Confidence Estimation in Language Models

Follow this topic Ask about this topic

The field studies whether language models can tell when their answers are likely to be wrong, which matters for deciding when to answer, abstain, retrieve evidence, or spend more computation. Methods include verbalized confidence, token probabilities and sampling-based agreement, as well as hidden-state probes, activation signals, post-hoc calibration, and conformal prediction. These approaches are tested across settings such as factual question answering, mathematical reasoning, code generation, multilingual tasks, and tool-using agents. A central challenge is that confidence signals can disagree or shift with prompts, training, and distribution changes; good calibration or error ranking alone does not guarantee useful decisions or reliable risk guarantees.

Latest papers

249 papers45 papers in the last four weeks, up 96% on the four weeks before. 0.3% of all new papers.

SCARL: Sequence Confidence-Aligned Reinforcement Learning
Oct 29, 2026Zhenghong Ke, Jingbo Zhou, Hua WuReinforcement Learning RL for Language Model Reasoning
Baidu Frontier Research Department · Nanjing University
Blind Spots: Where Small Language Models Fall Behind Frontier Models on Terminal Tasks
Oct 28, 2026Ashima Suvarna, Ruta Desai, Ahmed ElgoharySmall Language Models LLM Evaluation
University of California, Los Angeles · Microsoft Research
Is Token Confidence Just Chain Length? A Signal-Processing Audit of Reasoning Traces
Oct 27, 2026Pragna MallappaLLM Reasoning Language Model Reasoning Evaluation
Keshav Memorial Institute of Technology
Confidence Is Not a Promise: A Reference Audit of Typed Decision Models
Oct 26, 2026Pedro Carvalho BromLLM Reliability LLM Auditing
Department of Mathematics, Federal Institute of Brasília, Brasília, Brazil
Not One Belief: Answer Payloads, Persistence, and Dissociated Confidence Signals in Conversational Question Answering
Oct 26, 2026Duc-Vinh TranConversational QA Belief Updating in LLMs
School of Information and Communication Technology, Hanoi University of Science and Technology, Hanoi, 100000, Vietnam.
Test-Time Confidence Adaptation under RL-Induced Trajectory Shift
Oct 26, 2026Zhenzhen Ren, Xiaoqing Zhang, GuoBiao Li +4Confidence Estimation in Language Models Inference-Time Adaptation of Language Models
Fudan University · Zhongguancun Academy · Renmin University of China +4
CalibraRL: Confidence-Calibrated Reinforcement Learning for Mitigating Hallucinations in RAG
Oct 26, 2026Qianchi Zhang, Hainan Zhang, Liang Pang +2Hallucination in RAG RL for RAG
School of Artificial Intelligence, Beihang University, Beijing, China · Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China · Beijing Academy of Blockchain and Edge Computing, Beijing, China
Why Are LLMs Confidently Wrong? Correcting Overconfident Errors via Causal Head Intervention
Oct 26, 2026Jing Ren, Bowen Li, Ziqi Xu +2Language Model Calibration Confidence Estimation in Language Models
RMIT University
STEP: Soft Trajectory Estimation of Prediction Correctness via Per-Step Probability Scoring
Oct 25, 2026Chenyang Yu, Zifan Zeng, Qunli Zhang +1Confidence Estimation in Language Models Language Model Performance Prediction
Technical University of Munich · RAMS Lab, Huawei Heisenberg Research Center
Two Uncertainties Are Not One: Disagreement versus Decision-Risk in Cheap Structured Content Moderation
Oct 23, 2026Karan YadavContent Moderation LLM Evaluation
Google
Each Insufficient in Its Own Way: Calibrating Language Models Across Varying Evidence States
Oct 23, 2026Jiayi Li, Dennis L. Shung, Bradly C. StadieConfidence Estimation in Language Models Language Model Calibration
Northwestern University · Mayo Clinic
Disentangling Question, Response, and Verifier Languages in Multilingual P(True) Self-Evaluation
Oct 23, 2026Jewon Yeom, Jaewon Sok, Seonghyeon Park +3Language Model Self-Assessment Multilingual Language Model Evaluation
Graduate School of Data Science, Seoul National University
Which Answer Did the Seventeen-Year-Old Write? A Behavioral Case Study of Author-Age Misattribution, Miscalibrated Confidence, and Non-Corrective Self-Explanation in Six Frontier Language Models
Oct 22, 2026Brianne LeeDemographic Bias in Language Models Language Model Self-Assessment
Independent Researcher
Free Lunch for LLMs Confidence: The RL Critic Already Knows
Oct 22, 2026Shujin Wu, Heng JiConfidence Estimation in Language Models LLM Reliability
University of Illinois Urbana-Champaign
Would an LLM Match into a Residency? Uncertainty, Reasoning and Image Use of Open Language Models on the Spanish MIR Examination
Oct 22, 2026Miguel Fernandez-de-Retana, Unai Zulaika, Ruben Sanchez-Corcuera +2Clinical Language Model Evaluation Medical QA
Faculty of Engineering, University of Deusto (Bilbao, Spain)
The Value of System 1 Models in Medicine
Oct 22, 2026Xiufeng Yang, Konstantinos Vrettos, Zeineb Ben Chaaben +17Clinical Language Model Evaluation Multimodal Model Evaluation
Department of Diagnostic and Interventional Radiology, School of Medicine and Health, Technical University of Munich, Munich, Germany · Artificial Intelligence and Translational Imaging Lab, Department of Radiology, School of Medicine, University of Crete, Heraklion, Greece · Department of Medical Imaging, University Hospital of Heraklion, Heraklion, Crete, Greece +6
A Strong MoE Is All You Need for Process Supervision
Oct 21, 2026Guanzhi Deng, Zengyan Liu, Haibo Wang +5Process Supervision Process Reward Models
City University of Hong Kong · Carnegie Mellon University · University of Hong Kong
Beyond Calibration: Temporal LLM Confidence Assessment Across Languages and Local Settings
Oct 20, 2026Sahil KaleConfidence Estimation in Language Models Metacognition in Language Models
University of California, Los Angeles CA, USA
Learned Readout Tokens for Extracting Large Language Model Confidence
Oct 20, 2026Nima Seifi, Martin EsterConfidence Estimation in Language Models Language Model Probing
School of Computing Science, Simon Fraser University, 8888 University Drive, Burnaby, BC V5A 1S6, Canada
Measurement Validity and Representation Dynamics of Confidence Signals in Large Language Models
Oct 19, 2026Debojeet Ghosh, Shreyansh Kumar, Ayushman GargConfidence Estimation in Language Models Language Model Calibration
Can This LLM Annotate My Data? Estimating Zero-Shot LLM Performance
Oct 16, 2026Tong Lin, Grant Schoenebeck, Sabina TomkinsLanguage Model Performance Prediction LLM-Assisted Annotation
School of Information, University of Michigan
Truth in Contrast: Calibrated Factual Evidence from Language Model Likelihoods
Oct 16, 2026Ponhvoan Srey, Jason Lucas, Pittawat Taveekitworachai +2Confidence Estimation in Language Models Language Model Calibration
Nanyang Technological University · University of Colorado at Boulder · National University of Singapore +1
Who Pays for Efficient Inference? Confidence-Based Small-to-Large LLM Routing Across Tasks and Languages
Oct 16, 2026Md Sabbir Hossen, Anichur Rahman, Pabon Shaha +3LLM Routing LLM Inference Acceleration
Bangladesh University, Bangladesh · Georgia Southern University, USA · University of Southern Mississippi, USA
Are You Sure Your LLM Is Sure? Repairing and Auditing Confidence
Oct 15, 2026Arijit Dey, Bhartendu Kumar, Ramya Hebbalaguppe +1Confidence Estimation in Language Models Language Model Calibration
MBZUAI · TCS Research Labs, India · IIT Delhi
Effectiveness of Hidden-State Probes for Localizing Reasoning Failures
Oct 15, 2026Shubham Subhnil, Yannis Belkhiter, Erik Miehling +2Hidden-State Probing Reasoning Trace Analysis
Trinity College Dublin, Ireland · IBM Research, Dublin, Ireland · Umeå University, Sweden
Confidence-Calibrated Inference Expansion for Evaluator-Guided Test-Time Reasoning
Oct 14, 2026Weida Liang, Kenji KawaguchiEfficient Language Model Reasoning Inference-Time Compute Allocation
School of Computing, National University of Singapore
Can an Agent Learn to Predict Its Own Failure? Outcome-Grounded Confidence Calibration in LLM Agents
Oct 13, 2026Ashutosh HathidaraConfidence Estimation in Language Models LLM Agent Reliability
SAP
Does a Governed Confidence Score Know When a Text-to-SQL Answer Is Right? An Exploratory Study on BIRD and a Pre-registered Confirmation on Spider
Oct 12, 2026Raghu Venkat, Tricha AnjaliConfidence Estimation in Language Models Selective Prediction
Numberz.ai Inc., Alpharetta, Georgia, USA
Towards Universal Calibration for In-Context Classification
Oct 12, 2026Awik Dhar, Sunita Sarawagi, Arpit AgarwalLanguage Model Calibration Post-Hoc Calibration
Indian Institute of Technology Bombay
Strong yet Controllable LLM Abstention without Learning to Abstain
Oct 12, 2026Junwei Zhang, Yifei Wang, Wanli Yang +6Factual QA LLM Abstention
State Key Laboratory of AI Safety, Institute of Computing Technology, CAS · University of Chinese Academy of Sciences · Beijing Institute of Technology

