ACL 2026 — Mechanistic Interpretability

Mechanistic Interpretability

80 papers Links not yet available — ACL proceedings pending on ACL Anthology.

• Rhetorical Questions in LLM Representations: A Linear Probing Study
• Mechanistic Interpretability Should Prioritize Feature Consistency in Sparse Autoencoders
• Make Mechanistic Interpretability Auditable: A Call to Develop Guidelines via Continuous Collaborative Reviewing
• APPSI-139: A Parallel Corpus of English Application Privacy Policy Summarization and Interpretation
• Are Emotion and Rhetoric Neurons in LLM? Neuron Recognition and Adaptive Masking for Emotion-Rhetoric Prediction Steering
• REFLEX: Self-Refining Explainable Fact-Checking via Verdict-Anchored Style Control
• Interpretable Safety Alignment via SAE-Constructed Low-Rank Subspace Adaptation
• Revitalizing Black-Box Interpretability: Actionable Interpretability for LLMs via Proxy Models
• OCP: Outlier-Centric Probing for Dynamic Structured Pruning of LLMs
• From Insight to Action: A Novel Framework for Interpretability-Guided Data Selection in Large Language Models
• A Mechanistic Account of Attention Sinks in GPT-2: One Circuit, Broader Implications for Mitigation
• SCOUT: Selective Coupling via Optimal Unbalanced Transport for Interpretable Text Classification
• TriEx: A Game-based Tri-View Framework for Explaining Internal Reasoning in Multi-Agent LLMs
• SpeechLLM-as-Judges: Towards General and Interpretable Speech Quality Evaluation
• Investigating More Explainable and Partition-Free Compositionality Estimation for LLMs: A Rule-Generation Perspective
• One Battle After Another: Probing LLMs’ Limits on Multi-Turn Instruction Following with a Benchmark Evolving Framework
• SPEAK: Spiking Neurons as an Entropy-Aware Tokenizer for Large Language Models
• Correct When Paired, Wrong When Split: Decoupling and Editing Modality-Specific Neurons in MLLMs
• Beyond Single-View Detection: A Dual-Space Reasoning Framework for Interpretable Harmful Meme Understanding
• How Do LLMs and VLMs Understand Viewpoint Rotation Without Vision? An Interpretability Study
• WebCoderBench: Benchmarking Web Application Generation with Comprehensive and Interpretable Evaluation Metrics
• Efficient Test-Time Scaling of Multi-Step Reasoning by Probing Internal States of Large Language Models
• Probing for Reading Times
• Continuous Interpretive Steering for Scalar Diversity
• ICDAGENT: Empowering Agentic Large Language Models for Explainable Medical Coding
• Discovering and Causally Validating Emotion-Sensitive Neurons in Large Audio-Language Models
• Attention Under Attack: Analog Noise Effects and Mechanistic Vulnerabilities in Transformer Models
• Lost in Translation, and Found: Detecting and Interpreting Translation Effects
• From Isolation to Entanglement: When Do Interpretability Methods Identify and Disentangle Known Concepts?
• Tracing Logit Trajectories Across Layer Depth: Dataset-Level Explainability for Language Models
• Learning What Matters: Dynamic Dimension Selection and Aggregation for Interpretable Vision-Language Reward Modeling
• Why LLMs Hallucinate on Structured Knowledge: A Mechanistic Analysis of Reasoning over Linearized Representations
• Uncovering Sentiment Analysis Circuit in Large Language Model
• Bridging Internal Consistency and External Alignment: A Causal and Dynamic Interpretability Framework for LLM Generation
• Probing the Safety Robustness of LLMs in Latent Space
• NeuReasoner: Towards Explainable, Controllable, and Unified Reasoning via Mixture-of-Neurons
• Stop Hardening Everything: A Training-Free Neuron-Level Defense for Neural Ranking Models
• MedForge: Interpretable Medical Deepfake Detection via Forgery-aware Reasoning
• MicroC-KT: Modeling Community Effect via Learning Micro-Environment for Evidence-Grounded Explainable Knowledge Tracing
