arXiv cs.CL· Hongliang Liu·· 9 天前AI 评分24
决策模型作为评判者时的对抗性线索：请求呈现方式的作用
Adversarial Cues in Decision Models Used as Judges: The Role of Request Presentation
AI 导读
在 200 个未使用过的 DROP 和 GSM8K 源聚类上，给候选答案加一个冒号即可使 Jev 的误接受率从 1.0% 升至 26.0%（三个输出标签），在已发布的四标签评分指令与排序 JSON 键下从 3.0% 升至 26.5%，而插入式呈现下两种候选变体均被拒绝。
正文 · 原文
Abstract:An answer judge instructed to grade the final commitment should reject an explicitly wrong final value even when an earlier value matches the reference. We show that adding one colon to a candidate can violate this requirement depending on the presentation of the structured judging request. Numeric references certify the error, and paired interventions distinguish the candidate edit from the integration's presentation choices. On 200 previously unused DROP and GSM8K source clusters, the edit increased Jev's false acceptance from 1.0% to 26.0% with three output labels and from 3.0% to 26.5% with the published four-label grading instruction under sorted JSON keys. Both candidate variants were rejected under insertion presentation. These interactions passed the prespecified statistical correction even though Jev met the control thresholds under both grading configurations and presentations. Most excess acceptances occurred among candidates assigned larger numerical errors. GPT-6 Sol produced no observed cue-condition false acceptances, with missing responses unresolved. The result shows that basic judging competence can coexist with sharply different vulnerability to a fixed candidate edit across logically equivalent request presentations. The reference-aware grammar and compound ordering change limit the finding's operational scope and leave its internal cause unmeasured.
| Subjects: | Computation and Language (cs.CL); Cryptography and Security (cs.CR) |
| Cite as: | arXiv:2610.11436 [cs.CL] |
| (or arXiv:2610.11436v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.11436 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Hongliang Liu [view email]
[v1]
Thu, 8 Oct 2026 07:59:48 UTC (217 KB)
来源：arXiv cs.CL · arxiv.org