The Confidence Paradox: Can LLM Know When It's Wrong

• URL: http://arxiv.org/abs/2506.23464v1
• Date: Mon, 30 Jun 2025 02:06:54 GMT
• Title: The Confidence Paradox: Can LLM Know When It's Wrong
• Authors: Sahil Tripathi, Md Tabrez Nafis, Imran Hussain, Jiechao Gao,
• Abstract summary: We introduce HonestVQA, a self-supervised honesty calibration framework for ethically aligned DocVQA.<n>Our model-agnostic method quantifies uncertainty to identify knowledge gaps, aligns model confidence with actual correctness using weighted loss functions, and enforces ethical response behavior via contrastive learning.<n> Empirically, HonestVQA improves DocVQA accuracy by up to 4.3% and F1 by 4.3% across SpDocVQA, InfographicsVQA, and SROIE datasets.
• Score: 5.545086863155316
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Document Visual Question Answering (DocVQA) systems are increasingly deployed in real world applications, yet they remain ethically opaque-often producing overconfident answers to ambiguous questions or failing to communicate uncertainty in a trustworthy manner. This misalignment between model confidence and actual knowledge poses significant risks, particularly in domains requiring ethical accountability. Existing approaches such as LayoutLMv3, UDOP, and DONUT have advanced SOTA performance by focusing on architectural sophistication and accuracy; however, they fall short in ethical responsiveness. To address these limitations, we introduce HonestVQA, a self-supervised honesty calibration framework for ethically aligned DocVQA. Our model-agnostic method quantifies uncertainty to identify knowledge gaps, aligns model confidence with actual correctness using weighted loss functions, and enforces ethical response behavior via contrastive learning. We further introduce two principled evaluation metrics--Honesty Score (H-Score) and Ethical Confidence Index (ECI)--to benchmark alignment between confidence, accuracy, and ethical communication. Empirically, HonestVQA improves DocVQA accuracy by up to 4.3% and F1 by 4.3% across SpDocVQA, InfographicsVQA, and SROIE datasets. It reduces overconfidence, lowering H-Score and ECI by 0.072 and 0.078, respectively. In cross domain evaluation, it achieves up to 78.9% accuracy and 76.1% F1-score, demonstrating strong generalization. Ablation shows a 3.8% drop in accuracy without alignment or contrastive loss.

Related papers

• Confidence-Diversity Calibration of AI Judgement Enables Reliable Qualitative Coding [0.0]
  Analysing 5,680 coding decisions from eight state-of-the-art LLMs across ten thematic categories.<n>Adding model diversity-quantified as the normalised Shannon entropy of the panel's votes-turns this single cue into a dual signal that explains agreement almost completely.
  arXiv  Detail & Related papers  (2025-08-04T03:47:10Z)
• Trust, or Don't Predict: Introducing the CWSA Family for Confidence-Aware Model Evaluation [0.0]
  We introduce two new metrics Confidence-Weighted Selective Accuracy (CWSA) and its normalized variant CWSA+.<n>CWSA offers principled and interpretable way to evaluate predictive models under confidence thresholds.<n>We show that CWSA and CWSA+ both effectively detect nuanced failure modes and outperform classical metrics in trust-sensitive tests.
  arXiv  Detail & Related papers  (2025-05-24T10:07:48Z)
• REVAL: A Comprehension Evaluation on Reliability and Values of Large Vision-Language Models [59.445672459851274]
  REVAL is a comprehensive benchmark designed to evaluate the textbfREliability and textbfVALue of Large Vision-Language Models.<n>REVAL encompasses over 144K image-text Visual Question Answering (VQA) samples, structured into two primary sections: Reliability and Values.<n>We evaluate 26 models, including mainstream open-source LVLMs and prominent closed-source models like GPT-4o and Gemini-1.5-Pro.
  arXiv  Detail & Related papers  (2025-03-20T07:54:35Z)
• MCQA-Eval: Efficient Confidence Evaluation in NLG with Gold-Standard Correctness Labels [16.300463494913593]
  Large Language Models (LLMs) require robust confidence estimation.<n>McQCA-Eval is an evaluation framework for assessing confidence measures in Natural Language Generation.
  arXiv  Detail & Related papers  (2025-02-20T05:09:29Z)
• Learning Conformal Abstention Policies for Adaptive Risk Management in Large Language and Vision-Language Models [3.958317527488534]
  Large Language and Vision-Language Models (LLMs/VLMs) are increasingly used in safety-critical applications.<n>Uncertainty quantification helps assess prediction confidence and enables abstention when uncertainty is high.<n>We propose learnable abstention, integrating reinforcement learning (RL) with Conformal Prediction (CP) to optimize abstention thresholds.
  arXiv  Detail & Related papers  (2025-02-08T21:30:41Z)
• Language Models Prefer What They Know: Relative Confidence Estimation via Confidence Preferences [62.52739672949452]
  Language models (LMs) should provide reliable confidence estimates to help users detect mistakes in their outputs and defer to human experts when necessary.<n>We propose relative confidence estimation, where we match up questions against each other and ask the model to make relative judgments of confidence.<n>Treating each question as a "player" in a series of matchups against other questions and the model's preferences as match outcomes, we can use rank aggregation methods like Elo rating and Bradley-Terry to translate the model's confidence preferences into confidence scores.
  arXiv  Detail & Related papers  (2025-02-03T07:43:27Z)
• Fact-Level Confidence Calibration and Self-Correction [64.40105513819272]
  We propose a Fact-Level framework that calibrates confidence to relevance-weighted correctness at the fact level. We also develop Confidence-Guided Fact-level Self-Correction ($textbfConFix$), which uses high-confidence facts within a response as additional knowledge to improve low-confidence ones.
  arXiv  Detail & Related papers  (2024-11-20T14:15:18Z)
• When to Trust LLMs: Aligning Confidence with Response Quality [49.371218210305656]
  We propose CONfidence-Quality-ORDer-preserving alignment approach (CONQORD) It integrates quality reward and order-preserving alignment reward functions. Experiments demonstrate that CONQORD significantly improves the alignment performance between confidence and response accuracy.
  arXiv  Detail & Related papers  (2024-04-26T09:42:46Z)
• Binary Classification with Confidence Difference [100.08818204756093]
  This paper delves into a novel weakly supervised binary classification problem called confidence-difference (ConfDiff) classification. We propose a risk-consistent approach to tackle this problem and show that the estimation error bound the optimal convergence rate. We also introduce a risk correction approach to mitigate overfitting problems, whose consistency and convergence rate are also proven.
  arXiv  Detail & Related papers  (2023-10-09T11:44:50Z)
• Reliability-Aware Prediction via Uncertainty Learning for Person Image Retrieval [51.83967175585896]
  UAL aims at providing reliability-aware predictions by considering data uncertainty and model uncertainty simultaneously. Data uncertainty captures the noise" inherent in the sample, while model uncertainty depicts the model's confidence in the sample's prediction.
  arXiv  Detail & Related papers  (2022-10-24T17:53:20Z)

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.