• Beyond Self-Report: Bridging the Intention-Behavior Gap in Critical Thinking Assessment via Interpretable Multi-Agent System
• Probing Semantic Alignment, Lexical Invariance, and Syntactic Influence in LLM Metaphor Processing
• DVI-DTM: Dual-View Representation Learning for Interpretable Short Text Dynamic Topic Modeling
• Exploring and Distilling Multi-Dimensional Clues for Interpretable Social Bot Detection
• DEFT: Demystifying VLN Failures via a Unified Dual-View Explainability Framework for LLM-based Agents
• Embracing Anisotropy: Turning Massive Activations into Interpretable Control Knobs for Large Language Models
• Towards Interpretable Tabular Reasoning: Enhancing LLM Reasoning on Tabular Data with Pre-Constructed Logic Graph
• Explainable and Fine-Grained Safeguarding of LLM Multi-Agent Systems via Bi-Level Graph Anomaly Detection
• Neuron-Aware Active Few-Shot Learning for LLMs
• PRISM: Probing Reasoning, Instruction, and Source Memory in LLM Hallucinations
• Unlocking the Black Box of Latent Reasoning: An Interpretability-Guided Approach to Intervention
• Fiction Flows: A Replication and Reinterpretation of Narrative Sequentiality
• Experiments or Outcomes? Probing Scientific Feasibility in Large Language Models
• Towards Intrinsic Interpretability of Large Language Models: A Survey of Design Principles and Architectures
• Specializing Large Models for Oracle Bone Script Interpretation via Component-Grounded Multimodal Knowledge Augmentation
• Union-of-Experts: Neurons in Mixture-of-Experts are Secretly Routers
• Interpretable Traces, Unexpected Outcomes: Investigating the Disconnect in Trace-Based Knowledge Distillation
• Memorization, Emergence, and Explaining Reversal Failures: A Controlled Study of Relational Semantics in LLMs
• An Existence Proof for Neural Language Models That Can Explain Garden-Path Effects via Surprisal
• Probing Audio-Visual Reasoning in Multimodal Language Models through the Lens of Audio
• Towards Explainable Diagnosis: A Self-learned Explanatory Knowledge Base Approach
• R2IF: Aligning Reasoning with Decisions via Composite Rewards for Interpretable LLM Function Calling
• Vision-Language Introspection: Mitigating Overconfident Hallucinations in MLLMs via Interpretable Bi-Causal Steering
• Why Does Reinforcement Learning Generalize? A Feature-Level Mechanistic Study of Post-Training in Large Language Models
• SemCSE-Multi: Multifaceted and Decodable Embeddings for Aspect-Specific and Interpretable Scientific Domain Mapping
• From Nodes to Narratives: Explaining Graph Neural Networks with LLMs and Graph Context
• Constructing Interpretable Features from Compositional Neuron Groups
• LogosKG: Hardware-Optimized Scalable and Interpretable Knowledge Graph Retrieval
• CoDial: Interpretable Task-Oriented Dialogue Systems Through Dialogue Flow Alignment
• Explain the Synth: Interpretable Evaluation of LLM Data Synthesis
• Locate and Explain: Joint Multimodal Emotion Cause Extraction and Summarization in Conversation
• A Lightweight Explainable Guardrail for Prompt Safety
• Explainable Disentangled Representation Learning for Generalizable Authorship Attribution in the Era of Generative AI
• BanHADEX: Towards Explainable HAte Speech Detection in Bangla Using Human Annotated EXplanation
• Explaining Sources of Uncertainty in Automated Fact-Checking
• PRISMA: Preference-Reinforced Self-Training Approach for Interpretable Emotionally Intelligent Negotiation Dialogues
• Interpretable Coreference Resolution Evaluation Using Explicit Semantics
• TabReX: Tabular Referenceless eXplainable Evaluation
• Preference Heads in Large Language Models: A Mechanistic Framework for Interpretable Personalization
• Is Chain-of-Thought Really Not Explainability? Chain-of-Thought Can Be Faithful without Hint Verbalization
• Towards a Mechanistic Understanding of Large Reasoning Models: A Survey of Training, Inference, and Failures