{"total":69,"rows":[{"id":"2404.14082","title":"Mechanistic Interpretability for AI Safety -- A Review","created":"2024-04-22","cited_by_count":422,"area":"mechinterp"},{"id":"2401.06102","title":"Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models","created":"2024-01-11","cited_by_count":235,"area":"mechinterp"},{"id":"2407.02646","title":"A Practical Review of Mechanistic Interpretability for Transformer-Based Language Models","created":"2024-07-02","cited_by_count":118,"area":"mechinterp"},{"id":"2401.12181","title":"Universal Neurons in GPT2 Language Models","created":"2024-01-22","cited_by_count":102,"area":"mechinterp"},{"id":"2402.09221","title":"Spectral Filters, Dark Signals, and Attention Sinks","created":"2024-02-14","cited_by_count":56,"area":"mechinterp"},{"id":"2410.19750","title":"The Geometry of Concepts: Sparse Autoencoder Feature Structure","created":"2024-10-10","cited_by_count":52,"area":"mechinterp"},{"id":"2506.03093","title":"From Flat to Hierarchical: Extracting Sparse Representations with Matching Pursuit","created":"2025-06-03","cited_by_count":49,"area":"mechinterp"},{"id":"2411.04430","title":"Towards Unifying Interpretability and Control: Evaluation via Intervention","created":"2024-11-07","cited_by_count":32,"area":"mechinterp"},{"id":"2502.01615","title":"Large Language Models Are Human-Like Internally","created":"2025-02-03","cited_by_count":29,"area":"mechinterp"},{"id":"2410.09223","title":"The Same But Different: Structural Similarities and Differences in Multilingual Language Modeling","created":"2024-10-11","cited_by_count":28,"area":"mechinterp"},{"id":"2601.14004","title":"Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models","created":"2026-01-20","cited_by_count":27,"area":"mechinterp"},{"id":"2407.19200","title":"On Behalf of the Stakeholders: Trends in NLP Model Interpretability in the Era of LLMs","created":"2024-07-27","cited_by_count":25,"area":"mechinterp"},{"id":"2503.11667","title":"LogitLens4LLMs: Extending Logit Lens Analysis to Modern Large Language Models","created":"2025-02-24","cited_by_count":23,"area":"mechinterp"},{"id":"2502.17601","title":"Representation Engineering for Large-Language Models: Survey and Research Challenges","created":"2025-02-24","cited_by_count":21,"area":"mechinterp"},{"id":"2409.04185","title":"Residual Stream Analysis with Multi-Layer SAEs","created":"2024-09-06","cited_by_count":19,"area":"mechinterp"},{"id":"2501.10573","title":"The Intrinsic Dimension of Prompts in Internal Representations of Large Language Models","created":"2025-01-17","cited_by_count":17,"area":"mechinterp"},{"id":"2508.15882","title":"Beyond Transcription: Mechanistic Interpretability in ASR","created":"2025-08-21","cited_by_count":13,"area":"mechinterp"},{"id":"2502.07424","title":"RomanLens: The Role Of Latent Romanization In Multilinguality In LLMs","created":"2025-02-11","cited_by_count":12,"area":"mechinterp"},{"id":"2509.13316","title":"Do Activation Verbalization Methods Convey Privileged Information?","created":"2025-09-16","cited_by_count":10,"area":"mechinterp"},{"id":"2505.20229","title":"From What to How: Attributing CLIP's Latent Components Reveals Unexpected Semantic Reliance","created":"2025-05-26","cited_by_count":6,"area":"mechinterp"},{"id":"2511.19265","title":"Unboxing the Black Box: Mechanistic Interpretability for Algorithmic Understanding of Neural Networks","created":"2025-11-24","cited_by_count":6,"area":"mechinterp"},{"id":"2604.02608","title":"Steerable but Not Decodable: Function Vectors Operate Beyond the Logit Lens","created":"2026-04-03","cited_by_count":6,"area":"mechinterp"},{"id":"2603.12298","title":"Global Evolutionary Steering: Refining Activation Steering Control via Cross-Layer Consistency","created":"2026-03-12","cited_by_count":5,"area":"mechinterp"},{"id":"2407.13594","title":"Validating Mechanistic Interpretations: An Axiomatic Approach","created":"2024-07-18","cited_by_count":4,"area":"mechinterp"},{"id":"2604.15557","title":"Predicting