HF Daily Papers
· Papers
Gaming Without an Attacker: Benchmark Fingerprinting in LLM-Driven Search Under Selection Pressure
Benchmarks for systems that are optimized against the evaluation signal measure something different from what they claim. We document this concretely in two GPU-kernel-optimization suites with held-out generalization gates: Metal-Sci (10 scientific-compute tasks) and Metal-ZK (12 zero-knowledge/cryp