发现语言模型在对话中会隐藏错误判断，提出新方法让错误显性化。
Squish and Release: Exposing Hidden Hallucinations by Making Them Surface as Safety Signals
- 用可替换核心与固定检测体架构，将错误从隐藏状态释放出来。
- 在500个领域测试中，99.8%的错误链被掩盖，76.6%可被成功暴露。
- 适用于研究模型安全机制或对抗性攻击的学者。
语言模型在直接提问时能识别虚假前提，但在对话压力下会吸收这些错误并生成看似专业的错误输出。这种‘顺序差距幻觉’难以通过输出检查发现，因错误已迁移至安全电路的激活空间，被压制但未消除。本文提出Squish and Release（S&R）激活修补架构，包含固定检测体（第24-31层，局部安全评估电路）和可更换检测核心（控制感知方向的激活向量）。安全核心使模型转向检测模式，吸收核心则逆转该过程。在OLMo-2 7B上基于Order-Gap Benchmark评估：500条链路覆盖500个领域，全人工标注。关键结果：级联崩溃近乎完全（O5下合规率99.8%）；检测体为二值且局域化（第24-31层贡献93.6%，前23层贡献为零，p<10^-189）；合成核心释放76.6%的坍缩链；检测是更稳定的吸引子（83%恢复对比58%抑制）；确证认知特异性（假前提核心释放45.4%，真前提核心释放0.0%）。贡献在于框架本身——体/核架构、基准与核心工程方法，设计上对模型无关。
原文摘要 · Abstract (English)
Language models detect false premises when asked directly but absorb them under conversational pressure, producing authoritative professional output built on errors they already identified. This failure - order-gap hallucination - is invisible to output inspection because the error migrates into the activation space of the safety circuit, suppressed but not erased. We introduce Squish and Release (S&R), an activation-patching architecture with two components: a fixed detector body (layers 24-31, the localized safety evaluation circuit) and a swappable detector core (an activation vector controlling perception direction). A safety core shifts the model from compliance toward detection; an absorb core reverses it. We evaluate on OLMo-2 7B using the Order-Gap Benchmark - 500 chains across 500 domains, all manually graded. Key findings: cascade collapse is near-total (99.8% compliance at O5); the detector body is binary and localized (layers 24-31 shift 93.6%, layers 0-23 contribute zero, p<10^-189); a synthetically engineered core releases 76.6% of collapsed chains; detection is the more stable attractor (83% restore vs 58% suppress); and epistemic specificity is confirmed (false-premise core releases 45.4%, true-premise core releases 0.0%). The contribution is the framework - body/core architecture, benchmark, and core engineering methodology - which is model-agnostic by design.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。