Provable Joint Decontamination for Benchmarking Multiple Large Language Models

• URL: http://arxiv.org/abs/2605.21543v1
• Date: Wed, 20 May 2026 09:16:39 GMT
• Title: Provable Joint Decontamination for Benchmarking Multiple Large Language Models
• Abstract summary: We formalize benchmark decontamination as a joint selection problem and propose Joint Envelope Conformal Selection (JECS)<n>JECS computes per-model conformal p-values, aggregates them by the per-item maximum, and reconstructs a conservative envelope of the max-p null distribution.<n>Experiments across various models and benchmarks demonstrate that JECS achieves higher power than the max-p baseline while consistently maintaining the target GCR control.
• Score: 19.929035827959822
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Benchmark data contamination has become a central challenge in LLM evaluation: when evaluation examples appear in the training data of one or more audited models, reported performance can be inflated and cross-model comparisons become unreliable. A broad line of training-data detection work designs scores to quantify how strongly a model memorizes a given data point, but these score-based methods lack theoretical guarantees. Recent conformal approaches provide provable false-identification control for a single model; however, applying them separately to each model can produce model-specific benchmarks, undermining fair comparison across models. In this work, we formalize multi-model benchmark decontamination as a joint selection problem and propose Joint Envelope Conformal Selection (JECS), a conformal procedure that enables global contamination rate (GCR) control under stated assumptions. Specifically, JECS computes per-model conformal p-values, aggregates them by the per-item maximum, and reconstructs a conservative envelope of the max-p null distribution from right-tail observations above a data-driven threshold. By applying the adaptive Benjamini-Hochberg (BH) procedure to the envelope-rescaled values, we select a benchmark with provable GCR control. Extensive experiments across various models and benchmarks demonstrate that JECS achieves higher power than the max-p baseline while consistently maintaining the target GCR control.

Related papers

• DualEval: Joint Model-Item Calibration for Unified LLM Evaluation [67.99119926012686]
  DualEval is a latent model-item calibration framework that represents models and evaluation items in a shared space.<n>We apply DualEval across four domains: coding, math, miscellaneous domain-knowledge tasks, and generic everyday user queries.
  arXiv  Detail & Related papers  (2026-06-24T22:40:46Z)
• Post-Selection Distributional Model Evaluation [31.006941545235396]
  Post-selection distributional model evaluation (PS-DME)<n>PS-DME is a framework for statistically valid distributional model assessment after arbitrary data-dependent model pre-selection.<n>Results: PS-DME enables reliable comparison of candidate configurations across a range of reliability levels.
  arXiv  Detail & Related papers  (2026-03-24T10:51:30Z)
• IF-RewardBench: Benchmarking Judge Models for Instruction-Following Evaluation [85.56193980646981]
  We propose IF-RewardBench, a comprehensive meta-evaluation benchmark for instruction-following.<n>For each instruction, we construct a preference graph containing all pairwise preferences among multiple responses.<n>Experiments on IF-RewardBench reveal significant deficiencies in current judge models.
  arXiv  Detail & Related papers  (2026-03-05T02:21:17Z)
• Learnable Chernoff Baselines for Inference-Time Alignment [64.81256817158851]
  We introduce Learnable Chernoff Baselines as a method for efficiently and approximately sampling from exponentially tilted kernels.<n>We establish total-variation guarantees to the ideal aligned model, and demonstrate in both continuous and discrete diffusion settings that LCB sampling closely matches ideal rejection sampling.
  arXiv  Detail & Related papers  (2026-02-08T00:09:40Z)
• Model Correlation Detection via Random Selection Probing [62.093777777813756]
  Existing similarity-based methods require access to model parameters or produce scores without thresholds.<n>We introduce Random Selection Probing (RSP), a hypothesis-testing framework that formulates model correlation detection as a statistical test.<n>RSP produces rigorous p-values that quantify evidence of correlation.
  arXiv  Detail & Related papers  (2025-09-29T01:40:26Z)
• Score-based Generative Modeling for Conditional Independence Testing [35.0533359302886]
  We propose a novel CI testing method via score-based generative modeling, which achieves precise Type I error control and strong testing power.<n>We theoretically establish the error bound of conditional distributions modeled by score-based generative models and prove the validity of our CI tests.
  arXiv  Detail & Related papers  (2025-05-29T10:10:46Z)
• PaCoST: Paired Confidence Significance Testing for Benchmark Contamination Detection in Large Language Models [41.772263447213234]
  Large language models (LLMs) are known to be trained on vast amounts of data, which may unintentionally or intentionally include data from commonly used benchmarks.<n>This inclusion can lead to cheatingly high scores on model leaderboards, yet result in disappointing performance in real-world applications.<n>We introduce PaCoST, a Paired Confidence Significance Testing to effectively detect benchmark contamination in LLMs.
  arXiv  Detail & Related papers  (2024-06-26T13:12:40Z)
• ConStat: Performance-Based Contamination Detection in Large Language Models [7.305342793164905]
  ConStat is a statistical method that reliably detects and quantifies contamination by comparing performance between a primary and reference benchmark relative to a set of reference models. We demonstrate the effectiveness of ConStat in an extensive evaluation of diverse model architectures, benchmarks, and contamination scenarios.
  arXiv  Detail & Related papers  (2024-05-25T15:36:37Z)
• Autoregressive Score Matching [113.4502004812927]
  We propose autoregressive conditional score models (AR-CSM) where we parameterize the joint distribution in terms of the derivatives of univariable log-conditionals (scores) For AR-CSM models, this divergence between data and model distributions can be computed and optimized efficiently, requiring no expensive sampling or adversarial training. We show with extensive experimental results that it can be applied to density estimation on synthetic data, image generation, image denoising, and training latent variable models with implicit encoders.
  arXiv  Detail & Related papers  (2020-10-24T07:01:24Z)
• Control as Hybrid Inference [62.997667081978825]
  We present an implementation of CHI which naturally mediates the balance between iterative and amortised inference. We verify the scalability of our algorithm on a continuous control benchmark, demonstrating that it outperforms strong model-free and model-based baselines.
  arXiv  Detail & Related papers  (2020-07-11T19:44:09Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.