{"total":111,"rows":[{"id":"2404.14082","title":"Mechanistic Interpretability for AI Safety -- A Review","created":"2024-04-22","cited_by_count":422,"area":"mechinterp"},{"id":"2309.16042","title":"Towards Best Practices of Activation Patching in Language Models: Metrics and Methods","created":"2023-09-27","cited_by_count":359,"area":"mechinterp"},{"id":"2502.17424","title":"Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs","created":"2025-02-24","cited_by_count":172,"area":"safety"},{"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":"2310.06110","title":"Grokking as the Transition from Lazy to Rich Training Dynamics","created":"2023-10-09","cited_by_count":108,"area":null},{"id":"2410.21272","title":"Arithmetic Without Algorithms: Language Models Solve Math With a Bag of Heuristics","created":"2024-10-28","cited_by_count":97,"area":"lr"},{"id":"2405.00208","title":"A Primer on the Inner Workings of Transformer-based Language Models","created":"2024-04-30","cited_by_count":85,"area":null},{"id":"2311.18817","title":"Dichotomy of Early and Late Phase Implicit Biases Can Provably Induce Grokking","created":"2023-11-30","cited_by_count":80,"area":null},{"id":"2405.15071","title":"Grokked Transformers are Implicit Reasoners: A Mechanistic Journey to the Edge of Generalization","created":"2024-05-23","cited_by_count":79,"area":"lr"},{"id":"2503.01067","title":"All Roads Lead to Likelihood: The Value of Reinforcement Learning in Fine-Tuning","created":"2025-03-03","cited_by_count":59,"area":"alignment"},{"id":"2402.15555","title":"Deep Networks Always Grok and Here is Why","created":"2024-02-23","cited_by_count":51,"area":null},{"id":"2310.03789","title":"Grokking as a First Order Phase Transition in Two Layer Networks","created":"2023-10-05","cited_by_count":48,"area":null},{"id":"2501.04697","title":"Grokking at the Edge of Numerical Stability","created":"2025-01-08","cited_by_count":46,"area":null},{"id":"2407.12332","title":"Why Do You Grok? A Theoretical Analysis of Grokking Modular Addition","created":"2024-07-17","cited_by_count":36,"area":null},{"id":"2310.02541","title":"Benign Overfitting and Grokking in ReLU Networks for XOR Cluster Data","created":"2023-10-04","cited_by_count":33,"area":null},{"id":"2310.03262","title":"Predicting Emergent Abilities with Infinite Resolution Evaluation","created":"2023-10-05","cited_by_count":28,"area":null},{"id":"2407.20199","title":"Emergence in non-neural models: grokking modular arithmetic via average gradient outer product","created":"2024-07-29","cited_by_count":28,"area":null},{"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":"2312.06581","title":"Grokking Group Multiplication with Cosets","created":"2023-12-11","cited_by_count":24,"area":null},{"id":"2401.10463","title":"Critical Data Size of Language Models from a Grokking Perspective","created":"2024-01-19","cited_by_count":24,"area":null},{"id":"2404.16367","title":"Learning Syntax Without Planting Trees: Understanding Hierarchical Generalization in Transformers","created":"2024-04-25","cited_by_count":24,"area":null},{"id":"2402.15175","title":"Unified View of Grokking, Double Descent and Emergent Abilities: A Perspective from Circuits Competition","created":"2024-02-23","cited_by_count":23,"area":null},{"id":"2602.07852","title":"Emergent Misalignment is Easy, Narrow Misalignment is Hard","created":"2026-02-08","cited_by_count":23,"area":null},{"id":"2505.11581","title":"Questioning Representational Optimism in Deep Learning: The Fractured Entangled Representation Hypothesis","created":"2025-05-16","cited_by_count":22,"area":null},{"id":"2404.09562","title":"{\\sigma}-GPTs: A New Approach to Autoregressive Models","created":"2024-04-15","cited_by_count":20,"area":null},{"id":"2402.01909","title":"On Catastrophic Inheritance of Large Foundation Models","created":"2024-02-02","cited_by_count":15,"area":null},{"id":"2509.21519","title":"Provable Scaling Laws of Feature Emergence from Learning Dynamics