Skip to content
Search
paperAugust 2026Unreviewed

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

Víctor Gallego

View resourcePDF

Abstract

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/cryptographic tasks), in which three frontier LLMs (Opus 4.7, Gemini 3.1 Pro, GPT-5.5) propose Metal kernels inside a $(1{+}1)$ evolutionary loop with rich feedback. Although no model is prompted to act a

Categories

Benchmarks & Evaluation

Cite

@misc{gallego2026gaming,
  title = {{Gaming Without an Attacker: Benchmark Fingerprinting in LLM-Driven Search Under Selection Pressure}},
  author = {Víctor Gallego},
  year = {2026},
  month = aug,
  eprint = {2608.08722},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2608.08722}
}