mlsanityData to 7 Oct 2026

Steer LLM Latents for Hallucination Detection

arXiv 2503.01917 ↗First posted 1 Mar 2025cs.LG cs.AI cs.CLcited by 49

Not on the map. Why?

Abstract (arXiv)

Hallucinations in LLMs pose a significant concern to their safe deployment in real-world applications. Recent approaches have leveraged the latent space of LLMs for hallucination detection, but their embeddings, optimized for linguistic coherence rather than factual accuracy, often fail to clearly separate truthful and hallucinated content. To this end, we propose the Truthfulness Separator Vector (TSV), a lightweight and flexible steering vector that reshapes the LLM's representation space during inference to enhance the separation between truthful and hallucinated outputs, without altering model parameters. Our two-stage framework first trains TSV on a small set of labeled exemplars to form compact and well-separated clusters. It then augments the exemplar set with unlabeled LLM generations, employing an optimal transport-based algorithm for pseudo-labeling combined with a confidence-based filtering process. Extensive experiments demonstrate that TSV achieves state-of-the-art performance with minimal labeled data, exhibiting strong generalization across datasets and providing a practical solution for real-world LLM applications.

Key facts. Cited by 49 papers in the map (data through 7 Oct 2026). Builds on 49 of its 62 references that are in the map. First posted on arXiv 1 Mar 2025.

Builds on

49 of 62 references.

References · 49

  1. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  2. 2024The Llama 3 Herd of Modelscited by 14,192
  3. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  4. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  5. 2023Visual Instruction Tuningcited by 7,792
  6. 2024Qwen2.5 Technical Reportcited by 4,713
  7. 2024GPT-4o System Cardcited by 4,467
  8. 2021TruthfulQA: Measuring How Models Mimic Human Falsehoodscited by 2,919
  9. 2017TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehensioncited by 2,895
  10. 2020Unsupervised Learning of Visual Features by Contrasting Cluster Assignmentscited by 2,845
  11. 2023MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Modelscited by 2,689
  12. 2023A Survey of Large Language Modelscited by 2,537
37 more
  1. 2013Sinkhorn Distances: Lightspeed Computation of Optimal Transportation Distancescited by 2,277
  2. 2022Language Models (Mostly) Know What They Knowcited by 1,783
  3. 2023Evaluating Object Hallucination in Large Vision-Language Modelscited by 1,696
  4. 2023A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questionscited by 1,582
  5. 2023Inference-Time Intervention: Eliciting Truthful Answers from a Language Modelcited by 1,259
  6. 2020BLEURT: Learning Robust Metrics for Text Generationcited by 1,196
  7. 2023Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMscited by 946
  8. 2024Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMscited by 822
  9. 2023Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generationcited by 795
  10. 2023SelfCheckGPT: Zero-Resource Black-Box Hallucination Detection for Generative Large Language Modelscited by 787
  11. 2017Crowdsourcing Multiple Choice Science Questionscited by 742
  12. 2022Discovering Latent Knowledge in Language Models Without Supervisioncited by 733
  13. 2022Teaching Models to Express Their Uncertainty in Wordscited by 731
  14. 2023Aligning Large Multimodal Models with Factually Augmented RLHFcited by 703
  15. 2023Siren's Song in the AI Ocean: A Survey on Hallucination in Large Language Modelscited by 695
  16. 2023The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasetscited by 673
  17. 2018Object Hallucination in Image Captioningcited by 661
  18. 2023The Internal State of an LLM Knows When It's Lyingcited by 644
  19. 2023Mitigating Hallucination in Large Multi-Modal Models via Robust Instruction Tuningcited by 439
  20. 2024INSIDE: LLMs' Internal States Retain the Power of Hallucination Detectioncited by 317
  21. 2023Generating with Confidence: Uncertainty Quantification for Black-box Large Language Modelscited by 286
  22. 2024Semantic Entropy Probes: Robust and Cheap Hallucination Detection in LLMscited by 246
  23. 2024Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?cited by 227
  24. 2022Out-of-Distribution Detection and Selective Generation for Conditional Language Modelscited by 193
  25. 2023Med-HALT: Medical Domain Hallucination Test for Large Language Modelscited by 156
  26. 2024Devils in Middle Layers of Large Vision-Language Models: Interpreting, Detecting and Mitigating Object Hallucinations via Attention Lenscited by 139
  27. 2024Interpreting and Editing Vision-Language Representations to Mitigate Hallucinationscited by 103
  28. 2024HaloScope: Harnessing Unlabeled LLM Generations for Hallucination Detectioncited by 96
  29. 2024LoFiT: Localized Fine-tuning on LLM Representationscited by 56
  30. 2024Reducing Hallucinations in Vision-Language Models via Latent Space Steeringcited by 56
  31. 2024Nullu: Mitigating Object Hallucinations in Large Vision-Language Models via HalluSpace Projectioncited by 46
  32. 2023Truth Forest: Toward Multi-Scale Truthfulness in Large Language Models through Intervention without Tuningcited by 45
  33. 2024Characterizing Truthfulness in Large Language Model Generations with Local Intrinsic Dimensioncited by 45
  34. 2025A Unified Understanding and Evaluation of Steering Methodscited by 40
  35. 2024Prism: A Framework for Decoupling and Assessing the Capabilities of VLMscited by 32
  36. 2024ICT: Image-Object Cross-Level Trusted Intervention for Mitigating Object Hallucination in Large Vision-Language Modelscited by 28
  37. 2025HalluEntity: Benchmarking and Understanding Entity-Level Hallucination Detectioncited by 13

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 18; 2026: 31.

49 papers.

