Evaluating Confidence-Gated Retrieval with Matched Trajectory Replay

• URL: http://arxiv.org/abs/2608.26846v1
• Date: Thu, 27 Aug 2026 09:15:33 GMT
• Title: Evaluating Confidence-Gated Retrieval with Matched Trajectory Replay
• Abstract summary: We propose matched trajectory replay, a controlled protocol for comparing confidence-to-action mappings.<n>We use it to compare raw verbalized confidence with post-hoc isotonic calibration in a multi-hop question-answering system.<n>Overall accuracy improves by up to 15 percentage points on HotpotQA but falls by up to 17 percentage points on MuSiQue.
• Score: 3.888537003098864
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Interactive language-model agents use confidence signals to decide whether to answer immediately, retrieve additional evidence (from memory or external knowledge), or defer. Yet confidence is usually evaluated in isolation, without measuring the trajectory-level consequences of the actions it triggers. We propose matched trajectory replay, a controlled protocol for comparing confidence-to-action mappings. The protocol holds candidate answer states, evidence points, budgets, and action costs fixed. We use it to compare raw verbalized confidence with post-hoc isotonic calibration in a multi-hop question-answering system using Mistral, GPT, and Qwen models on HotpotQA and MuSiQue datasets. At the same numerical commitment threshold, calibration changes which questions agents ultimately commit to answering. Across all six model-dataset pairs, it increases accuracy among committed answers by up to 41 percentage points. However, it can reduce coverage and increase retrieval use. Overall accuracy improves by up to 15 percentage points on HotpotQA but falls by up to 17 percentage points on MuSiQue. These effects reflect a shift to a more selective, lower-risk operating point, not improved answers or confidence ranking. A calibration map fitted before retrieval improves held-out calibration through retrieval depths one and two, but is worse than raw confidence at depth three for all three models. Additional evidence helps on average, but this aggregate effect does not establish whether confidence identifies which individual episodes will benefit from another retrieval. Taken together, these results show that calibration can make commitment risk interpretable, but it does not estimate the expected benefit of another retrieval. Retrieval therefore requires a separate value-of-information or utility estimate. Evaluations should report held-out calibration, risk-coverage, and retrieval cost.

Related papers

• DirEAG: Dirichlet Evidence Aggregation for Calibrating Verbalized Confidence in Mathematical Reasoning [2.867517731896504]
  Black-box verbalized confidence is difficult to calibrate in large language models.<n>We propose DirEAG, a Dirichlet Evidence Aggregation method that converts each elicited answer-confidence observation into calibrated soft evidence.
  arXiv  Detail & Related papers  (2026-08-21T03:48:25Z)
• From token probabilities to calibrated confidence: An empirical study of mathematical question answering [1.0032961794537367]
  Prior work has shown that token probabilities are often overconfident.<n>We compare single-pass estimators, which reuse token probabilities from the original generation, with multi-pass estimators.<n>We find that aggregating token probabilities over the full sequence captures small but consistent differences between correct and incorrect generations.
  arXiv  Detail & Related papers  (2026-08-08T00:00:19Z)
• AB-RAG: Adaptive Budgeted Retrieval-Augmented Generation for Reliable Question Answering [0.0]
  AB-RAG is a training-free and backbone-agnostic framework that generates an answer.<n>It estimates its confidence from a combination of three signals.<n>It then decides whether to stop or to retrieve more evidence.
  arXiv  Detail & Related papers  (2026-06-27T21:08:14Z)
• Confidence Before Answering: A Paradigm Shift for Efficient LLM Uncertainty Estimation [47.91529693614168]
  Existing methods are predominantly answer-first, producing confidence only after generating an answer.<n>We study a confidence-first paradigm, where the model outputs its confidence before answering, interpreting this score as the model's probability of answering the question correctly.<n>We propose CoCA, a reinforcement learning framework that jointly optimize confidence calibration and answer accuracy via segmented credit assignment.
  arXiv  Detail & Related papers  (2026-03-06T04:03:13Z)
• Reference-Free Rating of LLM Responses via Latent Information [53.463883683503106]
  We study the common practice of asking a judge model to assign Likert-scale scores to free-text responses.<n>We then propose and evaluate Latent Judges, which derive scalar ratings from internal model signals.<n>Across a broad suite of pairwise and single-rating benchmarks, latent methods match or surpass standard prompting.
  arXiv  Detail & Related papers  (2025-09-29T12:15:52Z)
• Reconfidencing LLMs from the Grouping Loss Perspective [56.801251926946485]
  Large Language Models (LLMs) are susceptible to generating hallucinated answers in a confident tone. Recent findings show that controlling uncertainty must go beyond calibration. We construct a new evaluation dataset derived from a knowledge base to assess confidence scores given to answers of Mistral and LLaMA.
  arXiv  Detail & Related papers  (2024-02-07T15:40:22Z)
• POCO: 3D Pose and Shape Estimation with Confidence [99.91683561240549]
  We develop POCO, a novel framework for training HPS regressors to estimate not only a 3D human body, but also their confidence. Specifically, POCO estimates both the 3D body pose and a per-sample variance. In all cases, training the network to reason about uncertainty helps it learn to more accurately estimate 3D pose.
  arXiv  Detail & Related papers  (2023-08-24T17:59:04Z)
• Realistic Conversational Question Answering with Answer Selection based on Calibrated Confidence and Uncertainty Measurement [54.55643652781891]
  Conversational Question Answering (ConvQA) models aim at answering a question with its relevant paragraph and previous question-answer pairs that occurred during conversation multiple times. We propose to filter out inaccurate answers in the conversation history based on their estimated confidences and uncertainties from the ConvQA model. We validate our models, Answer Selection-based realistic Conversation Question Answering, on two standard ConvQA datasets.
  arXiv  Detail & Related papers  (2023-02-10T09:42:07Z)
• A Statistical Analysis of Summarization Evaluation Metrics using Resampling Methods [60.04142561088524]
  We find that the confidence intervals are rather wide, demonstrating high uncertainty in how reliable automatic metrics truly are. Although many metrics fail to show statistical improvements over ROUGE, two recent works, QAEval and BERTScore, do in some evaluation settings.
  arXiv  Detail & Related papers  (2021-03-31T18:28:14Z)

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.