Where Steering Vectors Succeed","created":"2026-04-16","cited_by_count":4,"area":"mechinterp"},{"id":"2605.06076","title":"Navigating by Old Maps: The Pitfalls of Static Mechanistic Localization in LLM Post-Training","created":"2026-05-07","cited_by_count":4,"area":"mechinterp"},{"id":"2606.11386","title":"Overcoming State Inertia in Full-Duplex Spoken Language Models via Activation Steering","created":"2026-06-09","cited_by_count":4,"area":"mechinterp"},{"id":"2503.02078","title":"Superscopes: Amplifying Internal Feature Representations for Language Model Interpretation","created":"2025-03-03","cited_by_count":3,"area":"mechinterp"},{"id":"2507.03668","title":"TRACE: Training and Inference-Time Interpretability Analysis for Language Models","created":"2025-07-04","cited_by_count":3,"area":"mechinterp"},{"id":"2509.25045","title":"Hyperdimensional Probe: Decoding LLM Representations via Vector Symbolic Architectures","created":"2025-09-29","cited_by_count":3,"area":"mechinterp"},{"id":"2510.04108","title":"Can Linear Probes Measure LLM Uncertainty?","created":"2025-10-05","cited_by_count":3,"area":"mechinterp"},{"id":"2602.09783","title":"Why Linear Interpretability Works: Invariant Subspaces as a Result of Architectural Constraints","created":"2026-02-10","cited_by_count":3,"area":"mechinterp"},{"id":"2606.06267","title":"Many Circuits, One Mechanism: Input Variation and Evaluation Granularity in Circuit Discovery","created":"2026-06-04","cited_by_count":3,"area":"mechinterp"},{"id":"2609.07876","title":"LLM Layers Immediately Correct Each Other","created":"2026-09-07","cited_by_count":3,"area":"mechinterp"},{"id":"2504.13763","title":"Decoding Vision Transformers: the Diffusion Steering Lens","created":"2025-04-18","cited_by_count":2,"area":"mechinterp"},{"id":"2605.27958","title":"Pressure-Testing Deception Probes in LLMs: Scaling, Robustness, and the Geometry of Deceptive Representations","created":"2026-05-27","cited_by_count":2,"area":"mechinterp"},{"id":"2312.03140","title":"FlexModel: A Framework for Interpretability of Distributed Large Language Models","created":"2023-12-05","cited_by_count":1,"area":"mechinterp"},{"id":"2601.09049","title":"Is Grokking Worthwhile? Functional Analysis and Transferability of Generalization Circuits in Transformers","created":"2026-01-14","cited_by_count":1,"area":"mechinterp"},{"id":"2601.22447","title":"Beyond Activation Patterns: A Weight-Based Out-of-Context Explanation of Sparse Autoencoder Features","created":"2026-01-30","cited_by_count":1,"area":"mechinterp"},{"id":"2603.13259","title":"How Transformers Reject Wrong Answers: Rotational Dynamics of Factual Constraint Processing","created":"2026-02-25","cited_by_count":1,"area":"mechinterp"},{"id":"2606.00356","title":"How Far Do Auto-Interpretation Labels Generalize: A Controlled Study Across Languages, Scripts, and Rewordings","created":"2026-05-29","cited_by_count":1,"area":"mechinterp"},{"id":"2606.28050","title":"Can LLMs Judge Better Than They Generate? Evaluating Task Asymmetry, Mechanistic Interpretability and Transferability for In-Context QA","created":"2026-06-26","cited_by_count":1,"area":"mechinterp"},{"id":"2607.16693","title":"Are Arithmetic Heuristic Neurons Form-Invariant? A Mechanistic Analysis of Symbols, Text, and Code in LLMs","created":"2026-07-18","cited_by_count":1,"area":"mechinterp"},{"id":"2607.28906","title":"Token-Level Diagnosis of Sycophancy in LLMs with Attribution-Guided Steering","created":"2026-07-31","cited_by_count":1,"area":"mechinterp"},{"id":"2608.22985","title":"What Does Activation Steering Control? Attribution Across Answer Encodings and Output-Sensitive Subspaces","created":"2026-08-24","cited_by_count":1,"area":"mechinterp"},{"id":"2609.05575","title":"Capsule Lens: Locating and Tracking Concept Geometry in Model Representations","created":"2026-09-04","cited_by_count":1,"area":"mechinterp"},{"id":"2609.22135","title":"Read-Best