{"total":115,"rows":[{"id":"2503.05613","title":"A Survey on Sparse Autoencoders: Interpreting the Internal Mechanisms of Large Language Models","created":"2025-03-07","cited_by_count":76,"area":"mechinterp"},{"id":"2503.01822","title":"Projecting Assumptions: The Duality Between Sparse Autoencoders and Concept Geometry","created":"2025-03-03","cited_by_count":61,"area":"mechinterp"},{"id":"2506.10805","title":"Detecting High-Stakes Interactions with Activation Probes","created":"2025-06-12","cited_by_count":53,"area":null},{"id":"2512.15674","title":"Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers","created":"2025-12-17","cited_by_count":52,"area":null},{"id":"2601.11516","title":"Building Production-Ready Probes For Gemini","created":"2026-01-16","cited_by_count":47,"area":"safety"},{"id":"2601.04603","title":"Constitutional Classifiers++: Efficient Production-Grade Defenses against Universal Jailbreaks","created":"2026-01-08","cited_by_count":41,"area":"safety"},{"id":"2507.16795","title":"Steering Out-of-Distribution Generalization with Concept Ablation Fine-Tuning","created":"2025-07-22","cited_by_count":39,"area":null},{"id":"2604.28119","title":"Do Sparse Autoencoders Capture Concept Manifolds?","created":"2026-04-30","cited_by_count":31,"area":"mechinterp"},{"id":"2505.16188","title":"SAE-SSV: Supervised Steering in Sparse Representation Spaces for Reliable Control of Language Models","created":"2025-05-22","cited_by_count":29,"area":"mechinterp"},{"id":"2510.13290","title":"To Steer or Not to Steer? Mechanistic Error Reduction with Abstention for Language Models","created":"2025-10-15","cited_by_count":29,"area":null},{"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":"2601.22594","title":"Language Model Circuits Are Sparse in the Neuron Basis","created":"2026-01-30","cited_by_count":25,"area":"mechinterp"},{"id":"2505.20322","title":"Beyond Prompt Engineering: Robust Behavior Control in LLMs via Steering Target Atoms","created":"2025-05-23","cited_by_count":24,"area":null},{"id":"2506.14261","title":"RL-Obfuscation: Can Language Models Learn to Evade Latent-Space Monitors?","created":"2025-06-17","cited_by_count":24,"area":null},{"id":"2506.06609","title":"Transferring Linear Features Across Language Models With Model Stitching","created":"2025-06-07","cited_by_count":21,"area":"mechinterp"},{"id":"2506.23845","title":"Position: Use Sparse Autoencoders to Discover Unknowns","created":"2025-06-30","cited_by_count":19,"area":"mechinterp"},{"id":"2511.05541","title":"Temporal Sparse Autoencoders: Leveraging the Sequential Nature of Language for Interpretability","created":"2025-10-30","cited_by_count":18,"area":"mechinterp"},{"id":"2505.11756","title":"Feature Hedging: Correlated Features Break Narrow Sparse Autoencoders","created":"2025-05-16","cited_by_count":16,"area":"mechinterp"},{"id":"2504.20271","title":"Investigating task-specific prompts and sparse autoencoders for activation monitoring","created":"2025-04-28","cited_by_count":15,"area":"mechinterp"},{"id":"2508.16560","title":"Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders","created":"2025-08-22","cited_by_count":14,"area":"mechinterp"},{"id":"2602.14111","title":"Sanity Checks for Sparse Autoencoders: Do SAEs Beat Random Baselines?","created":"2026-02-15","cited_by_count":14,"area":"mechinterp"},{"id":"2503.08980","title":"I Predict Therefore I Am: Is Next Token Prediction Enough to Learn Human-Interpretable Concepts from Data?","created":"2025-03-12","cited_by_count":13,"area":null},{"id":"2601.21571","title":"Shaping capabilities with token-level data filtering","created":"2026-01-29","cited_by_count":13,"area":null},{"id":"2512.18792","title":"The Dead Salmons of AI Interpretability","created":"2025-12-21","cited_by_count":12,"area":null},{"id":"2510.06036","title":"Refusal Falls off a Cliff: How Safety Alignment Fails in Reasoning?","created":"2025-10-07","cited_by_count":11,"area":"lr"},{"id":"2602.14161","title":"When Benchmarks Lie: Evaluating Malicious Prompt Classifiers Under True Distribution