of Grokking","created":"2025-09-25","cited_by_count":15,"area":null},{"id":"2310.05918","title":"Grokking as Compression: A Nonlinear Complexity Perspective","created":"2023-10-09","cited_by_count":14,"area":null},{"id":"2402.16726","title":"Towards Empirical Interpretation of Internal Circuits and Properties in Grokked Transformers on Modular Polynomials","created":"2024-02-26","cited_by_count":14,"area":null},{"id":"2504.13292","title":"Let Me Grok for You: Accelerating Grokking via Embedding Transfer from a Weaker Model","created":"2025-04-17","cited_by_count":13,"area":null},{"id":"2311.06597","title":"Understanding Grokking Through A Robustness Viewpoint","created":"2023-11-11","cited_by_count":12,"area":null},{"id":"2505.17923","title":"Language models can learn implicit multi-hop reasoning, but only if they have lots of training data","created":"2025-05-23","cited_by_count":11,"area":"lr"},{"id":"2602.16746","title":"Low-Dimensional and Transversely Curved Optimization Dynamics in Grokking","created":"2026-02-18","cited_by_count":11,"area":null},{"id":"2605.29548","title":"Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention","created":"2026-05-28","cited_by_count":10,"area":null},{"id":"2411.00247","title":"Deep Learning Through A Telescoping Lens: A Simple Model Provides Empirical Insights On Grokking, Gradient Boosting & Beyond","created":"2024-10-31","cited_by_count":9,"area":null},{"id":"2602.18523","title":"The Geometry of Multi-Task Grokking: Transverse Instability, Superposition, and Weight Decay Phase Structure","created":"2026-02-19","cited_by_count":9,"area":null},{"id":"2405.12755","title":"Progress Measures for Grokking on Real-world Tasks","created":"2024-05-21","cited_by_count":8,"area":null},{"id":"2505.11411","title":"Is Grokking a Computational Glass Relaxation?","created":"2025-05-16","cited_by_count":8,"area":null},{"id":"2506.04434","title":"Grokking and Generalization Collapse: Insights from \\texttt{HTSR} theory","created":"2025-06-04","cited_by_count":8,"area":null},{"id":"2507.08017","title":"Mechanistic Indicators of Understanding in Large Language Models","created":"2025-07-07","cited_by_count":8,"area":null},{"id":"2507.20057","title":"What Can Grokking Teach Us About Learning Under Nonstationarity?","created":"2025-07-26","cited_by_count":8,"area":null},{"id":"2601.19791","title":"To Grok Grokking: Provable Grokking in Ridge Regression","created":"2026-01-27","cited_by_count":8,"area":null},{"id":"2311.00863","title":"Training Dynamics of Contextual N-Grams in Language Models","created":"2023-11-01","cited_by_count":7,"area":null},{"id":"2506.13688","title":"What Happens During the Loss Plateau? Understanding Abrupt Learning in Transformers","created":"2025-06-16","cited_by_count":7,"area":null},{"id":"2603.13331","title":"The Norm-Separation Delay Law of Grokking: A First-Principles Theory of Delayed Generalization","created":"2026-03-05","cited_by_count":7,"area":null},{"id":"2310.19470","title":"Bridging Lottery Ticket and Grokking: Understanding Grokking from Inner Structure of Networks","created":"2023-10-30","cited_by_count":6,"area":null},{"id":"2410.01779","title":"Composing Global Solutions to Reasoning Tasks via Algebraic Objects in Neural Nets","created":"2024-10-02","cited_by_count":6,"area":null},{"id":"2505.00808","title":"A Mathematical Philosophy of Explanations in Mechanistic Interpretability -- The Strange Science Part I.i","created":"2025-05-01","cited_by_count":6,"area":"mechinterp"},{"id":"2505.15624","title":"Mechanistic Insights into Grokking from the Embedding Layer","created":"2025-05-21","cited_by_count":6,"area":null},{"id":"2410.11985","title":"The Fair Language Model Paradox","created":"2024-10-15","cited_by_count":4,"area":null},{"id":"2505.15811","title":"On the creation of narrow AI: hierarchy and nonlocality of neural network skills","created":"2025-05-21","cited_by_count":4,"area":null},{"id":"2509.06931","title":"Learning words in