Oct 29, 2026J/K move · Enter open · S save

SCARL: Sequence Confidence-Aligned Reinforcement Learning

Zhenghong Ke, Jingbo Zhou, Hua Wu

Baidu Frontier Research Department · Nanjing University

Reinforcement Learning RL for Language Model Reasoning Reinforcement Learning with Verifiable Rewards Confidence Estimation in Language Models

Open paper AskPDF arXiv

Reinforcement learning with verifiable rewards (RLVR) improves language-model reasoning by scoring sampled responses with a rule-based verifier and reinforcing those that pass. Where a classifier's training couples predicted probability to labels, nothing ties an RLVR policy's confidence in a response to its reward: aligned training would drive responses with the same advantage toward equal confidence, yet the advantage is blind to confidence. Under this sequence-confidence misalignment, a success the policy barely found earns the same update as one it produces with near certainty, and a confident error is corrected no harder than an exploratory one. We propose SCARL (Sequence Confidence-Aligned Reinforcement Learning), which reshapes, but never re-signs, each advantage with the alignment gap between the response's group-relative confidence and the outcome its advantage certifies, normalized into shares of each class's advantage budget. Misaligned samples absorb the force, responses with the same advantage are driven toward equal confidence, and repairing the misalignment doubles as a remedy for the entropy collapse that erodes exploration. SCARL provably reduces to its base estimator when a class's confidences agree, adds no extra passes and one hyperparameter. On Qwen3-4B-Base it lifts average Pass@32 by 2.4 points over DAPO and is best among nine compute-matched methods on both coverage and per-draw accuracy; applied unchanged across four estimators and four backbones, it improves 14 of 16 settings while preserving policy entropy and widening the confidence-correctness separation. Our code is available at https://github.com/kezhh163/scarl.