A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Unknown
Proposes conformal-calibrated rewards with procedural regeneration to mitigate benchmark contamination in scientific RLVR.
Unknown
A large-scale measurement of benchmark data contamination across 17 frontier language models, quantifying how evaluation data leaks into training corpora.
Unknown
This paper formalizes when benchmark contamination is statistically detectable, distinguishing clean benchmarks from low-power audits and proposing power-calibrated detection methods.
Unknown
Cultivar is a contrastive, locale-oriented translation benchmark that enables diagnosing benchmark contamination and quantifying true localisation robustness.
Floriane C. M. Braun
A nuisance-controlled residual-stream probing method for detecting benchmark contamination in language models.
Unknown
Proposes stratified per-question probability evaluation and step-wise mitigation to detect and reduce benchmark contamination in LLM pretraining.
Watchara Ruangsang, Patiyuth Pramkeaw
Cross-dataset evaluation of YOLOv8 for UAV fire and smoke detection reveals benchmark contamination and strong zero-shot transfer.
Unknown
The paper argues benchmark contamination is not a single failure but at least five distinct types, and proposes a taxonomy organized by which mitigation each type defeats.
Unknown
This paper identifies distribution shift and scale as failure modes of benchmark contamination detection, revealing a reliability gap in benchmark auditing.
Unknown
This paper proposes a hierarchical, session-isolated detection framework to identify benchmark contamination in LLM coding benchmarks, addressing solution leakage and test quality issues.
Unknown
This position paper argues that LLM benchmark datasets should be designed to be contamination-resistant, proposing criteria and methods to prevent data leakage from inflating model performance.
Unknown
MMLU-CF introduces a contamination-free multi-task language understanding benchmark to address benchmark contamination in LLM evaluation.