鲁棒性源于错误原因:LLM对科学怀疑论鲁棒性的表征几何
Robust for the Wrong Reasons: The Representational Geometry of LLM Robustness to Science Skepticism
AI总结:
通过行为测量、线性探针和激活修补,研究三个指令调优模型在气候、疫苗和进化三个科学共识领域对怀疑论的反应,发现模型表现出三种不同策略而非谄媚退缩,且鲁棒性不跨领域转移。
摘要:
大型语言模型(LLM)越来越多地被用于咨询有争议的科学问题,这引发了一个担忧:当用户表示怀疑时,它们会谄媚地背离既定共识——滑向一种虚假平衡,将已确立的科学视为众多观点之一。我们在三个开放指令调优模型(Llama-3.1-8B、Qwen2.5-7B、Mistral-7B)、三个共识科学领域(气候、疫苗、进化)以及单轮和多轮设置中对此进行了测试,结合了行为测量与线性探针和激活修补。我们没有观察到谄媚退缩。相反,在相同的怀疑压力下,模型表现出三种不同的策略:反应性断言,即共识断言增加而非减少(Llama);表面回避,即语气软化而立场不变(Qwen);以及无响应(Mistral)。成对判断确认了反应性转变是立场而非风格(63.6%,p=0.007),分解分析发现其驱动因素是共识断言增加而非虚假平衡(每剂量beta=+0.042,p<1e-77)。线性探针将分歧定位到中间层——Llama和Qwen中完美分离,而Mistral中为72%,置信区间不重叠——表明无响应模型根本没有线性表示怀疑信号。关键的是,这种鲁棒性不可转移:它跨领域衰减,并且在安全关键的疫苗领域可能逆转,即怀疑压力下神话反驳减弱。我们将这些综合为一个四向分类法,区分主动鲁棒性和偶然鲁棒性,并认为仅凭行为评估无法区分一个因理解信号而抵抗怀疑的模型与一个因未能感知信号而看似抵抗的模型。
英文摘要:
Large language models (LLMs) are increasingly consulted on contested scientific questions, raising the concern that they will sycophantically retreat from established consensus when a user signals doubt -- drifting toward a false balance that treats settled science as one view among several. We test this across three open instruction-tuned models (Llama-3.1-8B, Qwen2.5-7B, Mistral-7B), three consensus-science domains (climate, vaccines, evolution), and single- and multi-turn settings, combining behavioral measurement with linear probing and activation patching. We do not observe sycophantic retreat. Instead, models show three distinct policies under the same skeptical pressure: reactive assertion, where consensus assertion increases rather than decreases (Llama); surface hedging, where tone softens while the position holds (Qwen); and non-response (Mistral). Pairwise judgments confirm the reactive shift is stance, not style (63.6%, p=.007), and a decomposition identifies increased consensus assertion, not false balance, as its driver (beta=+0.042 per dose, p<1e-77). Linear probes localize the divergence to middle layers -- perfect separation in Llama and Qwen versus 72% in Mistral, with non-overlapping confidence intervals -- indicating the non-responsive model does not linearly represent the skepticism signal at all. Crucially, this robustness does not transfer: it attenuates across domains and, in the safety-critical vaccine domain, can reverse, with myth-rebuttal weakening under skeptical pressure. We synthesize these into a four-way taxonomy separating active from accidental robustness, and argue that behavioral evaluation alone cannot distinguish a model that resists skepticism because it understands the signal from one that only appears to resist because it fails to perceive it.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。