ICML 2025
3,384 papers
Top Research Directions
Research directions most represented by ICML 2025 papers.
Top Authors
Impactful Papers
Papers from ICML 2025 that shaped research clusters.
Foundational
Efficient Source-free Unlearning via Energy-Guided Data Synthesis and Discrimination-Aware Multitask
foundational
Resolving Gradient Conflict in Machine U
Reflection-Window Decoding: Text Generation with Selective Refinement
foundational
Real-Time Error Detection via Token-Leve
ELITE: Enhanced Language-Image Toxicity Evaluation for Safety
foundational
Emergent Safety Risks from Benign Multim
SafeArena: Evaluating the Safety of Autonomous Web Agents
foundational
The Safety Alignment Gap in Autonomous A
Scaling Large Motion Models with Million-Level Human Motions
foundational
Part Level Decomposition for Human Motio
RWKVQuant: Quantizing the RWKV Family with Proxy Guided Hybrid of Scalar and Vector Quantization
foundational
Activation-Aware Codebook Optimization f
any4: Learned 4-bit Numeric Representation for LLMs
foundational
Activation-Aware Codebook Optimization f
CLOVER: Cross-Layer Orthogonal Vectors Pruning
foundational
Spectral Decomposition of Transformer At
Olica: Efficient Structured Pruning of Large Language Models without Retraining
foundational
Spectral Decomposition of Transformer At
Weak-to-Strong Jailbreaking on Large Language Models
foundational
Why Safety Alignment Fails Beyond the In
Influential
Imagine While Reasoning in Space: Multimodal Visualization-of-Thought
influential
Interleaved Multimodal Chain of Thought
Proactive Agents for Multi-Turn Text-to-Image Generation Under Uncertainty
influential
Automated Prompt Refinement for Generati
Video Prediction Policy: A Generalist Robot Policy with Predictive Visual Representations
influential
Video Generation as a Foundation for Rob
Lightspeed Geometric Dataset Distance via Sliced Optimal Transport
influential
Resolving Peak Overlap in Powder Diffrac
Near-Optimal Sample Complexity for MDPs via Anchoring
influential
Stabilizing Average Reward Reinforcement
Communicating Activations Between Language Model Agents
influential
Latent State Communication in Neural Pip
A Memory Efficient Randomized Subspace Optimization Method for Training Large Language Models
influential
Activation Compression and Mixed Precisi