Deep Learning

The Knowledge Microscope: Features as Better Analytical Lenses than Neurons ACL 2025 The Language of Motion: Unifying Verbal and Non-verbal Language of 3D Human Motion CVPR 2025 The Last Iterate Advantage: Empirical Auditing and Principled Heuristic Analysis of Differentially Private SGD ICLR 2025 The late-stage training dynamics of (stochastic) subgradient descent on homogeneous neural networks COLT 2025 The Law of Knowledge Overshadowing: Towards Understanding, Predicting, and Preventing LLM Hallucination ACL 2025 The LLM Already Knows: Estimating LLM-Perceived Question Difficulty via Hidden Representations EMNLP 2025 The Local Learning Coefficient: A Singularity-Aware Complexity Measure AISTATS 2025 The Logical Implication Steering Method for Conditional Interventions on Transformer Generation ICML 2025 The Master Key Filters Hypothesis: Deep Filters Are General AAAI 2025 The Medium Is Not the Message: Deconfounding Document Embeddings via Linear Concept Erasure EMNLP 2025 The Mirage of Model Editing: Revisiting Evaluation in the Wild ACL 2025 The Missing Parts: Augmenting Fact Verification with Half Truth Detection EMNLP 2025 The Noisy Laplacian: a Threshold Phenomenon for Non-Linear Dimension Reduction ICML 2025 TheoremExplainAgent: Towards Video-based Multimodal Explanations for LLM Theorem Understanding ACL 2025 Theorem Prover as a Judge for Synthetic Data Generation ACL 2025 Theorem-Validated Reverse Chain-of-Thought Problem Generation for Geometric Reasoning EMNLP 2025 Theoretical Analysis of Hierarchical Language Recognition and Generation by Transformers without Positional Encoding ACL 2025 Theoretical Convergence Guarantees for Variational Autoencoders AISTATS 2025 Theoretical Insights into Fine-Tuning Attention Mechanism: Generalization and Optimization IJCAI 2025 Theoretically Unmasking Inference Attacks Against LDP-Protected Clients in Federated Vision Models ICML 2025 Theory, Analysis, and Best Practices for Sigmoid Self-Attention ICLR 2025 Theory-Inspired Deep Multi-View Multi-Label Learning with Incomplete Views and Noisy Labels CVPR 2025 Theory of Agreement-on-the-Line in Linear Models and Gaussian Data AISTATS 2025 Theory on Mixture-of-Experts in Continual Learning ICLR 2025 Theory on Score-Mismatched Diffusion Models and Zero-Shot Conditional Samplers ICLR 2025