Is Not Steer-Best: A Probing--Steering Layer Dissociation in Omni-Modal Large Language Models","created":"2026-08-24","cited_by_count":1,"area":"mechinterp"},{"id":"2502.15090","title":"ExpertLens: Activation steering features are highly interpretable","created":"2025-02-20","cited_by_count":0,"area":"mechinterp"},{"id":"2510.11390","title":"Medical Interpretability and Knowledge Maps of Large Language Models","created":"2025-10-13","cited_by_count":0,"area":"mechinterp"},{"id":"2602.13567","title":"DistillLens: Symmetric Knowledge Distillation Through Logit Lens","created":"2026-02-14","cited_by_count":0,"area":"mechinterp"},{"id":"2603.30002","title":"Tracking Equivalent Mechanistic Interpretations Across Neural Networks","created":"2026-03-31","cited_by_count":0,"area":"mechinterp"},{"id":"2604.06483","title":"Distributed Interpretability and Control for Large Language Models","created":"2026-04-07","cited_by_count":0,"area":"mechinterp"},{"id":"2604.11467","title":"From Attribution to Action: A Human-Centered Application of Activation Steering","created":"2026-04-13","cited_by_count":0,"area":"mechinterp"},{"id":"2604.27167","title":"How a Cooperative-Override Circuit Suppresses Nash Play in Large Language Models","created":"2026-04-29","cited_by_count":0,"area":"mechinterp"},{"id":"2605.08295","title":"In-Context Fixation: When Demonstrated Labels Override Semantics in Few-Shot Classification","created":"2026-05-08","cited_by_count":0,"area":"mechinterp"},{"id":"2606.07617","title":"Query Lens: Interpreting Sparse Key-Value Features with Indirect Effects","created":"2026-05-30","cited_by_count":0,"area":"mechinterp"},{"id":"2606.27941","title":"VASAE: Naming SAE Dictionary Directions with Vocabulary-Aligned Anchoring","created":"2026-06-26","cited_by_count":0,"area":"mechinterp"},{"id":"2606.32008","title":"Surrogate Fidelity: When Can Open LLMs Explain Closed Ones?","created":"2026-06-30","cited_by_count":0,"area":"mechinterp"},{"id":"2607.00083","title":"Harnessing the Latent Space: From Steering Vectors to Model Calibrators for Control and Trust","created":"2026-06-30","cited_by_count":0,"area":"mechinterp"},{"id":"2608.02486","title":"Cultural Awareness is Represented but Not Decoded: Tracing Mythological Knowledge across 18 Open-Source LLMs","created":"2026-08-03","cited_by_count":0,"area":"mechinterp"},{"id":"2608.05160","title":"The Ignition Index: Measuring Global Workspace Dynamics in Language Models","created":"2026-05-26","cited_by_count":0,"area":"mechinterp"},{"id":"2608.06429","title":"Recovering Lesion Parameters from Aphasic Picture Naming Error Profiles in Large Language Models","created":"2026-08-05","cited_by_count":0,"area":"mechinterp"},{"id":"2608.06638","title":"MI-MIDI: Mechanistic Interpretability of Text-to-MIDI Generation Models via Probing, Lenses and Steering","created":"2026-08-06","cited_by_count":0,"area":"mechinterp"},{"id":"2608.24758","title":"RACE: Scalable Statistical Estimation of Functional Consistency in LLM Neurons","created":"2026-08-25","cited_by_count":0,"area":"mechinterp"},{"id":"2609.06879","title":"AutoLexSteer: Automatic Contrast Construction for Lexical Activation Steering","created":"2026-09-06","cited_by_count":0,"area":"mechinterp"},{"id":"2609.09902","title":"Contrastive Projection: Reading Transformer Internals by Differencing Logit Lenses","created":"2026-09-09","cited_by_count":0,"area":"mechinterp"},{"id":"2609.30405","title":"Adaptive Multi-Value Control in LLMs via Causal Activation Steering","created":"2026-09-24","cited_by_count":0,"area":"mechinterp"},{"id":"2609.39263","title":"Concept Subspaces Compute Beyond the Logit Lens: A Weights-Only Test for Locating Representations Upstream of Readout","created":"2026-09-30","cited_by_count":0,"area":"mechinterp"},{"id":"2610.14958","title":"What Survives When Activation Steering Is Used as Memory","created":"2026-10-10","cited_by_count":0,"area":"mechinterp"}],"next_cursor":null}