Zero Gap Is Not Restoration: Stratified Per-Question Probability Evaluation and Step-wise Mitigation of Benchmark Contamination

• URL: http://arxiv.org/abs/2608.07341v1
• Date: Fri, 07 Aug 2026 15:37:03 GMT
• Title: Zero Gap Is Not Restoration: Stratified Per-Question Probability Evaluation and Step-wise Mitigation of Benchmark Contamination
• Abstract summary: We propose textbfSA-AP (textbfStratified textbfAggregate of textbfPer-question textbfProbability textbfGaps): estimate each question's solve probability by sampling, difference it against the clean model per question, and aggregate within groups defined by the clean model's solve probability.
• Score: 8.221134173409895
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Test data from public benchmarks inevitably leaks into pretraining corpora, inflating evaluation scores once memorized. \textbf{Contamination mitigation evaluation} intervenes in the decoding process to suppress memorization and restore a contaminated model's genuine capability, but its prevailing metric, the \textbf{G-AP} (\textbf{G}ap of \textbf{A}ggregate \textbf{P}erformance), is flawed. Discrete correct/incorrect readouts cannot characterize per-question performance, averaging before differencing lets over- and under-suppression cancel out, and uniform per-question weighting invites strategies to push solve probabilities onto the clean model's high-frequency values. We propose \textbf{SA-PPG} (\textbf{S}tratified \textbf{A}ggregate of \textbf{P}er-question \textbf{P}robability \textbf{G}aps): estimate each question's solve probability by sampling, difference it against the clean model per question, and aggregate within groups defined by the clean model's solve probability. Existing mitigation strategies first estimate where contamination lies and then operate on the estimate, so they are only as correct as the estimate. \textbf{RailCap} instead judges contamination during generation: whenever a sample falls back onto the greedy trajectory, the next trajectory token is capped to the runner-up, accumulating suppression until the response distribution becomes sufficiently dispersed. Across multiple contaminated models and benchmarks, SA-PPG reveals that prior strategies' restoration is substantially overestimated, while RailCap attains the lowest SA-PPG.

