Language Model Auditing

Follow this topic Ask about this topic

Language model auditing examines whether a model’s behavior, internal signals, training history, and deployed tools match claims about its reliability, safety, fairness, and compliance. Auditors use controlled prompts and counterfactual tests, multi-turn and domain-specific benchmarks, black-box output or likelihood analysis, white-box activation probes, and execution logs or claim-level evidence. These methods are applied to settings such as jailbreaks, retrieval-augmented generation, agent workflows, demographic bias, data leakage, and finetuning. A central challenge is validity: proxy scores and automated judges can miss meaningful failures or produce false alarms, so audits need calibrated thresholds, matched controls, realistic settings, and evidence that can be independently checked.

Latest papers

160 papers16 papers in the last four weeks, down 43% on the four weeks before. 0.1% of all new papers.

Identifying and Mitigating Gender Cues in Academic Recommendation Letters: An Interpretability Case Study
Apr 14, 2026Charlotte S. Alexander, Shane Storks, Souradip Pal +4Gender Bias in Language Models Language Model Auditing
Scheller College of Business, Georgia Institute of Technology, Atlanta, GA 30308, USA · Weinberg Institute for Cognitive Science, University of Michigan - Ann Arbor, Ann Arbor, MI 48109, USA · School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN 47906, USA +3
Rater State Bias in RLHF Preference Data: An Audit Framework
Apr 14, 2026Elena Kopteva, Vitaliy Hlynianyi-ZhukHuman Preference Evaluation Human-in-the-Loop Annotation
The Grainger College of Engineering, Department of Physics & Illinois Center for Advanced Studies of the Universe, University of Illinois Urbana-Champaign, Urbana, Illinois 61801, USA · Faculty of Applied Mathematics, Oles Honchar Dnipro National University, Dnipro, 49045, Ukraine · Department of Clinical Psychology, Kyiv Institute of Modern Psychology and Psychotherapy, Kyiv, 01133, Ukraine
Functional Emotions or Situational Contexts? A Discriminating Test from the Mythos Preview System Card
Apr 9, 2026Hiranya V. PeirisEmotion Representation in Language Models Language Model Auditing
Institute of Astronomy & Kavli Institute for Cosmology, University of Cambridge, Madingley Road, Cambridge CB3 0HA, UK · Cavendish Laboratory, Department of Physics, University of Cambridge, JJ Thomson Avenue, Cambridge, CB3 0HE, UK
LatentAudit: Real-Time White-Box Faithfulness Monitoring for Retrieval-Augmented Generation with Verifiable Deployment
Apr 7, 2026Zhe Yu, Wenpeng Xing, Meng HanLLM Auditing RAG Evaluation
Binjiang Institute of Zhejiang University · Communication University of Zhejiang · Zhejiang University
When Only the Final Text Survives: Implicit Execution Tracing for Multi-Agent Auditing
Mar 18, 2026Yi Nian, Haosen Cao, Shenzhe Zhu +4Data Provenance in Language Models AI Agent Auditing
University of Southern California · University of Toronto · University of Illinois Chicago +1
Neuro-Symbolic Financial Reasoning via Deterministic Fact Ledgers and Adversarial Low-Latency Hallucination Detector
Mar 4, 2026Pedram AgandFinancial Reasoning in Language Models Neuro-Symbolic Reasoning
FactAI Lab, Vancouver, BC, Canada.
AuditBench: Evaluating Alignment Auditing Techniques on Models with Hidden Behaviors
Feb 26, 2026Abhay Sheshadri, Aidan Ewart, Elias Kempf +8Benchmark Auditing AI Agent Evaluation
Anthropic Fellows Program · Anthropic
FinReasoning: A Hierarchical Benchmark for Reliable Financial Research Reporting
Feb 25, 2026Yiyun Zhu, Yidong Jiang, Ziwen Xu +4Financial Reasoning in Language Models Financial Reasoning Benchmarks
School of Computer Science and Technology, Tongji University, Shanghai, China · Shanghai Artificial Intelligence Laboratory, Shanghai, China
All Leaks Count, Some Count More: Interpretable Temporal Contamination Detection and Mitigation in LLM Backtesting
Feb 19, 2026Zeyu Zhang, Ryan Chen, Bradly C. StadieData Contamination in Language Models Temporal Data Leakage
Department of Statistics and Data Science, Northwestern University · Bridgewater AIA Labs
The Emergence of Lab-Driven Alignment Signatures: A Psychometric Framework for Auditing Latent Bias and Compounding Risk in Generative AI
Feb 19, 2026Dusan BosnjakovicLanguage Model Auditing Language Model Bias Evaluation

Apr 14, 2026J/K move · Enter open · S save

Identifying and Mitigating Gender Cues in Academic Recommendation Letters: An Interpretability Case Study

Charlotte S. Alexander, Shane Storks, Souradip Pal, Sayak Chakrabarty+3

Scheller College of Business, Georgia Institute of Technology, Atlanta, GA 30308, USA · Weinberg Institute for Cognitive Science, University of Michigan - Ann Arbor, Ann Arbor, MI 48109, USA · School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN 47906, USA+3

Gender Bias in Language Models Language Model Auditing

Open paper AskPDF arXiv

Letters of recommendation (LoRs) can carry patterns of implicitly gendered language that can inadvertently influence downstream decisions, e.g. in hiring and admissions. In this work, we investigate the extent to which Transformer-based encoder models as well as Large Language Models (LLMs) can infer the gender of applicants in academic LoRs submitted to an U.S. medical-residency program after explicit identifiers like names and pronouns are de-gendered. While using three models (DistilBERT, RoBERTa, and Llama 2) to classify the gender of anonymized and de-gendered LoRs, significant gender leakage was observed as evident from up to 68% classification accuracy. Text interpretation methods, like TF-IDF and SHAP, demonstrate that certain linguistic patterns are strong proxies for gender, e.g. "emotional'' and "humanitarian'' are commonly associated with LoRs from female applicants. As an experiment in creating truly gender-neutral LoRs, these implicit gender cues were remove resulting in a drop of up to 5.5% accuracy and 2.7% macro

F_1

score on re-training the classifiers. However, applicant gender prediction still remains better than chance. In this case study, our findings highlight that 1) LoRs contain gender-identifying cues that are hard to remove and may activate bias in decision-making and 2) while our technical framework may be a concrete step toward fairer academic and professional evaluations, future work is needed to interrogate the role that gender plays in LoR review. Taken together, our findings motivate upstream auditing of evaluative text in real-world academic letters of recommendation as a necessary complement to model-level fairness interventions.