aiqbench analysis
Current statistical methods for detecting benchmark contamination fail in realistic auditing scenarios due to distribution shifts and scale constraints, resulting in a significant reliability gap where only 59% of evaluations yield correct outcomes. The study demonstrates that existing paradigms like LLM Dataset Inference and Post-Hoc Dataset Inference suffer from false positives and underpowered detection, respectively, proving that statistical signals cannot yet replace transparent data provenance. This work highlights the urgent need for robust validation frameworks that account for non-IID assumptions and the massive disparity between benchmark sizes and pre-training corpora.
The Reliability Gap in Benchmark Auditing: Distribution Shift and Scale as Failure Modes of Contamination Detection
arXiv:2606.03305v1 Announce Type: new Abstract: Benchmark contamination, where evaluation examples appear in a model's training data, threatens the validity of LLM assessment. Statistical tools for detecting training-data membership exist, but have been validated almost exclusively in controlled academic regimes: large, homogeneous pre-training corpora and transparent, single-stage training pipelines. Whether these methods remain reliable in realistic auditing scenarios remains unclear. We identify two under-studied failure modes: distribution shift, which arises when suspect and validation sets violate the IID assumption, and scale constraints, which arise because benchmarks are orders of magnitude smaller than pre-training corpora. We systematically evaluate three leading paradigms: LLM Dataset Inference, Post-Hoc Dataset Inference, and CoDeC across 27 models from multiple families (including Pythia, OLMo~2, and specialised cultural and medical LLMs) and scales (up to 27B). We then further extend our analysis to frontier industry models. Across 335 evaluations, only 199 yield correct outcomes. LLM Dataset Inference results in false positives under distribution shift, Post-Hoc Dataset Inference is underpowered at benchmark scale, and CoDeC provides only coarse provenance signals that are insufficient to verify individual benchmark splits. Our results reveal a systematic reliability gap between controlled validation and practical benchmark auditing, and show that statistical detection cannot yet replace transparent data provenance. We open-source our benchmark for further research.
Abstract:Benchmark contamination, where evaluation examples appear in a model's training data, threatens the validity of LLM assessment. Statistical tools for detecting training-data membership exist, but have been validated almost exclusively in controlled academic regimes: large, homogeneous pre-training corpora and transparent, single-stage training pipelines. Whether these methods remain reliable in realistic auditing scenarios remains unclear. We identify two under-studied failure modes: distribution shift, which arises when suspect and validation sets violate the IID assumption, and scale constraints, which arise because benchmarks are orders of magnitude smaller than pre-training corpora. We systematically evaluate three leading paradigms: LLM Dataset Inference, Post-Hoc Dataset Inference, and CoDeC across 27 models from multiple families (including Pythia, OLMo~2, and specialised cultural and medical LLMs) and scales (up to 27B). We then further extend our analysis to frontier industry models. Across 335 evaluations, only 199 yield correct outcomes. LLM Dataset Inference results in false positives under distribution shift, Post-Hoc Dataset Inference is underpowered at benchmark scale, and CoDeC provides only coarse provenance signals that are insufficient to verify individual benchmark splits. Our results reveal a systematic reliability gap between controlled validation and practical benchmark auditing, and show that statistical detection cannot yet replace transparent data provenance. We open-source our benchmark for further research.
Artificial Intelligence (cs.AI)
arXiv:2606.03305 [cs.AI]
(or arXiv:2606.03305v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2606.03305 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 299 words
Related coverage
[AINews] Meta Muse Wonder & Wonder Full: MSL’s Claude 5.5-tier Frontier Model
Latent Space · Oct 13, 2026
[AINews] High Taste Testers Ep. 1
Latent Space · Oct 17, 2026
Presentation: Building Evals for AI Adoption: from Principles to Practice
InfoQ · May 29, 2026
[AINews] AI Engineer NYC: Gemini 4 Argon, Vals AI Benchmarks
Latent Space · Oct 14, 2026