Sort:

By area:

  1. 2025SAFE: Multitask Failure Detection for Vision-Language-Action Modelscited by 74
  2. 2025Large Language Models Hallucination: A Comprehensive Surveycited by 45
  3. 2026DynHD: Hallucination Detection for Diffusion Large Language Models via Denoising Dynamics Deviation Learningcited by 13
  4. 2025GLSim: Detecting Object Hallucinations in LVLMs via Global-Local Similaritycited by 11
  5. 2025LUMINA: Detecting Hallucinations in RAG System with Context-Knowledge Signalscited by 8
  6. 2025Investigating CoT Monitorability in Large Reasoning Modelscited by 8
  7. 2025TraceDet: Hallucination Detection from the Decoding Trace of Diffusion Large Language Modelscited by 7
  8. 2026TDGNet: Hallucination Detection in Diffusion Language Models via Temporal Dynamic Graphscited by 7
  9. 2025Hallucination reduction with CASAL: Contrastive Activation Steering For Amortized Learningcited by 6
  10. 2026Hide-and-Seek in Trajectories: Discovering Failure Signals for VLA Runtime Monitoringcited by 6
  11. 2026VAUQ: Vision-Aware Uncertainty Quantification for LVLM Self-Evaluationcited by 5
  12. 2026Curveball Steering: The Right Direction To Steer Isn't Always Linearcited by 5
37 more
  1. 2025HARP: Hallucination Detection via Reasoning Subspace Projectioncited by 4
  2. 2025LLM Hallucination Detection: A Fast Fourier Transform Method Based on Hidden Layer Temporal Signalscited by 4
  3. 2026Beyond Steering Vector: Flow-based Activation Steering for Inference-Time Interventioncited by 4
  4. 2026Student Guides Teacher: Weak-to-Strong Inference via Spectral Orthogonal Explorationcited by 3
  5. 2026Steering Large Reasoning Models towards Concise Reasoning via Flow Matchingcited by 3
  6. 2025Shaking to Reveal: Perturbation-Based Detection of LLM Hallucinationscited by 2
  7. 2025Decoding Memories: An Efficient Pipeline for Self-Consistency Hallucination Detectioncited by 2
  8. 2025TPA: Next Token Probability Attribution for Detecting Hallucinations in RAGcited by 2
  9. 2026LatentRefusal: Latent-Signal Refusal for Unanswerable Text-to-SQL Queriescited by 2
  10. 2026Hallucination Basins: A Dynamic Framework for Understanding and Controlling LLM Hallucinationscited by 2
  11. 2025Steering Conceptual Bias via Transformer Latent-Subspace Activationcited by 1
  12. 2025Grounding the Ungrounded: A Spectral-Graph Framework for Quantifying Hallucinations in Multimodal LLMscited by 1
  13. 2025LLM Hallucination Detection: HSADcited by 1
  14. 2026AP-OOD: Attention Pooling for Out-of-Distribution Detectioncited by 1
  15. 2026SAKED: Mitigating Hallucination in Large Vision-Language Models via Stability-Aware Knowledge Enhanced Decodingcited by 1
  16. 2026GRADE: Probing Knowledge Gaps in LLMs through Gradient Subspace Dynamicscited by 1
  17. 2026HIVE: Hidden-Evidence Verification for Hallucination Detection in Diffusion Large Language Modelscited by 1
  18. 2026OpenHalDet: A Unified Benchmark for Hallucination Detection across Diverse Generation Scenarioscited by 1
  19. 2026Does Bielik Know What It Doesn't Know? Activation Dispersion Separates Entity Familiarity from Factual Reliability Across Model Scalecited by 1
  20. 2025MMD-Flagger: Leveraging Maximum Mean Discrepancy to Detect Hallucinationscited by 0
  21. 2025AI-Generated Content in Cross-Domain Applications: Research Trends, Challenges and Propositionscited by 0
  22. 2025Training-free Truthfulness Detection via Sparse MLP Value Vectorscited by 0
  23. 2026Harnessing Reasoning Trajectories for Hallucination Detection via Answer-agreement Representation Shapingcited by 0
  24. 2026Disentangling Deception and Hallucination Failures in LLMscited by 0
  25. 2026Mind the Shift: Decoding Monetary Policy Stance from FOMC Statements with Large Language Modelscited by 0
  26. 2026Two Birds, One Projection: Harmonizing Safety and Utility in LVLMs via Inference-time Feature Projectioncited by 0
  27. 2026OSCAR: Orchestrated Self-verification and Cross-path Refinementcited by 0
  28. 2026Logical Consistency as a Bridge: Improving LLM Hallucination Detection via Label Constraint Modeling between Responses and Self-Judgmentscited by 0
  29. 2026Steer-to-Detect: Probing Hidden Representations for Detection of LLM-Generated Textscited by 0
  30. 2026Micro-Macro Retrieval: Reducing Long-Form Hallucination in Large Language Modelscited by 0
  31. 2026FLaG: Fine-Grained Latent Grouping for Hallucination Detectioncited by 0
  32. 2026Graded Entity-Familiarity Readouts in Language Models: Polish Adaptation, Cross-Language Robustness, and Refusal Steeringcited by 0
  33. 2026TRE: Training-Free Hallucination Detection for Diffusion Language Modelscited by 0
  34. 2026DeMTS: Denoising Trajectories as Multivariate Time Series for Hallucination Detection in Diffusion Language Modelscited by 0
  35. 2026Evidence-Aligned Entity Verification for Hallucination Detection in Retrieval-Augmented Generationcited by 0
  36. 2026Local Sparsity Enables Unsupervised LLM Safety Detectioncited by 0
  37. 2026When Helpful Text Hurts: Option-Redirecting Bias in Vision-Language Modelscited by 0