Interpretability

Understanding and Mitigating Hallucination in Large Vision-Language Models via Modular Attribution and Intervention ICLR 2025 Understanding and Mitigating Overrefusal in LLMs from an Unveiling Perspective of Safety Decision Boundary EMNLP 2025 Understanding Complexity in VideoQA via Visual Program Generation ICML 2025 Understanding Fixed Predictions via Confined Regions ICML 2025 Understanding GUI Agent Localization Biases through Logit Sharpness EMNLP 2025 Understanding How CodeLLMs (Mis)Predict Types with Activation Steering EMNLP 2025 Understanding How Value Neurons Shape the Generation of Specified Values in LLMs EMNLP 2025 Understanding Impact of Human Feedback via Influence Functions ACL 2025 Understanding Individual Agent Importance in Multi-Agent System via Counterfactual Reasoning AAAI 2025 Understanding Model Reprogramming for CLIP via Decoupling Visual Prompts ICML 2025 Understanding Refusal in Language Models with Sparse Autoencoders EMNLP 2025 Understanding Subword Compositionality of Large Language Models EMNLP 2025 Understanding Synthetic Context Extension via Retrieval Heads ICML 2025 Understanding the Dark Side of LLMs’ Intrinsic Self-Correction ACL 2025 Understanding the Language Model to Solve the Symbolic Multi-Step Reasoning Problem from the Perspective of Buffer Mechanism EMNLP 2025 Understanding the Logic of Direct Preference Alignment through Logic ICML 2025 Understanding the Repeat Curse in Large Language Models from a Feature Perspective ACL 2025 Understanding the Thinking Process of Reasoning Models: A Perspective from Schoenfeld’s Episode Theory EMNLP 2025 Understanding Verbatim Memorization in LLMs Through Circuit Discovery ACL 2025 UniDebugger: Hierarchical Multi-Agent Framework for Unified Software Debugging EMNLP 2025 UniT: One Document, Many Revisions, Too Many Edit Intention Taxonomies ACL 2025 Universal Sparse Autoencoders: Interpretable Cross-Model Concept Alignment ICML 2025 Unlearning-based Neural Interpretations ICLR 2025 Unlocking General Long Chain-of-Thought Reasoning Capabilities of Large Language Models via Representation Engineering ACL 2025 Unlocking the Capabilities of Large Vision-Language Models for Generalizable and Explainable Deepfake Detection ICML 2025