Language Model Error Detection

Follow this topic Ask about this topic

Language model error detection studies how to recognize mistakes such as hallucinated facts, faulty reasoning, incorrect code or tool calls, and errors in specialized domains. Methods use output signals such as token probabilities, entropy, and agreement across samples or models, as well as internal activations, probes, retrieval, execution tests, and formal verifiers. Benchmarks span question answering, mathematical reasoning, coding, clinical and legal tasks, and multilingual settings, but results can be distorted by benchmark artifacts or fail to transfer across models and domains. A central challenge is that detecting an error does not ensure that a system can correct it: confident mistakes may be hard to observe, while interventions can introduce new errors.

Latest papers

91 papers21 papers in the last four weeks, up 110% on the four weeks before. 0.1% of all new papers.

Legible Failures: Detecting and Repairing In-Context Binding Errors
Sep 10, 2026Manas Venkata Sai Ravulapalli, Samrath Singh Chadha, Abhinav M. HariLanguage Model Error Detection LLM Reliability
Efficient Computation Inc.
Solving versus Verifying: Catching Contradictions in Tax Reasoning Systems
Sep 5, 2026Albert Sadowski, Jarosław A. ChudziakLLM Reliability Language Model Error Detection
Faculty of Electronics and Information Technology Warsaw University of Technology Warsaw, Poland
MMPCBench: Benchmarking Multimodal Large Language Models on Proactive Critique of Flawed Inputs
Aug 29, 2026Jinzhe Li, Gengxu Li, Jinnan Li +2Language Model Error Detection Multimodal Large Language Models
School of Artificial Intelligence, Jilin University · International Center for Future Science, Jilin University · Engineering Research Center of Knowledge-Driven Human-Machine Intelligence, MOE, China
The Calls are Coming from Inside the Model: Investigating Probe-based Detection of Tool-Calling Errors in LLMs
Aug 27, 2026Eric Yeats, Brendan Kennedy, Loc Truong +5Language Model Error Detection LLM Tool Use
Pacific Northwest National Laboratory · National Security Agency
Prediction of Prediction (PoP): Inter-Layer Activation Fusion for Single-Pass Hallucination Detection in Large Language Models
Aug 27, 2026Himal BaduLanguage Model Error Detection LLM Hallucination Detection
Not Just Reason, Not Just Scan: Reinforcement Learning for Proactive Scientific Error Verification over Academic Paper
Aug 27, 2026Rongjin Li, Yuanxin Liu, Hao Zhou +3Language Model Error Detection Scientific Reasoning in Language Models
Beijing University of Posts and Telecommunications · State Key Laboratory for Multimedia Information Processing, School of Computer Science, Peking University · WeChat AI, Tencent Inc., China
FaithSieve: Fine-Grained Evaluation of Math Proofs with Faithful Formal Evidence
Aug 26, 2026Ziyu Wang, Qiming Dai, Yishan Wu +1LLM Evaluation Language Model Error Detection
Academy for Advanced Interdisciplinary Studies Peking University · School of Mathematical Sciences Peking University · Theory Lab, 2012 Labs Huawei Technologies Co., Ltd. +1
From Passive Response to Proactive Correction: Enhancing LLM Robustness Against Input Fact Perturbations
Aug 26, 2026Ping Wang, Xiangguo Sun, Bingbing Xu +2Language Model Robustness LLM Hallucination Mitigation
Renmin University of China · Southeast University · Zhejiang University
Confidently Wrong, Silently So: Auditing Undetectable Failures of a Deployed On-Device Language Model
Aug 24, 2026Shashwat Pandey, Satwik Pandey, Suresh RaghuLLM Auditing LLM Reliability
University of California, Santa Cruz · Independent Researcher
Toward Better Assessment of LLMs' Performance in Clinical Error Detection
Aug 17, 2026Yifan Zhang, Rahmatollah BeheshtiClinical Language Model Evaluation LLM Evaluation
University of Delaware, Newark, DE, USA
Prior Audit-Repair Context Shifts LLM Verifier Thresholds Toward Leniency
Aug 17, 2026Parsa Mazaheri, Kasra MazaheriLanguage Model Error Detection Bias in LLM-as-a-Judge
University of California, Santa Cruz · Massachusetts Institute of Technology
ACTS-SQL: Agentic and Critic-Oriented Tree-Structured SQL Correctness with Large Language Models
Aug 15, 2026Xinmei Huang, Jie Song, Peng Li +10Text-to-SQL LLM Self-Correction
Renmin University of China · ByteDance Inc.
No Universal Signal Predicts Sample-Level LLM Regression under Version Updates
Aug 11, 2026Jia Sheng, Yiwei LuLanguage Model Error Detection LLM Reliability
University of Ottawa · Vector Institute
Reasoning Errors Have a Region and a Direction in the Residual-Stream Trajectory of LLMs
Aug 6, 2026Hamed Damirchi, Ignacio Meza De la Jara, Damith Ranasinghe +2Language Model Error Detection LLM Interpretability
Australian Institute for Machine Learning · Adelaide University · Naval Group Pacific +1
UHP Detection: LVLMs have their Unique Hallucination Pattern in the Consistency Space
Aug 4, 2026Amir Mohammad Ezzati, Kiyan Rezaee, Bardiya Kariminia +4Hallucination Detection in VLMs Vision-Language Models
Sharif University of Technology · Michigan State University · Shahid Beheshti University +2
D-Score: A Spectral Hidden-State Signal for Hallucination Detection in Large Language Models
Jul 27, 2026Bianca Raimondi, Davide Evangelista, Maurizio Gabbrielli +1Language Model Error Detection Hallucination in Language Models
Department of Computer Science and Engineering, University of Bologna, Bologna, Italy
The Knowing-Saying Gap: When Probes See Errors that Confidence Misses
Jul 21, 2026Jyotin Goel, Ipshita Bandyopadhyay, Justin ShenkLanguage Model Error Detection Confidence Estimation in Language Models
Pre-Generation Hallucination Detection in Large Language Models via Soft-Target Attention Probing
Jun 20, 2026Amina Miftakhova, Alexey ZaytsevLanguage Model Error Detection LLM Hallucination Detection
Applied AI Institute
Peeking Inside LLMs: Leveraging Internal Artifacts of LLMs for Enhancing Reliability in Legal Classification
Jun 18, 2026Sudipta Santra, Debtanu Datta, Saptarshi GhoshLegal NLP Language Model Error Detection
Indian Institute of Technology Kharagpur, Kharagpur, 721302, West Bengal, India
Measurement Under Selection: Decoy-Calibrated Failure Audits for Language Models
Jun 8, 2026Vyzantinos Repantis, Ameya Gawde, Harshvardhan SinghLLM Evaluation Language Model Error Detection
The Educational Approaches to Virtual Reality Technologies Lab, University of Ioannina, Greece · Meta Platforms
How Language Models Fail: Token-Level Signatures of Committed and Persistent Reasoning Failures
Jun 4, 2026Tanvi Thoria, Kiana Jafari, Marc R. Schlichting +1Language Model Error Detection LLM Reliability
Department of Computer Science, Stanford University · Department of Aeronautics and Astronautics, Stanford University
Ekka: Automated Diagnosis of Silent Errors in LLM Inference
Jun 3, 2026Yile Gu, Zhen Zhang, Shaowei Zhu +4Software Engineering LLM Serving
University of Washington · 2Amazon Web Services.
DECK: A Consistency x Confidence Taxonomy of LLM Hallucinations
Jun 1, 2026Mohit Singh ChauhanLanguage Model Error Detection LLM Hallucination Detection
TriLens: Per-Layer Logit-Lens Entropy for White-Box Hallucination Detection
May 31, 2026Bohan Yang, Yijun Gong, Zhi Zhang +3Language Model Error Detection LLM Reliability
Binjiang Institute of Zhejiang University · Beijing Normal-Hong Kong Baptist University · Great Bay University +1
BenHalluEval: A Multi-Task Hallucination Evaluation Framework for Large Language Models on Bengali
May 29, 2026Shefayat E Shams Adib, Ahmed Alfey Sani, Ekramul Alam Esham +3LLM Reliability Multilingual Language Model Evaluation
Department of Computer Science and Engineering, Islamic University of Technology · Department of Computer Science and Engineering, University of California
Learning the Error Patterns of Language Models
May 27, 2026Jinwoo Kim, Taylor Berg-KirkPatrick, Loris D'AntoniConstrained Decoding Language Model Error Detection
Entropy Distribution as a Fingerprint for Hallucinations in Generative Models
May 27, 2026Mattia J. Villani, Pranav Deshpande, Akshay Seshadri +2Language Model Error Detection Language Model Generation Evaluation
Global Technology Applied Research, JPMorganChase, New York, NY 10001, USA
Detection Without Correction: A Two-Parameter Decomposition of Multi-Stage LLM Pipelines
May 26, 2026Prashanti Nilayam, Kiran Ramanna, Prashil TumbadeLLM Self-Correction LLM Evaluation
Servicenow CA, USA
CyberCorrect: A Cybernetic Framework for Closed-Loop Self-Correction in Large Language Models
May 17, 2026Yuning Wu, Yingmin Liu, Yang ShuLLM Self-Correction LLM Evaluation
School of Software, Henan University, Kaifeng, China · 2Zhejiang University, Hangzhou, China
PARALLAX: Separating Genuine Hallucination Detection from Benchmark Construction Artifacts
May 16, 2026Khizar Hussain, Murat KantarciogluLLM Evaluation Language Model Error Detection
Virginia Tech