groups: fusion algebras, tensor ranks and grokking","created":"2025-09-08","cited_by_count":4,"area":null},{"id":"2509.20829","title":"Explaining Grokking and Information Bottleneck through Neural Collapse Emergence","created":"2025-09-25","cited_by_count":4,"area":null},{"id":"2605.09724","title":"Model Capacity Determines Grokking through Competing Memorisation and Generalisation Speeds","created":"2026-05-10","cited_by_count":4,"area":null},{"id":"2606.13753","title":"The Weight Norm Sets the Grokking Timescale: A Causal Delay Law","created":"2026-06-11","cited_by_count":4,"area":null},{"id":"2405.17479","title":"A rationale from frequency perspective for grokking in training neural network","created":"2024-05-24","cited_by_count":3,"area":null},{"id":"2505.20076","title":"ExPLAIND: Unifying Model, Data, and Training Attribution to Study Model Behavior","created":"2025-05-26","cited_by_count":3,"area":null},{"id":"2604.13082","title":"The Long Delay to Arithmetic Generalization: When Learned Representations Outrun Behavior","created":"2026-03-30","cited_by_count":3,"area":null},{"id":"2607.04333","title":"Structure-Specific Representational Priors Causally Control the Grokking Delay","created":"2026-07-05","cited_by_count":3,"area":null},{"id":"2505.01751","title":"A dynamic view of some anomalous phenomena in SGD","created":"2025-05-03","cited_by_count":2,"area":null},{"id":"2505.03997","title":"Quiet Feature Learning in Algorithmic Tasks","created":"2025-05-06","cited_by_count":2,"area":null},{"id":"2510.04930","title":"Egalitarian Gradient Descent: A Simple Approach to Accelerated Grokking","created":"2025-10-06","cited_by_count":2,"area":null},{"id":"2602.02859","title":"Late-Stage Generalization Collapse in Grokking: Detecting anti-grokking with Weightwatcher","created":"2026-02-02","cited_by_count":2,"area":null},{"id":"2602.06702","title":"Explaining Grokking in Transformers through the Lens of Inductive Bias","created":"2026-02-06","cited_by_count":2,"area":null},{"id":"2604.13123","title":"Spectral Entropy Collapse as a Phase Transition in Delayed Generalisation: An Interventional and Predictive Framework for Grokkin","created":"2026-04-13","cited_by_count":2,"area":null},{"id":"2605.08237","title":"Distributional Spectral Diagnostics for Localizing Grokking Transitions","created":"2026-05-07","cited_by_count":2,"area":null},{"id":"2605.20441","title":"Weight Decay Regimes in Grokking Transformers: Cheap Online Diagnostics","created":"2026-05-19","cited_by_count":2,"area":null},{"id":"2605.27078","title":"Two Speeds of Learning: A Representation-Readout Decomposition of Grokking and Double Descent","created":"2026-05-26","cited_by_count":2,"area":null},{"id":"2606.12966","title":"Learned Structure Shows Up Earlier in the Task's Own Basis: A Controlled Look at a Synchronisation Precursor in Grokking","created":"2026-06-11","cited_by_count":2,"area":null},{"id":"2608.07436","title":"Post-Grokking Collapse at the Representation-Readout Interface in Muon-Trained Transformers","created":"2026-08-07","cited_by_count":2,"area":null},{"id":"2504.12700","title":"A Two-Phase Perspective on Deep Learning Dynamics","created":"2025-04-17","cited_by_count":1,"area":null},{"id":"2505.09855","title":"An evolutionary perspective on modes of learning in Transformers","created":"2025-05-14","cited_by_count":1,"area":null},{"id":"2505.23013","title":"Scalable Complexity Control Facilitates Reasoning Ability of LLMs","created":"2025-05-29","cited_by_count":1,"area":"lr"},{"id":"2511.04760","title":"When Data Falls Short: Grokking Below the Critical Threshold","created":"2025-11-06","cited_by_count":1,"area":null},{"id":"2601.03162","title":"On the Convergence Behavior of Preconditioned Gradient Descent Toward the Rich Learning Regime","created":"2026-01-06","cited_by_count":1,"area":null},{"id":"2602.01442","title":"Hidden Heroes and Gradient Bloats: Layer-Wise Redundancy Inverts Attribution in