Does Visual Token Pruning Improve Calibration? An Empirical Study on Confidence in MLLMs

• URL: http://arxiv.org/abs/2604.12035v1
• Date: Mon, 13 Apr 2026 20:24:03 GMT
• Title: Does Visual Token Pruning Improve Calibration? An Empirical Study on Confidence in MLLMs
• Abstract summary: We study how visual token pruning affects model calibration, that is, whether predicted confidence matches actual correctness.<n>Our results show that pruning does not simply trade reliability for efficiency.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Visual token pruning is a widely used strategy for efficient inference in multimodal large language models (MLLMs), but existing work mainly evaluates it with task accuracy. In this paper, we study how visual token pruning affects model calibration, that is, whether predicted confidence matches actual correctness. Using LLaVA-1.5-7B on POPE and ScienceQA-IMG, we evaluate Expected Calibration Error (ECE), Brier score, and AURC under several pruning strategies, including SCOPE with different saliency weights, saliency-only pruning, FastV, and random pruning, across multiple token budgets. Our results show that pruning does not simply trade reliability for efficiency. On POPE, a pure-coverage setting in SCOPE achieves substantially lower ECE than the full unpruned model while maintaining similar accuracy. An internal alpha-sweep further shows a consistent trend: reducing the saliency weight improves calibration at all tested token budgets, while accuracy changes only slightly. In contrast, saliency-based pruning leads to worse calibration, and real FastV causes severe performance degradation in our setting. On ScienceQA-IMG, pruning also reduces ECE, with accuracy remaining stable or slightly improving. We additionally study the gap power exponent in coverage-based selection and find that its default setting is not always optimal. Overall, our results suggest that visual token pruning should be evaluated not only by accuracy, but also by confidence quality, especially for multimodal systems that need reliable decisions.

Related papers

• Don't Go Breaking My LLM: The Impact of Pruning Attention Layers on Explanation Faithfulness and Confidence Calibration [17.14392068503777]
  Pruning Large Language Models (LLMs) reduces memory and inference costs by removing parts of the network.<n>We study how pruning attention layers affects explanation faithfulness and confidence calibration across five LLMs and eight datasets.<n>Our findings suggest that layer pruning can affect LLMs' interpretability and reliability in ways not captured by accuracy and efficiency measures alone.
  arXiv  Detail & Related papers  (2026-06-23T11:07:59Z)
• Decoupling Reasoning and Confidence: Resurrecting Calibration in Reinforcement Learning from Verifiable Rewards [71.19033708090389]
  Reinforcement Learning from Verifiable Rewards (RLVR) significantly enhances large language models (LLMs) reasoning but severely suffers from calibration degeneration.<n>We propose DCPO, a framework that systematically decouples reasoning and calibration objectives.
  arXiv  Detail & Related papers  (2026-03-10T02:47:59Z)
• On Calibration of Large Language Models: From Response To Capability [66.59139960234326]
  Large language models (LLMs) are widely deployed as general-purpose problem solvers.<n>We introduce capability calibration, which targets the model's expected accuracy on a query.<n>Our results demonstrate that capability-calibrated confidence improves pass@$k$ prediction and inference budget allocation.
  arXiv  Detail & Related papers  (2026-02-14T01:07:45Z)
• Balancing Classification and Calibration Performance in Decision-Making LLMs via Calibration Aware Reinforcement Learning [10.123352394689134]
  Well-calibrated confidence enables downstream systems to decide when to trust a model and when to defer to fallback mechanisms.<n>We show that while RLVR improves task performance, it produces extremely overconfident models.<n>We propose a calibration-aware reinforcement learning formulation that directly adjusts decision-token probabilities.
  arXiv  Detail & Related papers  (2026-01-19T18:31:31Z)
• Can Less Precise Be More Reliable? A Systematic Evaluation of Quantization's Impact on CLIP Beyond Accuracy [38.91808019403202]
  We evaluate the impact of quantization on CLIP's performance beyond accuracy.<n>We show that quantization consistently improves calibration for typically underconfident pre-trained models.<n>We identify specific quantization-aware training (QAT) methods that yield simultaneous gains in zero-shot accuracy, calibration, and OOD robustness.
  arXiv  Detail & Related papers  (2025-09-25T13:54:34Z)
• NIRVANA: Structured pruning reimagined for large language models compression [50.651730342011014]
  We introduce NIRVANA, a novel pruning method designed to balance immediate zero-shot preservation accuracy with robust fine-tuning.<n>To further address the unique challenges posed by structured pruning, NIRVANA incorporates an adaptive sparsity allocation mechanism across layers and modules.<n>Experiments conducted on Llama3, Qwen, T5 models demonstrate that NIRVANA outperforms existing structured pruning methods under equivalent sparsity constraints.
  arXiv  Detail & Related papers  (2025-09-17T17:59:00Z)
• Beyond Binary Rewards: Training LMs to Reason About Their Uncertainty [59.97939500426759]
  This paper describes RLCR, an approach to training reasoning models that jointly improves accuracy and confidence estimation.<n>We show that across diverse datasets, RLCR substantially improves calibration with no loss in accuracy.<n>We also demonstrate that verbalized confidence can be leveraged at test time to improve accuracy and calibration.
  arXiv  Detail & Related papers  (2025-07-22T17:56:01Z)
• A Confidence Interval for the $\ell_2$ Expected Calibration Error [35.88784957918326]
  We develop confidence intervals $ell$ Expected the Error (ECE)<n>We consider top-1-to-$k$ calibration, which includes both the popular notion of confidence calibration as well as calibration.<n>For a debiased estimator of the ECE, we show normality, but with different convergence rates and variances for calibrated and misd models.
  arXiv  Detail & Related papers  (2024-08-16T20:00:08Z)
• Towards Calibrated Robust Fine-Tuning of Vision-Language Models [97.19901765814431]
  This work proposes a robust fine-tuning method that improves both OOD accuracy and confidence calibration simultaneously in vision language models. We show that both OOD classification and OOD calibration errors have a shared upper bound consisting of two terms of ID data. Based on this insight, we design a novel framework that conducts fine-tuning with a constrained multimodal contrastive loss enforcing a larger smallest singular value.
  arXiv  Detail & Related papers  (2023-11-03T05:41:25Z)
• Bridging Precision and Confidence: A Train-Time Loss for Calibrating Object Detection [58.789823426981044]
  We propose a novel auxiliary loss formulation that aims to align the class confidence of bounding boxes with the accurateness of predictions. Our results reveal that our train-time loss surpasses strong calibration baselines in reducing calibration error for both in and out-domain scenarios.
  arXiv  Detail & Related papers  (2023-03-25T08:56:21Z)

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.