arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2604.28031cs.CLcs.AI
收藏

模型记住它们违反的内容:多轮LLM构思中的约束遵守

Models Recall What They Violate: Constraint Adherence in Multi-Turn LLM Ideation

Technical University of Munich(慕尼黑技术大学)

Garvin Kruthof

AI总结:

研究探讨多轮LLM构思中约束遵守问题,通过DriftBench基准发现迭代压力增加结构复杂性但降低约束遵守,模型在记忆与行为上存在脱节,约束违反率高达99%。

摘要:

当研究人员逐步完善想法时,大型语言模型是否保持对原始目标的忠实?我们引入DriftBench,用于评估多轮LLM辅助科学构思中的约束遵守。在2146次评分基准运行中,涵盖七种模型、四种交互条件和38个科学领域,发现迭代压力可靠增加结构复杂性并经常降低对原始约束的遵守。重述探测揭示了声明性记忆与行为遵守之间的脱节,模型准确重述了它们同时违反的约束。知道但违反(KBV)率,衡量尽管保留记忆但约束不合规的比率,在模型间从8%到99%不等。结构检查点部分降低了KBV率,但未消除脱节,复杂性膨胀持续存在。人类验证确认LLM判官低估了约束违规,使报告的约束遵守分数保守。敏感性分析确认结果在温度(0.7 vs. 1.0)和压力类型(新颖性 vs. 严谨性)方面稳健。我们发布所有简报、提示、评分标准、转录和评分作为开放基准。

英文摘要:

When researchers refine ideas with large language models over several turns, do the models preserve the original brief's constraints? We introduce DriftBench, a benchmark for constraint adherence in multi-turn LLM-assisted scientific ideation. It pairs 38 research briefs from 24 scientific domains, each with hard constraints and banned moves, with four interaction conditions and seven models from five providers (two open-weight), scored on 2,116 runs. Iterative pressure to improve a proposal raises structural complexity in every model and lowers adherence to the brief's hard constraints. A restatement probe shows that the constraints stay legible and reproducible on demand while the proposal departs from them. The knows-but-violates (KBV) rate, the share of pressure runs scored non-compliant on a hard constraint despite a correct restatement, exceeds 50% for most models. A crossed-judge design measures the judge offset. GPT-5.4 scores adherence 0.535 points lower than Opus 4.6 on the same runs. The design finds no judge-by-condition interaction on adherence, so the pressure effect on adherence is judge-stable while KBV levels and the bottom of the model ranking are not. Re-eliciting the brief at checkpoints lowers KBV for some models but not others, and complexity inflation persists. Blind human raters find more constraint violations than the LLM judge, so reported adherence is conservative. We release all briefs, prompts, rubrics, transcripts, and scores.

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。

↑