Most confidence estimators judge a response solely from the current inference trace. This paper argues that an LLM's own history — graded past episodes annotated with outcome and lessons — is a powerful signal for estimating whether a new answer will be correct. XConf uses recall + reflection over the model's experience to produce a revised, empirically grounded confidence score.
AI Agent Papers2026
Confidence Comes from Experience: Experiential Confidence Estimation from Reasoning to Agents
Estimates a model's probability of correctness by leveraging its past graded episodes: XConf recalls similar past cases and prompts the model to reflect and restate confidence, improving calibration and discrimination across reasoning, coding, multimodal QA, and agent tasks at much lower generation cost.
Introduction
Information
- Websitearxiv.org
- AuthorsCaiqi Zhang, Xiaochen Zhu, Chengzu Li, Yulong Chen, Dharshan Kumaran, Nigel Collier
- Published date2026/09/15
More Items
AI Agent Papers2026
Builds executable professional tasks by mapping required information dependencies and repairing gaps revealed during test runs. It produces 112,686 tasks across 612 occupations for collecting teacher demonstrations and training working agents.
Decides when a diffusion vision-language model needs to process an image by checking whether text-only predictions agree with its latest image-conditioned predictions. Retains 99.6% of baseline scores across ten benchmarks at 35.3% of decoding FLOPs.
Yifan Sun, Naicheng Yu +3University of Illinois Urbana-Champaign, University of California San Diego +1
Treats a benchmark as a distribution over item difficulty, measuring both mismatch with a target distribution and uncovered difficulty regions. BenchFill rewrites items to fill the largest gaps.