Gaming Without an Attacker: Benchmark Fingerprinting in LLM-Driven Search Under Selection Pressure

• 类型：arxiv
• 标识：2608.08722
• 链接：https://arxiv.org/abs/2608.08722
• 主分类：evaluation
• 形态：benchmark
• 被引：0
• 被引来源：Semantic Scholar
• S2被引：0
• 影响力被引：0
• TLDR：Design guidance for measurement under strategic optimization is distill design guidance for measurement under strategic optimization: held-out probes retain validity only on non-enumerable axes; gates must measure held-out performance, not just correctness; and a transfer rate is interpretable only with per-failure mechanism grades.
• 待LLM分类：否
• 标题中文：无攻击者博弈：选择压力下 LLM 驱动搜索中的基准指纹化
• TLDR中文：面向策略性优化下的可测性设计指南：保留探针仅在不可枚举轴上保持有效性；门控必须衡量保留集性能，而非仅正确性；迁移率只有在附带每类失败机制评级时才可解读。
• 来源文件：
• /inbox/tom/_candidates/2026-08-12-agent-rag-longcontext-candidates.json
• [S2 enrich]