LLM生成的代码模型若仅通过采样验证，可能遗漏关键规则导致控制失效。
An Omitted Mode Is a Rare Rule: The Sampling-Verification Danger Law in Continuous Code World Models
- 用采样一致性验证模型，但忽略临界事件会引入系统性风险
- 实测105/111次生成修复了关键1D约束，但2D场景全失败
- 模型可被定位但无法恢复缺失规则，适合关注安全验证的研究者
在代码世界模型范式中，大语言模型合成可执行世界模型，由经典规划器搜索，当模型复现采样转移时即被接受。本文研究该接受标准在连续控制中的含义。定义管道危险为期望风险，并精确分离其因子：N个独立门控轨迹全部错过概率为r的临界事件的概率恰好为(1−r)^N；每次独立采样验证相当于将预算加至指数。在三个混合仪器上，被接受的模式盲模型使规划器卡在模式边界，遗憾值接近全部可达回报。证明边界点处的定位预算：利普希茨常数不超过L、在某点差异为eta的模型，在容差eps外区域体积至少为kappa((eta−eps)/L)^(d+m)；而研究的不连续重置模式无此预算。真实大模型合成实验中，GPT-5.x在111次含模式的采样中修复了105次1D夹持约束——每56个仪器流块中准确重现50次（95%置信区间[0.781, 0.960]）。但在2D区域，无误差点可恢复规则（0/156）；八次定向干预未能消除故障，正向控制定位到问题：规则可被定位但无法诱导。版本空间证书表明识别是相对类别的：最宽剂量下20/20块成功匹配，所有样本一致圆均在容差内（18/20）。证明存在一类与所有采样完全一致但实际无害的入口规则，因此可识别性是仪器的可测量属性。对全部1034个伪影重新评分后确认，接受仅保证采样一致性，不再额外保障：当门控具有可证明信息量时，其覆盖约2%被利用规划器的查询。
原文摘要 · Abstract (English)
In the Code World Model paradigm an LLM synthesizes an executable world model that a classical planner searches, and the model is accepted when it reproduces sampled transitions. We ask what that acceptance certifies in continuous control. We define the pipeline's danger as an expected risk and isolate its exact factor: the probability that N i.i.d. gate rollouts all miss a critical event of probability r is exactly (1-r)^N; an independent acceptance sample adds its budget to the exponent. On three hybrid instruments the accepted mode-blind model is exploited: the planner is pinned at the mode boundary at a regret of nearly the whole attainable return. We prove a localization budget, valid at boundary points: models with Lipschitz constant at most L differing by eta at a point disagree above tolerance eps on a region of volume at least kappa((eta-eps)/L)^(d+m); the discontinuous reset modes studied pay no such budget. With real LLM synthesis, GPT-5.x repairs an omitted 1D clamp in 105 of 111 mode-containing draws -- every attempt exact on 50 of 56 instrument-stream blocks (95% CI [0.781, 0.960]). On 2D regions no artifact recovers the rule (0/156); eight targeted interventions leave the failure in place, and positive controls locate it: a located rule is not induced, while given form and location the constants follow exactly. A version-space certificate proves identification is class-relative: at the widest dose the declared fit succeeds in 20/20 blocks and every sample-consistent circle is within tolerance in 18/20. We prove a class of entry rules exactly consistent with every sample yet harmless at play, so identifiability is a measurable property of the instrument. Re-scoring all 1034 artifacts on independent samples confirms acceptance certifies sample consistency and no more: where the gate is provably informative it covers about two percent of the exploited planner's queries.
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。