{"total":118,"rows":[{"id":"2404.14082","title":"Mechanistic Interpretability for AI Safety -- A Review","created":"2024-04-22","cited_by_count":422,"area":"mechinterp"},{"id":"2410.13835","title":"Active-Dormant Attention Heads: Mechanistically Demystifying Extreme-Token Phenomena in LLMs","created":"2024-10-17","cited_by_count":56,"area":null},{"id":"2501.16265","title":"Training Dynamics of In-Context Learning in Linear Attention","created":"2025-01-27","cited_by_count":46,"area":null},{"id":"2408.10147","title":"In-Context Learning with Representations: Contextual Generalization of Trained Transformers","created":"2024-08-19","cited_by_count":38,"area":null},{"id":"2504.10957","title":"When is Task Vector Provably Effective for Model Editing? A Generalization Analysis of Nonlinear Transformers","created":"2025-04-15","cited_by_count":38,"area":"safety"},{"id":"2405.05219","title":"Conv-Basis: A New Paradigm for Efficient Attention Inference and Gradient Computation in Transformers","created":"2024-05-08","cited_by_count":32,"area":null},{"id":"2502.21212","title":"Transformers Learn to Implement Multi-step Gradient Descent with Chain of Thought","created":"2025-02-28","cited_by_count":32,"area":"lr"},{"id":"2406.02128","title":"Iteration Head: A Mechanistic Study of Chain-of-Thought","created":"2024-06-04","cited_by_count":31,"area":"lr"},{"id":"2411.02544","title":"Pretrained transformer efficiently learns low-dimensional target functions in-context","created":"2024-11-04","cited_by_count":31,"area":null},{"id":"2410.05459","title":"From Sparse Dependence to Sparse Attention: Unveiling How Chain-of-Thought Enhances Transformer Sample Efficiency","created":"2024-10-07","cited_by_count":29,"area":"lr"},{"id":"2604.21691","title":"There Will Be a Scientific Theory of Deep Learning","created":"2026-04-23","cited_by_count":29,"area":null},{"id":"2408.09503","title":"Out-of-distribution generalization via composition: a lens through induction heads in Transformers","created":"2024-08-18","cited_by_count":28,"area":null},{"id":"2405.04669","title":"Towards a Theoretical Understanding of the 'Reversal Curse' via Training Dynamics","created":"2024-05-07","cited_by_count":27,"area":"lr"},{"id":"2407.17686","title":"Transformers on Markov Data: Constant Depth Suffices","created":"2024-07-25","cited_by_count":27,"area":null},{"id":"2502.01694","title":"Metastable Dynamics of Chain-of-Thought Reasoning: Provable Benefits of Search, RL and Distillation","created":"2025-02-02","cited_by_count":25,"area":"lr"},{"id":"2505.17863","title":"The emergence of sparse attention: impact of data distribution and benefits of repetition","created":"2025-05-23","cited_by_count":25,"area":null},{"id":"2406.06893","title":"Transformers Provably Learn Sparse Token Selection While Fully-Connected Nets Cannot","created":"2024-06-11","cited_by_count":22,"area":null},{"id":"2502.13913","title":"How Do LLMs Perform Two-Hop Reasoning in Context?","created":"2025-02-19","cited_by_count":20,"area":"lr"},{"id":"2403.08699","title":"Implicit Regularization of Gradient Flow on One-Layer Softmax Attention","created":"2024-03-13","cited_by_count":19,"area":null},{"id":"2410.23042","title":"Toward Understanding In-context vs. In-weight Learning","created":"2024-10-30","cited_by_count":19,"area":null},{"id":"2406.11779","title":"Compact Proofs of Model Performance via Mechanistic Interpretability","created":"2024-06-17","cited_by_count":18,"area":"mechinterp"},{"id":"2410.01537","title":"Attention layers provably solve single-location regression","created":"2024-10-02","cited_by_count":17,"area":null},{"id":"2508.12837","title":"Learning In-context n-grams with Transformers: Sub-n-grams Are Near-stationary Points","created":"2025-08-18","cited_by_count":16,"area":null},{"id":"2509.23365","title":"Emergence of Superposition: Unveiling the Training Dynamics of Chain of Continuous Thought","created":"2025-09-27","cited_by_count":16,"area":"lr"},{"id":"2405.15302","title":"Understanding the Language Model to Solve the Symbolic Multi-Step Reasoning Problem from the Perspective of Buffer