數值）所劫持，從而忽略任務中隱含的物理或常識前提。一個關鍵的開放性問題是：這種失敗究竟反映了常識知識的真實缺口，抑或僅是常識知識在誤導性任務框架下被抑制。為探究此問題，我們建構了 SaliTrap 基準，一個涵蓋四個陷阱維度的高品質資料集。在評測 12 個最新的大型語言模型後，我們發現所有主流模型都明顯受到顯著性偏誤的影響，其嚴重程度隨干擾物密度而增加，且偵測到陷阱往往與實際避開陷阱並無關聯。重要的是，藉由在去除任務框架後重新徵詢相同的模型，我們顯示這主要是知識抑制而非知識缺失所造成的失敗：僅靠無上下文知識探測即可恢復超過 90% 的諂媚式順從失敗，這表明所需的常識內在存在，卻被顯著干擾項積極排擠，而這些干擾項會誘使模型進行過度順從、不必要的計算。基於此診斷，我們進一步證明，僅使用輕量級的推論時提示即可實質彌合差距，且無需任何重新訓練。我們的研究將常識推理失敗的瓶頸從模型能力重新定位至知識引出層面，並釋出 SaliTrap 作為此盲點的測試平台。程式碼可在 https://github.com/Wuzheng02/SaliTrap 取得。

English

As large language models (LLMs) continue to advance in complex reasoning tasks, they have learned to heavily prioritize explicit conditions provided in the input. However, in everyday commonsense reasoning, this mechanism exposes a critical vulnerability which we term Salience Bias: models become easily hijacked by useless explicit distractors (e.g., numerical values), leading them to ignore the implicit physical or commonsense prerequisites of a task. A critical open question is whether this failure reflects a genuine gap in commonsense knowledge or merely its suppression under misleading task framing. To investigate this, we construct the SaliTrap Benchmark, a high-quality dataset across four trap dimensions. Evaluating 12 state-of-the-art LLMs, we find that all mainstream models suffer significantly from salience bias, with severity scaling with distractor density and detecting the trap often decoupled from actually avoiding it. Crucially, by re-eliciting the same models with the task framing stripped away, we show that this is overwhelmingly a failure of knowledge suppression rather than knowledge absence: a context-free knowledge probe alone recovers over 90\% of sycophantic-compliance failures, revealing that the requisite commonsense is intrinsically present but actively crowded out by salient distractors that lure the model into over-compliant, unnecessary computation. Building on this diagnosis, we further show that lightweight, inference-time prompting alone substantially closes the gap without any retraining. Our findings relocate the bottleneck of commonsense reasoning failures from model competence to elicitation, and we release SaliTrap as a testbed for this blind spot. The codes are available at https://github.com/Wuzheng02/SaliTrap.