The papers

Most central to this idea first, not most recent.

• Chinese Techcs.CL

  Token-level certainty predicts LLM question difficulty better than response correctness

  Shanghai Jiao Tong University, Alibaba Group · Oct 2026

• Research Labcs.AI

  Jailbreak prompts improve uncertainty calibration in black-box reasoning models

  Petscraft, Inria · Sept 2026

• Top Universitycs.AI

  Teaching Models Confidence Can Shorten Reasoning Without Explicit Stop Training

  University of Maryland, AI Foundations, Capital One · Sept 2026

• Chinese Techcs.LG

  Differentiable confidence readouts improve reasoning accuracy and calibration together

  University of Science and Technology of China, Qwen Business Unit of Alibaba · Sept 2026

• Research Labcs.CL

  Companion model sharpens LLM confidence without harming task performance

  State Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences · Sept 2026

• Big Techcs.LG

  Prediction error helps Decision Transformers reject unreliable rollout contexts

  International Institute of Information Technology Bangalore, IBM Research Bangalore · Sept 2026

• Industrycs.LGcode

  Training Forecasters to Search Improves Calibration and Reduces Tool Use

  Future Principle · Oct 2026

• Research Labcs.LG

  Event graphs improve LLM stock forecasts across two major markets

  Zircon Security, Utrecht University · Oct 2026

• Top Universitycs.AI

  Typed selective control cuts strong-model calls while preserving agent success

  Nanyang Technological University · Sept 2026

• Academiccs.AI

  Reasoning makes language models repeat the same wrong answers

  Oklahoma State University · Sept 2026

• Research Labcs.AIcode

  Agent performance depends more on system configuration than model size

  Technical University of Munich, MCML · Oct 2026

• Big Techcs.CL

  Reconstructing text under safety labels produces more reliable guard models

  University of Neuchâtel, Delft University of Technology · Sept 2026

• Academiccs.CL

  Short natural context can flip decision models to wrong answers

  University of Southern California · Sept 2026

• Top Universitycs.CR

  Unverified state text can flip calibrated model decisions

  Institute of Information Engineering, Chinese Academy of Sciences, University of Chinese Academy of Sciences · Sept 2026

• Independentcs.LG

  Fixed LLM judges can mismeasure upgrades as agents change

  Sept 2026