Transformers","created":"2026-02-01","cited_by_count":1,"area":"mechinterp"},{"id":"2602.18649","title":"Global Low-Rank, Local Full-Rank: The Holographic Encoding of Learned Algorithms","created":"2026-02-20","cited_by_count":1,"area":null},{"id":"2602.22600","title":"Transformers converge to invariant algorithmic cores","created":"2026-02-26","cited_by_count":1,"area":null},{"id":"2603.15492","title":"Grokking as a Variance-Limited Phase Transition: Spectral Gating and the Epsilon-Stability Threshold","created":"2026-03-16","cited_by_count":1,"area":null},{"id":"2603.29262","title":"Grokking From Abstraction to Intelligence","created":"2026-03-31","cited_by_count":1,"area":null},{"id":"2604.00316","title":"Breaking Data Symmetry is Needed For Generalization in Feature Learning Kernels","created":"2026-03-31","cited_by_count":1,"area":null},{"id":"2604.16697","title":"Surgical Repair of Insecure Code Generation in LLMs","created":"2026-04-17","cited_by_count":1,"area":null},{"id":"2605.18845","title":"First-Passage Prediction of Grokking Delay: ACalibrated Law under AdamW with Causal Validation","created":"2026-05-13","cited_by_count":1,"area":null},{"id":"2606.05863","title":"Deciphering Two Training Clocks in Grokking via Deep Linear Network Theory with Conditional ReLU Reduction","created":"2026-06-04","cited_by_count":1,"area":null},{"id":"2606.18465","title":"What Does the Weight Norm Control in Grokking? Logit-Scale Mediation under Cross-Entropy","created":"2026-06-16","cited_by_count":1,"area":null},{"id":"2606.26050","title":"Natural Ungrokking: Asymmetric Control of Which Rules Survive Pretraining","created":"2026-06-24","cited_by_count":1,"area":null},{"id":"2607.06639","title":"At-Grok Is Not Converged:A Measurement-Validity Audit for Grokking Representation Metrics","created":"2026-07-07","cited_by_count":1,"area":null},{"id":"2411.03541","title":"Do Mice Grok? Glimpses of Hidden Progress During Overtraining in Sensory Cortex","created":"2024-11-05","cited_by_count":0,"area":null},{"id":"2503.22059","title":"Low Rank and Sparse Fourier Structure in Recurrent Networks Trained on Modular Addition","created":"2025-03-28","cited_by_count":0,"area":null},{"id":"2507.11645","title":"Tracing the Path to Grokking: Embeddings, Dropout, and Network Activation","created":"2025-07-15","cited_by_count":0,"area":null},{"id":"2510.25791","title":"The Kinetics of Reasoning: How Chain-of-Thought Shapes Learning in Transformers?","created":"2025-10-28","cited_by_count":0,"area":"lr"},{"id":"2512.03437","title":"Grokked Models are Better Unlearners","created":"2025-12-03","cited_by_count":0,"area":"safety"},{"id":"2604.17673","title":"Grokking of Diffusion Models: Case Study on Modular Addition","created":"2026-04-20","cited_by_count":0,"area":"diffusion"},{"id":"2605.14659","title":"Slower Generalization, Faster Memorization: A Sweet Spot in Algorithmic Learning","created":"2026-05-14","cited_by_count":0,"area":null},{"id":"2605.28975","title":"A Training-Time Diagnostic for Generalization via the Log-Alignment Ratio","created":"2026-05-27","cited_by_count":0,"area":null},{"id":"2606.20737","title":"Repeated Shared Access Enables Grokking, but Edit Propagation Depends on an Addressable Memory","created":"2026-06-17","cited_by_count":0,"area":null},{"id":"2606.30388","title":"A Stochastic--Geometric Theory of Scaling Laws in Grokking","created":"2026-06-29","cited_by_count":0,"area":null},{"id":"2607.05113","title":"Rating the Pitch, Not the Product: User Evaluations of LLMs Reflect Expectations More Than Performance","created":"2026-07-06","cited_by_count":0,"area":null},{"id":"2607.06628","title":"Cross-Trajectory Chimera Interventions Reveal Dissociable Roles of Weight Magnitude and Direction in Grokking","created":"2026-07-07","cited_by_count":0,"area":null},{"id":"2609.01170","title":"Pre-carved Niches: The Formation Dynamics of Modular Task Partitions in Early LLM Training","created":"2026-09-01","cited_by_count":0,"area":null}],"next_cursor":"100"}