Related papers

• Uncertainty-based Debiasing and Unlearning for Decontamination [12.593872705071114]
  Uncertainty-Based Decontamination (UBD) estimates per-sample memorization without requiring a contaminated model or knowledge of which samples are contaminated.<n>UBD produces per-sample output distributions substantially closer to those of an uncontaminated model than paraphrasing or choice-permutation baselines.
  arXiv  Detail & Related papers  (2026-06-22T13:26:29Z)
• Quotient DAGs for Off-Policy Evaluation:Forward-Flow Importance Sampling and Exact Slate Propensities [4.751925814658969]
  Off-policy evaluation estimates how a target policy would perform using data collected by a different behavior policy.<n>Standard importance sampling reweights each logged trajectory, but it can treat details of the generation process as meaningful even when the evaluation target ignores them.<n>We introduce a quotient-DAG view that merges histories equivalent for evaluation and assigns weights using target-to-behavior forward-flow ratios.
  arXiv  Detail & Related papers  (2026-05-28T07:23:40Z)
• Learning from Emptiness: De-biasing Listwise Rerankers with Content-Agnostic Probability Calibration [76.08899010904652]
  CapCal is a training-free framework that mechanically decouples positional bias from ranking decisions.<n>It achieves superior performance among training-free methods while preserving single-pass efficiency.
  arXiv  Detail & Related papers  (2026-04-11T10:47:22Z)
• Robust Regression with Adaptive Contamination in Response: Optimal Rates and Computational Barriers [65.58071581164043]
  We study robust regression under a contamination model in which covariates are clean while the responses may be corrupted in an adaptive manner.<n>We show that the additional information can be carefully exploited to construct an estimator that a better estimation rate than achieves that attainable under Huber contamination.<n>Despite the improvement over the Huber model from an information-theoretic perspective, we provide formal evidence that the problem exhibits strong information-computation gaps.
  arXiv  Detail & Related papers  (2026-04-05T19:07:24Z)
• Fair regression under localized demographic parity constraints [9.082079838093371]
  We introduce a novel ($ell$, Z)-fair predictor, which imposes groupwise CDF constraints of the form F f |S=s (z m ) = $ell$ m for prescribed pairs.<n> Experiments on synthetic and real datasets illustrate an interpretable fairness-accuracy trade-off, enabling targeted corrections at decision-relevant quantiles or thresholds.
  arXiv  Detail & Related papers  (2026-03-26T09:21:05Z)
• Sharp Convergence Rates for Masked Diffusion Models [53.117058231393834]
  We develop a total-variation based analysis for the Euler method that overcomes limitations.<n>Our results relax assumptions on score estimation, improve parameter dependencies, and establish convergence guarantees.<n>Overall, our analysis introduces a direct TV-based error decomposition along the CTMC trajectory and a decoupling-based path-wise analysis for FHS.
  arXiv  Detail & Related papers  (2026-02-26T00:47:51Z)
• Post-Hoc Split-Point Self-Consistency Verification for Efficient, Unified Quantification of Aleatoric and Epistemic Uncertainty in Deep Learning [5.996056764788456]
  Uncertainty quantification (UQ) is vital for trustworthy deep learning, yet existing methods are either computationally intensive or provide only partial, task-specific estimates.<n>We propose a post-hoc single-forward-pass framework that jointly captures aleatoric and epistemic uncertainty without modifying or retraining pretrained models.<n>Our method applies emphSplit-Point Analysis (SPA) to decompose predictive residuals into upper and lower subsets, computing emphMean Absolute Residuals (MARs) on each side.
  arXiv  Detail & Related papers  (2025-09-16T17:16:01Z)
• COIN: Uncertainty-Guarding Selective Question Answering for Foundation Models with Provable Risk Guarantees [51.5976496056012]
  COIN is an uncertainty-guarding selection framework that calibrates statistically valid thresholds to filter a single generated answer per question.<n>COIN estimates the empirical error rate on a calibration set and applies confidence interval methods to establish a high-probability upper bound on the true error rate.<n>We demonstrate COIN's robustness in risk control, strong test-time power in retaining admissible answers, and predictive efficiency under limited calibration data.
  arXiv  Detail & Related papers  (2025-06-25T07:04:49Z)
• Rejection via Learning Density Ratios [50.91522897152437]
  Classification with rejection emerges as a learning paradigm which allows models to abstain from making predictions.<n>We propose a different distributional perspective, where we seek to find an idealized data distribution which maximizes a pretrained model's performance.<n>Our framework is tested empirically over clean and noisy datasets.
  arXiv  Detail & Related papers  (2024-05-29T01:32:17Z)
• Selective Regression Under Fairness Criteria [30.672082160544996]
  In some cases, the performance of minority group can decrease while we reduce the coverage. We show that such an unwanted behavior can be avoided if we can construct features satisfying the sufficiency criterion.
  arXiv  Detail & Related papers  (2021-10-28T19:05:12Z)
• Evaluating probabilistic classifiers: Reliability diagrams and score decompositions revisited [68.8204255655161]
  We introduce the CORP approach, which generates provably statistically Consistent, Optimally binned, and Reproducible reliability diagrams in an automated way. Corpor is based on non-parametric isotonic regression and implemented via the Pool-adjacent-violators (PAV) algorithm.
  arXiv  Detail & Related papers  (2020-08-07T08:22:26Z)
• Pre-training Is (Almost) All You Need: An Application to Commonsense Reasoning [61.32992639292889]
  Fine-tuning of pre-trained transformer models has become the standard approach for solving common NLP tasks. We introduce a new scoring method that casts a plausibility ranking task in a full-text format. We show that our method provides a much more stable training phase across random restarts.
  arXiv  Detail & Related papers  (2020-04-29T10:54:40Z)

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.