Shift","created":"2026-02-15","cited_by_count":11,"area":"agents"},{"id":"2602.14687","title":"SynthSAEBench: Evaluating Sparse Autoencoders on Scalable Realistic Synthetic Data","created":"2026-02-16","cited_by_count":11,"area":"synth"},{"id":"2508.11277","title":"Probing the Representational Power of Sparse Autoencoders in Vision Models","created":"2025-08-15","cited_by_count":10,"area":"diffusion"},{"id":"2510.00404","title":"AbsTopK: Rethinking Sparse Autoencoders For Bidirectional Features","created":"2025-10-01","cited_by_count":9,"area":"lr"},{"id":"2512.11949","title":"Neural Chameleons: Language Models Can Learn to Hide Their Thoughts from Unseen Activation Monitors","created":"2025-12-12","cited_by_count":9,"area":"lr"},{"id":"2502.19964","title":"Do Sparse Autoencoders Generalize? A Case Study of Answerability","created":"2025-02-27","cited_by_count":8,"area":"mechinterp"},{"id":"2507.06445","title":"Can Interpretation Predict Behavior on Unseen Data?","created":"2025-07-08","cited_by_count":8,"area":null},{"id":"2510.01025","title":"Hypothesis-Driven Feature Manifold Analysis in LLMs via Supervised Multi-Dimensional Scaling","created":"2025-10-01","cited_by_count":8,"area":null},{"id":"2601.22588","title":"Rethinking LLM-as-a-Judge: Representation-as-a-Judge with Small Language Models via Semantic Capacity Asymmetry","created":"2026-01-30","cited_by_count":8,"area":"lr"},{"id":"2602.05027","title":"AudioSAE: Towards Understanding of Audio-Processing Models with Sparse AutoEncoders","created":"2026-02-04","cited_by_count":8,"area":"mechinterp"},{"id":"2602.06964","title":"Learning a Generative Meta-Model of LLM Activations","created":"2026-02-06","cited_by_count":8,"area":"mechinterp"},{"id":"2602.16698","title":"Causality is Key for Interpretability Claims to Generalise","created":"2026-02-18","cited_by_count":8,"area":null},{"id":"2604.02178","title":"The Expert Strikes Back: Interpreting Mixture-of-Experts Language Models at Expert Level","created":"2026-04-02","cited_by_count":7,"area":"moe"},{"id":"2506.13060","title":"Rethinking Explainability in the Era of Multimodal AI","created":"2025-06-16","cited_by_count":6,"area":null},{"id":"2512.10805","title":"Interpretable and Steerable Concept Bottleneck Sparse Autoencoders","created":"2025-12-11","cited_by_count":6,"area":"mechinterp"},{"id":"2602.12418","title":"Sparse Autoencoders are Capable LLM Jailbreak Mitigators","created":"2026-02-12","cited_by_count":6,"area":"safety"},{"id":"2603.13277","title":"Learning Retrieval Models with Sparse Autoencoders","created":"2026-02-27","cited_by_count":6,"area":"mechinterp"},{"id":"2603.28744","title":"Stop Probing, Start Coding: Why Linear Probes and Sparse Autoencoders Fail at Compositional Generalisation","created":"2026-03-30","cited_by_count":6,"area":"mechinterp"},{"id":"2505.21364","title":"Towards Interpretability Without Sacrifice: Faithful Dense Layer Decomposition with Mixture of Decoders","created":"2025-05-27","cited_by_count":5,"area":null},{"id":"2603.04198","title":"Stable and Steerable Sparse Autoencoders with Weight Regularization","created":"2026-03-04","cited_by_count":5,"area":"mechinterp"},{"id":"2605.18549","title":"Monitoring the Internal Monologue: Probe Trajectories Reveal Reasoning Dynamics","created":"2026-05-18","cited_by_count":5,"area":"lr"},{"id":"2508.15094","title":"Evaluating Sparse Autoencoders for Monosemantic Representation","created":"2025-08-20","cited_by_count":4,"area":"mechinterp"},{"id":"2602.11910","title":"TADA! Tuning Audio Diffusion Models through Activation Steering","created":"2026-02-12","cited_by_count":4,"area":"diffusion"},{"id":"2604.13386","title":"Linear Probe Accuracy Scales with Model Size and Benefits from Multi-Layer Ensembling","created":"2026-04-15","cited_by_count":4,"area":"alignment"},{"id":"2605.18229","title":"Are Sparse Autoencoder Benchmarks Reliable?","created":"2026-05-18","cited_by_count":4,"area":"mechinterp"},{"id":"2606.11722","title":"ICA Lens: Interpreting Language Models Without Training Another