研究：历史案例的标签与事实对 LLM 判定方向相反
Precedent Labels and Facts Shift LLM Approval Decisions in Opposite Directions
三个主流模型做审批时都会被历史案例带偏：结论标签拉高批准率，案例事实反而压低。做 RAG 或判例检索的值得看看这两部分怎么分开测。
论文用退款/理赔类审批任务测试 DeepSeek V4.1 Flash、GPT-5.6 Luna 和 Claude Haiku 5.5。当历史案例与请求近乎相同（事实仅改动少量细节）时，把案例记录的结论从 DENY 改成 APPROVE，可使 DeepSeek V4.1 Flash 和 GPT-5.6 Luna 的批准率上升约 30 到 55 个百分点。反过来，只展示案例事实而不带结论时，事实支持批准的案例反而使这两个模型更少批准，作者推测是对比让请求显得更弱。在九条由真实政策改写而来的公开政策上，当历史案例属于另一个请求时，改动结论只带来 9 个百分点的变化，而事实影响仍有 22 个百分点。论文建议检索历史判例的系统把案例结论与案例事实两部分分开测试。
Precedent Labels and Facts Shift LLM Approval Decisions in Opposite Directions
When a language model decides refund or claim requests under written policies, the system can show it similar past cases. Each past case has two parts: its facts and the decision recorded for it. We separate the two. A model sees a policy, a request and three past cases, and we change either the recorded decisions or the facts, keeping the request fixed. We select requests that are close calls: deciding without past cases, DeepSeek V4.1 Flash approved each in some of four tries and denied it in others. When the past cases are near-copies of the request (its facts with a few details changed), switching their recorded decisions from DENY to APPROVE raises approval by about 30 to 55 percentage points in DeepSeek V4.1 Flash, GPT-5.6 Luna and Claude Haiku 5.5 on policies written for the study, and DeepSeek and Luna show the same on nine public policies paraphrased from real ones. The facts push the other way: shown without their recorded decisions, past cases whose facts support approval make these two models approve less often than past cases whose facts support denial. One reading is that the request looks weaker next to them. This push is not just a correction toward the right answer: in an additional DeepSeek analysis, it was of similar estimated size on requests that human annotators judged approvable and on those they judged deniable. In a separate DeepSeek run on the public policies, when the past cases are near-copies of a different request, switching their decisions moves approval by 9 points instead of 43, while facts that support approval still lower it by 22 points against facts that support denial. Systems that retrieve past decisions should test the two parts separately.