Sep 10, 2026J/K move · Enter open · S save

Legible Failures: Detecting and Repairing In-Context Binding Errors

Manas Venkata Sai Ravulapalli, Samrath Singh Chadha, Abhinav M. Hari

Efficient Computation Inc.

Language Model Error Detection LLM Reliability In-Context Learning Language Model Probing

Open paper AskPDF arXiv

A wrong answer does not show whether the model lacked the needed information or held it and failed to use it. On an entity-obligation binding task, a language model can emit an incorrect prompt-supplied binding while a linear probe can recover the correct one from its frozen hidden state. We measure how often this occurs across 16 public checkpoints, each evaluated with three seeds. We fit a probe on a training fold, select its layer on a validation fold, and report results on a disjoint test fold. On the trials each model gets wrong, probe accuracy exceeds the strict present-obligation baseline, 1/K = 0.125, by +0.196 (95% CI [+0.101, +0.296], bootstrapped over models). A query-entity counterfactual rules out token presence and recency. A score built from the sign of probe-output disagreement improves failure detection over the model's own confidence by +0.079 AUROC (95% CI [+0.036, +0.126]). Raw probe confidence gives no measurable improvement over model confidence. Steering the residual stream toward the probe-decoded binding, with no gold label, raises accuracy on all eight models tested by a mean of +0.168 (95% CI [+0.066, +0.280]). Where recent studies report that probe-detected errors are resistant to interventions, we find that in-context binding is a setting in which probes are actionable.