Dictionary","created":"2026-06-10","cited_by_count":4,"area":null},{"id":"2505.17769","title":"Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models","created":"2025-05-23","cited_by_count":3,"area":"mechinterp"},{"id":"2505.17998","title":"TRACE for Tracking the Emergence of Semantic Representations in Transformers","created":"2025-05-23","cited_by_count":3,"area":null},{"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":"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":"2604.14865","title":"Segment-Level Coherence for Robust Harmful Intent Probing in LLMs","created":"2026-04-16","cited_by_count":3,"area":"safety"},{"id":"2505.10375","title":"Are Sparse Autoencoders Useful for Java Function Bug Detection?","created":"2025-05-15","cited_by_count":2,"area":"mechinterp"},{"id":"2506.20040","title":"Cross-Layer Discrete Concept Discovery for Interpreting Language Models","created":"2025-06-24","cited_by_count":2,"area":"mechinterp"},{"id":"2509.18792","title":"Beyond the Leaderboard: Understanding Performance Disparities in Large Language Models via Model Diffing","created":"2025-09-23","cited_by_count":2,"area":null},{"id":"2511.20910","title":"Emergence and Localisation of Semantic Role Circuits in LLMs","created":"2025-11-25","cited_by_count":2,"area":"mechinterp"},{"id":"2601.20075","title":"Sparse CLIP: Co-Optimizing Interpretability and Performance in Contrastive Learning","created":"2026-01-27","cited_by_count":2,"area":"mllm"},{"id":"2605.12874","title":"Descriptive Collision in Sparse Autoencoder Auto-Interpretability: When One Explanation Describes Many Features","created":"2026-05-13","cited_by_count":2,"area":"mechinterp"},{"id":"2605.15164","title":"Position: Behavioural Assurance Cannot Verify the Safety Claims Governance Now Demands","created":"2026-05-14","cited_by_count":2,"area":null},{"id":"2607.08046","title":"What LLM Forecasters Know but Don't Say: Probing Internal Representations for Calibration and Faithfulness","created":"2026-07-09","cited_by_count":2,"area":"lr"},{"id":"2607.26865","title":"Think Short, Defer Smart, Act, and Repeat: Calibrated Reasoning and Uncertainty-Aware Deferral for Edge LLM Agents","created":"2026-07-29","cited_by_count":2,"area":"lr"},{"id":"2609.18080","title":"Decodability is Not Causality: Dissociating Probe Readouts from Behavioral Drivers via SAE Decomposition","created":"2026-09-16","cited_by_count":2,"area":"mechinterp"},{"id":"2507.06458","title":"Automated Neuron Labelling Enables Generative Steering and Interpretability in Protein Language Models","created":"2025-07-08","cited_by_count":1,"area":"mechinterp"},{"id":"2509.10809","title":"Rethinking Sparse Autoencoders: Select-and-Project for Fairness and Control from Encoder Features Alone","created":"2025-09-13","cited_by_count":1,"area":"lr"},{"id":"2601.14660","title":"NeuroFilter: Activation-Based Guardrails for Privacy-Conscious LLM Agents","created":"2026-01-21","cited_by_count":1,"area":"agents"},{"id":"2602.08713","title":"Towards Understanding Multimodal Fine-Tuning: Spatial Features","created":"2026-02-06","cited_by_count":1,"area":"mllm"},{"id":"2604.06495","title":"Improving Robustness In Sparse Autoencoders via Masked Regularization","created":"2026-04-07","cited_by_count":1,"area":"mechinterp"},{"id":"2604.11662","title":"A Robust Evaluation of Probe Robustness: Lessons for Reliable OOD Uncertainty Quantification","created":"2026-04-13","cited_by_count":1,"area":null},{"id":"2605.08936","title":"Self-ReSET: Learning to Self-Recover from Unsafe Reasoning Trajectories","created":"2026-05-09","cited_by_count":1,"area":"lr"},{"id":"2605.09224","title":"SMIXAE: Towards Unsupervised Manifold Discovery in Language Models","created":"2026-05-09","cited_by_count":1,"area":"mechinterp"},{"id":"2606.15092","title":"High-Dimensional Random Projection for Activation Steering in Language Models","created":"2026-06-13","cited_by_count":1,"area":"mechinterp"},{"id":"2607.16448","title":"Retrieval is Enough: Training-Free Interpretability with a Tool-Using Agent","created":"2026-07-17","cited_by_count":1,"area":"agents"},{"id":"2609.03026","title":"ObserverBench: