Conditional Evaluation of Language Models with Cheap Auxiliary Signals

• URL: http://arxiv.org/abs/2608.16210v1
• Date: Mon, 17 Aug 2026 07:44:22 GMT
• Title: Conditional Evaluation of Language Models with Cheap Auxiliary Signals
• Abstract summary: Estimating conditional performance profiles from gold labels alone is expensive.<n>We propose LACE (Local Augmented Control-Variate Evaluation), a semi-supervised estimator for conditional LLM evaluation.<n>We empirically evaluate the primary performance-profile estimator on MATH-500, ScienceQA, MMLU, WinoGrande, HellaSwag, TruthfulQA, GSM8K, and ARC.
• Score: 7.275460635772361
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Aggregate accuracy hides where models succeed and fail. Estimating conditional performance profiles from gold labels alone is expensive, while cheap auxiliary signals such as LLM-judge scores, pairwise comparisons, confidence scores, and judge-disagreement features can be collected for every benchmark item but are often biased or miscalibrated. We propose LACE (Local Augmented Control-Variate Evaluation), a semi-supervised estimator for conditional LLM evaluation. The key step is local centering: after subtracting the conditional mean of a cheap signal within the target profile region, any linear augmentation has zero conditional mean and therefore cannot change the estimand. The augmentation coefficient is used only for efficiency, and a local ridge control variate combines a gold-label residual mean from the labeled subset with a cheap-signal mean from the full item pool. We prove calibration-free identification, unbiasedness for grouped profiles, local oracle optimality within centered linear augmentations, and first-order adaptivity to the estimated coefficient. The resulting gain formula is governed by a population local $R^2$, which characterizes how the efficiency attainable from the cheap signals varies across profile values. We also derive corresponding estimators for direct paired model gaps and deployment-weighted scores. We empirically evaluate the primary performance-profile estimator on MATH-500, ScienceQA, MMLU, WinoGrande, HellaSwag, TruthfulQA, GSM8K, and ARC.

Related papers

• 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)
• Heteroskedastic Signals in Budgeted LLM Verification: Structural Heterogeneity Limits Optimization Gains [4.560961083890857]
  Large language model (LLM) systems increasingly use uncertainty signals to allocate limited computation across verification, test-time scaling, tool execution, and other selective-compute decisions.<n>We show that uncertainty quality is heteroskedastic across cost strata, with some regions exhibiting near-random discriminability despite concentrating many errors.
  arXiv  Detail & Related papers  (2026-06-14T14:45:57Z)
• Instance-Optimal Estimation with Multiple LLM Judges on a Budget [84.31744861038106]
  We formalize this question as *budgeted heteroskedastic multi-judge estimation*.<n>Given $K$ prompt-response pairs, $J$ judges with known costs, and unknown query-judge variances, the goal is to estimate a bounded score vector while minimizing an $ell_p$-error.<n>We prove that EST-IVWE matches the oracle IVWE rate up to lower-order terms in the budget.
  arXiv  Detail & Related papers  (2026-05-22T08:26:08Z)
• OPPO: Bayesian Value Recursion for Token-Level Credit Assignment in LLM Reasoning [17.98540130851038]
  Reinforcement learning with verifiable rewards has become the standard recipe for improving LLM reasoning.<n>But the dominant algorithm GRPO assigns a single trajectory-level advantage to every token.<n>We propose Oracle-Prompted Policy Optimization (OPPO), which rests on a single observation.
  arXiv  Detail & Related papers  (2026-05-21T00:55:13Z)
• When Individually Calibrated Models Become Collectively Miscalibrated [3.556355987197792]
  We show that individually calibrated predictors can become collectively miscalibrated when their predictions interact strategically.<n>We prove that under Brier-score-based aggregation with positively correlated beliefs, each agent's individually optimal report systematically underestimates the positive-class probability.<n>In contrast, VCG-based aggregation aligns incentives by rewarding marginal contribution, achieving dominant-strategy incentive compatibility and near-optimal performance.
  arXiv  Detail & Related papers  (2026-05-14T05:25:16Z)
• Evidence-based Distributional Alignment for Large Language Models [58.65469623911573]
  We propose Evi-DA, an evidence-based alignment technique that improves the fidelity and robustness of LLM-based distribution estimation.<n>Given a target country, Evi-DA retrieves related World Values Survey items and their answer distributions, predicts a coarse Welzel value signature for each option, and infers the country-conditioned answer distribution in a structured format.
  arXiv  Detail & Related papers  (2026-03-03T03:34:06Z)
• Evaluating LLMs When They Do Not Know the Answer: Statistical Evaluation of Mathematical Reasoning via Comparative Signals [18.612081365101464]
  We develop a framework that combines standard labeled outcomes with pairwise comparison signals obtained by having models judge auxiliary reasoning chains.<n>Across simulations, our one-step estimator substantially improves ranking accuracy with gains increasing as model output noise grows.<n>Experiments on GPQA Diamond, AIME 2025 and GSM8K further demonstrate more precise performance estimation and more reliable model rankings.
  arXiv  Detail & Related papers  (2026-02-03T03:40:01Z)
• KAIROS: Scalable Model-Agnostic Data Valuation [8.766103946679435]
  KAIROS is a scalable, model-agnostic valuation framework that assigns each example a distributional influence score.<n> KAIROS consistently outperforms state-of-the-art model-, Shapley-, and Wasserstein-based baselines in both accuracy and runtime.
  arXiv  Detail & Related papers  (2025-06-30T12:44:28Z)
• 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)
• Stochastic Optimization of Areas Under Precision-Recall Curves with Provable Convergence [66.83161885378192]
  Area under ROC (AUROC) and precision-recall curves (AUPRC) are common metrics for evaluating classification performance for imbalanced problems. We propose a technical method to optimize AUPRC for deep learning.
  arXiv  Detail & Related papers  (2021-04-18T06:22:21Z)
• SLOE: A Faster Method for Statistical Inference in High-Dimensional Logistic Regression [68.66245730450915]
  We develop an improved method for debiasing predictions and estimating frequentist uncertainty for practical datasets. Our main contribution is SLOE, an estimator of the signal strength with convergence guarantees that reduces the computation time of estimation and inference by orders of magnitude.
  arXiv  Detail & Related papers  (2021-03-23T17:48:56Z)

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.