Mechanism","created":"2024-05-24","cited_by_count":15,"area":"lr"},{"id":"2406.18400","title":"Do LLMs dream of elephants (when told not to)? Latent concept association and associative memory in transformers","created":"2024-06-26","cited_by_count":15,"area":null},{"id":"2410.11778","title":"On the Training Convergence of Transformers for In-Context Classification of Gaussian Mixtures","created":"2024-10-15","cited_by_count":15,"area":null},{"id":"2504.12691","title":"Why and How LLMs Hallucinate: Connecting the Dots with Subsequence Associations","created":"2025-04-17","cited_by_count":15,"area":null},{"id":"2408.04532","title":"How Transformers Utilize Multi-Head Attention in In-Context Learning? A Case Study on Sparse Linear Regression","created":"2024-08-08","cited_by_count":14,"area":null},{"id":"2410.02167","title":"Training Nonlinear Transformers for Chain-of-Thought Inference: A Theoretical Generalization Analysis","created":"2024-10-03","cited_by_count":14,"area":"lr"},{"id":"2511.07378","title":"Transformers Provably Learn Chain-of-Thought Reasoning with Length Generalization","created":"2025-11-10","cited_by_count":14,"area":"lr"},{"id":"2409.17335","title":"Non-asymptotic Convergence of Training Transformers for Next-token Prediction","created":"2024-09-25","cited_by_count":13,"area":null},{"id":"2502.10178","title":"From Markov to Laplace: How Mamba In-Context Learns Markov Chains","created":"2025-02-14","cited_by_count":11,"area":"ssm"},{"id":"2505.00926","title":"How Transformers Learn Regular Language Recognition: A Theoretical Study on Training Dynamics and Implicit Bias","created":"2025-05-02","cited_by_count":11,"area":null},{"id":"2506.10887","title":"Generalization or Hallucination? Understanding Out-of-Context Reasoning in Transformers","created":"2025-06-12","cited_by_count":10,"area":"lr"},{"id":"2510.06954","title":"From Condensation to Rank Collapse: A Two-Stage Analysis of Transformer Training Dynamics","created":"2025-10-08","cited_by_count":10,"area":null},{"id":"2405.15926","title":"Dissecting the Interplay of Attention Paths in a Statistical Mechanics Theory of Transformers","created":"2024-05-24","cited_by_count":9,"area":null},{"id":"2410.09605","title":"Training Dynamics of Transformers to Recognize Word Co-occurrence via Gradient Flow Analysis","created":"2024-10-12","cited_by_count":9,"area":null},{"id":"2410.01265","title":"Transformers Handle Endogeneity in In-Context Linear Regression","created":"2024-10-02","cited_by_count":8,"area":null},{"id":"2410.11474","title":"How Transformers Get Rich: Approximation and Dynamics Analysis","created":"2024-10-15","cited_by_count":8,"area":null},{"id":"2505.05145","title":"Understanding In-context Learning of Addition via Activation Subspaces","created":"2025-05-08","cited_by_count":8,"area":null},{"id":"2509.08184","title":"Selective Induction Heads: How Transformers Select Causal Structures In Context","created":"2025-09-09","cited_by_count":8,"area":null},{"id":"2510.27015","title":"Quantitative Bounds for Length Generalization in Transformers","created":"2025-10-30","cited_by_count":8,"area":null},{"id":"2501.18187","title":"On the Role of Transformer Feed-Forward Layers in Nonlinear In-Context Learning","created":"2025-01-30","cited_by_count":7,"area":null},{"id":"2407.05591","title":"On the Power of Convolution Augmented Transformer","created":"2024-07-08","cited_by_count":6,"area":null},{"id":"2409.12293","title":"In-Context Learning of Linear Systems: Generalization Theory and Applications to Operator Learning","created":"2024-09-18","cited_by_count":6,"area":null},{"id":"2410.13981","title":"On the Learn-to-Optimize Capabilities of Transformers in In-Context Sparse Recovery","created":"2024-10-17","cited_by_count":6,"area":null},{"id":"2506.05200","title":"Transformers Meet In-Context Learning: A Universal Approximation Theory","created":"2025-06-05","cited_by_count":6,"area":null},{"id":"2512.18634","title":"From Shortcut to Induction Head: How Data Diversity Shapes Algorithm Selection in Transformers","created":"2025-12-21","cited_by_count":6,"area":null},{"id":"2601.15158","title":"Outcome-Based RL Provably Leads