Testing Mechanistic Estimates for Intervention and Control","created":"2026-09-02","cited_by_count":1,"area":"mechinterp"},{"id":"2609.06862","title":"Feature Superposition in Neural Networks: From Theory to Practice","created":"2026-09-06","cited_by_count":1,"area":"mechinterp"},{"id":"2609.19150","title":"Sampling Reveals Style: Unsupervised, Training-Free Discovery of Prompt-Conditional Stylistic Axes in LLM Activations","created":"2026-07-20","cited_by_count":1,"area":null},{"id":"2609.36490","title":"LLMs Learn to Evade Latent Monitors from Prior Feedback Alone","created":"2026-09-29","cited_by_count":1,"area":"agents"},{"id":"2507.10106","title":"BlueGlass: A Framework for Composite AI Safety","created":"2025-07-14","cited_by_count":0,"area":null},{"id":"2509.22015","title":"Concept-SAE: A Controllable and Invertible Concept Interface for Sparse Autoencoders","created":"2025-09-26","cited_by_count":0,"area":"mechinterp"},{"id":"2509.26560","title":"Estimating Dimensionality of Neural Representations from Finite Samples","created":"2025-09-30","cited_by_count":0,"area":null},{"id":"2510.01288","title":"Microsaccade-Inspired Probing: Positional Encoding Perturbations Reveal LLM Misbehaviours","created":"2025-10-01","cited_by_count":0,"area":null},{"id":"2511.07572","title":"SCALAR: Benchmarking SAE Interaction Sparsity in Toy LLMs","created":"2025-11-10","cited_by_count":0,"area":"mechinterp"},{"id":"2512.21670","title":"The Deepfake Detective: Interpreting Neural Forensics Through Sparse Features and Manifolds","created":"2025-12-25","cited_by_count":0,"area":"mechinterp"},{"id":"2601.20420","title":"Concept Component Analysis: A Principled Approach for Concept Extraction in LLMs","created":"2026-01-28","cited_by_count":0,"area":"mechinterp"},{"id":"2602.15730","title":"Causal Effect Estimation with Latent Textual Treatments","created":"2026-02-17","cited_by_count":0,"area":null},{"id":"2602.23826","title":"GLUScope: A Tool for Analyzing GLU Neurons in Transformer Language Models","created":"2026-02-27","cited_by_count":0,"area":null},{"id":"2604.01619","title":"Automatic Image-Level Morphological Trait Annotation for Organismal Images","created":"2026-04-02","cited_by_count":0,"area":null},{"id":"2604.08846","title":"Dictionary-Aligned Concept Control for Safeguarding Multimodal LLMs","created":"2026-04-10","cited_by_count":0,"area":"mllm"},{"id":"2605.03160","title":"Steering grids for sparse-autoencoder features: when a top-context label names an activation regime rather than a causal axis","created":"2026-05-04","cited_by_count":0,"area":"mechinterp"},{"id":"2605.28649","title":"Interpretability-Guided Layer Selection over Subspace Projection: SAEs as Stethoscopes, Not Scalpels, for Raw Task Vector Model Editing","created":"2026-05-27","cited_by_count":0,"area":"mechinterp"},{"id":"2606.14040","title":"Decompose Sparsely Where You Should, Absorb Densely Where You Should No","created":"2026-06-12","cited_by_count":0,"area":"mechinterp"},{"id":"2606.19750","title":"Manifold Bandits: Bayesian Curriculum Learning over the Latent Geometry of Large Language Models","created":"2026-06-18","cited_by_count":0,"area":"lr"},{"id":"2607.10578","title":"Laguerre Geometry for Interpreting Large Language Models","created":"2026-07-12","cited_by_count":0,"area":null},{"id":"2607.14070","title":"Screening of Biosecurity Features in Metagenomic Data with Evo 2 Probes","created":"2026-07-15","cited_by_count":0,"area":null},{"id":"2607.19364","title":"SAE-StatSteer: Statistical Consensus Feature Selection for Optimization-Free Activation Steering of Large Language Models","created":"2026-06-05","cited_by_count":0,"area":"mechinterp"},{"id":"2607.19386","title":"Building Fast, Evaluating Slow: Pipeline Choices Dominate Autointerpretability Score Variance","created":"2026-07-01","cited_by_count":0,"area":"mechinterp"},{"id":"2608.03507","title":"ChronoLens: Measuring Language Change Across Time, Languages, and Linguistic Levels","created":"2026-08-04","cited_by_count":0,"area":null}],"next_cursor":"100"}