Towards Reliable LLM Evaluation: Correcting the Winner's Curse in Adaptive Benchmarking

• URL: http://arxiv.org/abs/2605.05973v1
• Date: Thu, 07 May 2026 10:18:56 GMT
• Title: Towards Reliable LLM Evaluation: Correcting the Winner's Curse in Adaptive Benchmarking
• Abstract summary: We study inference for this procedure-level target under explicit tuning budgets.<n>We propose SIREN, a selection-aware repeated-split reporting protocol.<n>Controlled simulations and MMLU-Pro tuning experiments show that winner-based reporting can be optimistic.
• Score: 40.44372785257615
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Adaptive prompt and program search makes LLM evaluation selection-sensitive. Once benchmark items are reused inside tuning, the observed winner's score need not estimate the fresh-data performance of the full tune-then-deploy procedure. We study inference for this procedure-level target under explicit tuning budgets. We propose SIREN, a selection-aware repeated-split reporting protocol that freezes the post-search shortlist, separates splitwise selection from held-out evaluation, and uses an item-level Gaussian multiplier bootstrap for uncertainty quantification. In a fixed-shortlist regime with smooth stabilized selection, the estimator admits a first-order item-level representation, and the bootstrap yields valid simultaneous inference on a finite budget grid. This supports confidence intervals for procedure-performance curves and pre-specified equal-budget and cross-budget comparisons. Controlled simulations and MMLU-Pro tuning experiments show that winner-based reporting can be optimistic and can change deployment conclusions, while SIREN remains close to the finite-sample reporting target.

Related papers

• SVR: Self-Verifying Refinement via Joint Verdict-Confidence Reinforcement Learning for Adaptive Test-Time Compute [62.3458279176813]
  Self-Verifying Refinement is an oracle-free multi-turn reinforcement learning framework.<n>It learns to use self-verification as a compute-control policy.<n>It achieves a macro-average accuracy of 0.563 with only 2.99 inference turns on average.
  arXiv  Detail & Related papers  (2026-07-30T16:20:58Z)
• LLM-as-a-Verifier: A General-Purpose Verification Framework [74.40111651545979]
  We introduce LLM-as-a-Verifier, a general-purpose verification framework.<n>It provides fine-grained feedback for agentic tasks without requiring additional training.<n>It achieves state-of-the-art performance on several benchmarks.
  arXiv  Detail & Related papers  (2026-07-06T17:59:35Z)
• Embedding by Elicitation: Dynamic Representations for Bayesian Optimization of System Prompts [22.060204303535187]
  System prompts are a central control mechanism in modern AI systems, shaping behavior across conversations, tasks, and user populations.<n>We study this aggregate feedback setting as sample-constrained black-box optimization over discrete, variable-length text.<n>We introduce ReElicit, a Bayesian optimization framework based on emphembedding by elicitation.
  arXiv  Detail & Related papers  (2026-05-18T20:28:17Z)
• Confident Rankings with Fewer Items: Adaptive LLM Evaluation with Continuous Scores [25.638175689769934]
  We present a principled extension of IRT-based adaptive testing to continuous bounded scores (ROUGE, BLEU, LLM-as-a-Judge)<n>We introduce an uncertainty aware ranker with adaptive stopping criteria that achieves reliable model ranking while testing as few items as possible.<n>Our method uses 2% of the items while improving ranking correlation by 0.12 over random sampling, with 95% accuracy on confident predictions.
  arXiv  Detail & Related papers  (2026-01-20T11:59:13Z)
• Reinforce-Ada: An Adaptive Sampling Framework for Reinforce-Style LLM Training [47.26632817047513]
  Reinforcement learning applied to large language models (LLMs) for reasoning tasks is often bottlenecked by unstable gradient estimates.<n>We propose Reinforce-Ada, an adaptive sampling framework for online RL post-training of LLMs.<n>Unlike conventional two-stage allocation methods, Reinforce-Ada interleaves estimation and sampling in an online successive elimination process.
  arXiv  Detail & Related papers  (2025-10-06T16:34:09Z)
• Unsupervised Conformal Inference: Bootstrapping and Alignment to Control LLM Uncertainty [49.19257648205146]
  We propose an unsupervised conformal inference framework for generation.<n>Our gates achieve close-to-nominal coverage and provide tighter, more stable thresholds than split UCP.<n>The result is a label-free, API-compatible gate for test-time filtering.
  arXiv  Detail & Related papers  (2025-09-26T23:40:47Z)
• SCOPE: Stochastic and Counterbiased Option Placement for Evaluating Large Language Models [0.27309692684728604]
  Large Language Models (LLMs) can achieve inflated scores on multiple-choice tasks by exploiting inherent biases in option positions or labels.<n>This study introduces SCOPE, an evaluation framework designed to measure and mitigate such selection bias in a dataset-independent manner.
  arXiv  Detail & Related papers  (2025-07-24T08:28:17Z)
• A Principled Approach to Randomized Selection under Uncertainty: Applications to Peer Review and Grant Funding [61.86327960322782]
  We propose a principled framework for randomized decision-making based on interval estimates of the quality of each item.<n>We introduce MERIT, an optimization-based method that maximizes the worst-case expected number of top candidates selected.<n>We prove that MERIT satisfies desirable axiomatic properties not guaranteed by existing approaches.
  arXiv  Detail & Related papers  (2025-06-23T19:59:30Z)
• CalibraEval: Calibrating Prediction Distribution to Mitigate Selection Bias in LLMs-as-Judges [21.580762639442913]
  We introduce CalibraEval, a novel label-free method for mitigating selection bias during inference. CalibraEval reformulates debiasing as an optimization task aimed at adjusting observed prediction distributions to align with unbiased prediction distributions. We show that CalibraEval effectively mitigates selection bias and improves performance compared to existing debiasing methods.
  arXiv  Detail & Related papers  (2024-10-20T13:47:39Z)
• Self-Evaluation Improves Selective Generation in Large Language Models [54.003992911447696]
  We reformulate open-ended generation tasks into token-level prediction tasks. We instruct an LLM to self-evaluate its answers. We benchmark a range of scoring methods based on self-evaluation.
  arXiv  Detail & Related papers  (2023-12-14T19:09:22Z)

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.