Interpretability

Safe-Unsafe Concept Separation Emerges from a Single Direction in Language Models Activation Space EACL 2026 Safe Vision-Language Models via Unsafe Weights Manipulation WACV 2026 SAGA: Source Attribution of Generative AI Videos CVPR 2026 SAGE: An Agentic Explainer Framework for Interpreting SAE Features in Language Models EACL 2026 SageLM: A Multi-aspect and Explainable Large Language Model for Speech Judgement AAAI 2026 SAGE: Structured Attribute-Guided Enhancement for GZSL AAAI 2026 Saliency-R1: Enforcing Interpretable and Faithful Vision-language Reasoning via Saliency-map Alignment Reward CVPR 2026 Same Attention, Different Truths: Put Logit-Lens over Visual Attention to Detect and Mitigate LVLM Object Hallucination CVPR 2026 Same Benchmark, Same Subspace: Task-Selective Convergence in LLM Representations UAI 2026 Sampling-Free Uncertainty Quantification via Hidden State Dynamics in Language Models AAAI 2026 SARA: Leveraging LLM Agents and Jurisprudential Ontologies for Automated Legal Reasoning AAAI 2026 SAVE: Sparse Autoencoder-Driven Visual Information Enhancement for Mitigating Object Hallucination WACV 2026 Say It Another Way: Auditing LLMs with a User-Grounded Automated Paraphrasing Framework EACL 2026 Scalable Vision-Guided Crop Yield Estimation AAAI 2026 SceneEval: Evaluating Semantic Coherence in Text-Conditioned 3D Indoor Scene Synthesis WACV 2026 SceneProp: Combining Neural Network and Markov Random Field for Scene-Graph Grounding WACV 2026 Scientifically-Interpretable Reasoning Network (ScIReN): Discovering Hidden Relationships in the Carbon Cycle and Beyond AAAI 2026 Scope Delineation Before Localization: A Two-Stage Framework for Enhancing Failure Attribution in Multi-Agent Systems AAAI 2026 SCoUT: A Framework for Structured Stereotype Analysis in Language Models AAAI 2026 SearchLLM: Detecting LLM Paraphrased Text by Measuring the Similarity with Regeneration of the Candidate Source via Search Engine EACL 2026 Seeing Is Believing: Rich-Context Hallucination Detection for MLLMs via Backward Visual Grounding AAAI 2026 Seeing through the Conflict: Transparent Knowledge Conflict Handling in Retrieval-Augmented Generation AAAI 2026 SEG4SEG: Identifying Systematic Failure Modes in Segmentation by Subgroup Discovery Methods MIDL 2026 SegCompass: Exploring Interpretable Alignment with Sparse Autoencoders for Enhanced Reasoning Segmentation CVPR 2026 Select, Hypothesize and Verify: Towards Verified Neuron Concept Interpretation CVPR 2026