Transformers to Reason, but Only With the Right Data","created":"2026-01-21","cited_by_count":6,"area":"lr"},{"id":"2405.16845","title":"On Mesa-Optimization in Autoregressively Trained Transformers: Emergence and Capability","created":"2024-05-27","cited_by_count":5,"area":null},{"id":"2406.10557","title":"Explain the Black Box for the Sake of Science: the Scientific Method in the Era of Generative Artificial Intelligence","created":"2024-06-15","cited_by_count":5,"area":null},{"id":"2511.01033","title":"On the Emergence of Induction Heads for In-Context Learning","created":"2025-11-02","cited_by_count":5,"area":null},{"id":"2502.08136","title":"In-Context Learning of Linear Dynamical Systems with Transformers: Approximation Bounds and Depth-Separation","created":"2025-02-12","cited_by_count":4,"area":null},{"id":"2502.15132","title":"CoT-ICL Lab: A Synthetic Framework for Studying Chain-of-Thought Learning from In-Context Demonstrations","created":"2025-02-21","cited_by_count":4,"area":"lr"},{"id":"2507.20443","title":"Provable In-Context Learning of Nonlinear Regression with Transformers","created":"2025-07-28","cited_by_count":4,"area":null},{"id":"2509.23213","title":"One-Shot Multi-Label Causal Discovery in High-Dimensional Event Sequences","created":"2025-09-27","cited_by_count":4,"area":null},{"id":"2511.07368","title":"Distributional Biases in Post-Training: A Markovian Analysis of Reasoning Trajectories","created":"2025-11-10","cited_by_count":4,"area":"lr"},{"id":"2511.07372","title":"Provable Benefit of Curriculum in Transformer Tree-Reasoning Post-Training","created":"2025-11-10","cited_by_count":4,"area":"lr"},{"id":"2405.04161","title":"Decoding complexity: how machine learning is redefining scientific discovery","created":"2024-05-07","cited_by_count":3,"area":null},{"id":"2406.02663","title":"Demystifying Spectral Bias on Real-World Data","created":"2024-06-04","cited_by_count":3,"area":null},{"id":"2410.05493","title":"An Information-Theoretic Approach to Understanding Transformers' In-Context Learning of Variable-Order Markov Chains","created":"2024-10-07","cited_by_count":3,"area":null},{"id":"2501.15857","title":"Are Transformers Able to Reason by Connecting Separated Knowledge in Training Data?","created":"2025-01-27","cited_by_count":3,"area":null},{"id":"2502.17024","title":"Towards Auto-Regressive Next-Token Prediction: In-Context Learning Emerges from Generalization","created":"2025-02-24","cited_by_count":3,"area":null},{"id":"2506.01582","title":"Bayes optimal learning of attention-indexed models","created":"2025-06-02","cited_by_count":3,"area":null},{"id":"2509.25741","title":"Test time training enhances in-context learning of nonlinear functions","created":"2025-09-30","cited_by_count":3,"area":null},{"id":"2601.19208","title":"How Do Transformers Learn to Associate Tokens: Gradient Leading Terms Bring Mechanistic Interpretability","created":"2026-01-27","cited_by_count":3,"area":"mechinterp"},{"id":"2603.20969","title":"Understanding Contextual Recall in Transformers: How Finetuning Enables In-Context Reasoning over Pretraining Knowledge","created":"2026-03-21","cited_by_count":3,"area":"lr"},{"id":"2604.14602","title":"CausalDetox: Causal Head Selection and Intervention for Language Model Detoxification","created":"2026-04-16","cited_by_count":3,"area":null},{"id":"2605.05755","title":"Transformers Provably Implement In-Context Reinforcement Learning with Policy Improvement","created":"2026-05-07","cited_by_count":3,"area":"deeprl"},{"id":"2507.05644","title":"The Features at Convergence Theorem: a first-principles alternative to the Neural Feature Ansatz for how networks learn representations","created":"2025-07-08","cited_by_count":2,"area":null},{"id":"2509.24125","title":"The Impossibility of Inverse Permutation Learning in Transformer Models","created":"2025-09-28","cited_by_count":2,"area":null},{"id":"2510.25542","title":"Transformers Provably Learn Directed Acyclic Graphs via Kernel-Guided Mutual Information","created":"2025-10-29","cited_by_count":2,"area":null},{"id":"2511.23239","title":"Towards Understanding Transformers in Learning Random Walks","created":"2025-11-28","cited_by_count":2,"area":null},{"id":"2512.18452","title":"Secret mixtures of experts inside your LLM","created":"2025-12-20","cited_by_count":2,"area":"moe"},{"id":"2601.10058","title":"Unlabeled Data Can Provably Enhance In-Context Learning of Transformers","created":"2026-01-15","cited_by_count":2,"area":null},{"id":"2602.02470","title":"Breaking the Reversal Curse in Autoregressive Language Models via Identity Bridge","created":"2026-02-02","cited_by_count":2,"area":"lr"},{"id":"2602.19143","title":"Incremental Learning of Sparse Attention Patterns in Transformers","created":"2026-02-22","cited_by_count":2,"area":null},{"id":"2605.28600","title":"Transformers Provably Learn to Internalize Chain-of-Thought","created":"2026-05-27","cited_by_count":2,"area":"lr"},{"id":"2606.22938","title":"Provable Benefits of RLVR over SFT for Reasoning Models: Learning to Backtrack Efficiently","created":"2026-06-22","cited_by_count":2,"area":"lr"},{"id":"2607.11875","title":"Invariant Learning Dynamics of Transformers in Inductive Reasoning Tasks","created":"2026-07-13","cited_by_count":2,"area":"lr"},{"id":"2407.10949","title":"Representing Rule-based Chatbots with Transformers","created":"2024-07-15","cited_by_count":1,"area":null},{"id":"2407.15425","title":"Empirical Capacity Model for Self-Attention Neural Networks","created":"2024-07-22","cited_by_count":1,"area":null},{"id":"2410.11444","title":"A Theoretical Survey on Foundation Models","created":"2024-10-15","cited_by_count":1,"area":null},{"id":"2505.17282","title":"Attention with Trained Embeddings Provably Selects Important Tokens","created":"2025-05-22","cited_by_count":1,"area":null},{"id":"2506.01919","title":"Transformers as Multi-task Learners: Decoupling Features in Hidden Markov Models","created":"2025-06-02","cited_by_count":1,"area":null},{"id":"2507.13920","title":"Causal Process Models: Reframing Dynamic Causal Graph Discovery as a Reinforcement Learning Problem","created":"2025-07-18","cited_by_count":1,"area":"deeprl"},{"id":"2508.02126","title":"Understanding Learning Dynamics Through Structured Representations","created":"2025-08-04","cited_by_count":1,"area":null},{"id":"2511.19328","title":"Understanding the Staged Dynamics of Transformers in Learning Latent Structure","created":"2025-11-24","cited_by_count":1,"area":null},{"id":"2601.16514","title":"Finite-Time Analysis of Gradient Descent for Shallow Transformers","created":"2026-01-23","cited_by_count":1,"area":null},{"id":"2601.22450","title":"Tuning the Implicit Regularizer of Masked Diffusion Language Models: Enhancing Generalization via Insights from $k$-Parity","created":"2026-01-30","cited_by_count":1,"area":"diffusion"},{"id":"2604.06440","title":"Visual prompting reimagined: The power of the Activation Prompts","created":"2026-04-07","cited_by_count":1,"area":null},{"id":"2604.10848","title":"Transformers Learn Latent Mixture Models In-Context via Mirror Descent","created":"2026-04-12","cited_by_count":1,"area":null},{"id":"2605.04330","title":"The Scaling Properties of Implicit Deductive Reasoning in Transformers","created":"2026-05-05","cited_by_count":1,"area":"lr"},{"id":"2605.27774","title":"Fine-Tuning Dynamics of In-Context Factual Recall in Transformers","created":"2026-05-26","cited_by_count":1,"area":null},{"id":"2607.02800","title":"Induction Heads Interpolate N-Grams","created":"2026-07-02","cited_by_count":1,"area":null},{"id":"2607.03660","title":"Sequential Correlations Change In-Context Learning: Effective Context Length and Architectural Mismatch","created":"2026-07-04","cited_by_count":1,"area":null},{"id":"2411.13264","title":"Transformers with Sparse Attention for Granger Causality","created":"2024-11-20","cited_by_count":0,"area":null},{"id":"2412.11459","title":"Rethinking Associative Memory Mechanism in Induction Head","created":"2024-12-16","cited_by_count":0,"area":"lr"},{"id":"2509.23178","title":"Limit Analysis for Symbolic Multi-step Reasoning Tasks with Information Propagation Rules Based on Transformers","created":"2025-09-27","cited_by_count":0,"area":null}],"next_cursor":"100"}