SPEED-Bench: A Unified and Diverse Benchmark for Speculative Decoding
Paper • 2604.09557 • Published • 10
Yury Panikov
panikov
AI & ML interests
None yet
Recent Activity
upvoted a paper about 8 hours ago
SPEED-Bench: A Unified and Diverse Benchmark for Speculative Decoding upvoted a paper about 8 hours ago
From Reasoning to Agentic: Credit Assignment in Reinforcement Learning for Large Language Models upvoted a paper about 8 hours ago
Mobile GUI Agent Privacy Personalization with Trajectory Induced Preference OptimizationOrganizations
None yet
upvoted 12
papers about 8 hours ago
From Reasoning to Agentic: Credit Assignment in Reinforcement Learning for Large Language Models
Paper • 2604.09459 • Published • 11
Mobile GUI Agent Privacy Personalization with Trajectory Induced Preference Optimization
Paper • 2604.11259 • Published • 11
TRACE: Capability-Targeted Agentic Training
Paper • 2604.05336 • Published • 13
Agentic Aggregation for Parallel Scaling of Long-Horizon Agentic Tasks
Paper • 2604.11753 • Published • 13
Solving Physics Olympiad via Reinforcement Learning on Physics Simulators
Paper • 2604.11805 • Published • 15
Efficient RL Training for LLMs with Experience Replay
Paper • 2604.08706 • Published • 17
Tracing the Roots: A Multi-Agent Framework for Uncovering Data Lineage in Post-Training LLMs
Paper • 2604.10480 • Published • 17
Introspective Diffusion Language Models
Paper • 2604.11035 • Published • 19
CocoaBench: Evaluating Unified Digital Agents in the Wild
Paper • 2604.11201 • Published • 32
Pseudo-Unification: Entropy Probing Reveals Divergent Information Patterns in Unified Multimodal Models
Paper • 2604.10949 • Published • 38
CodeTracer: Towards Traceable Agent States
Paper • 2604.11641 • Published • 36
upvoted 8
papers about 9 hours ago
Strips as Tokens: Artist Mesh Generation with Native UV Segmentation
Paper • 2604.09132 • Published • 48
Attention Sink in Transformers: A Survey on Utilization, Interpretation, and Mitigation
Paper • 2604.10098 • Published • 69
The Past Is Not Past: Memory-Enhanced Dynamic Reward Shaping
Paper • 2604.11297 • Published • 92
Cactus: Accelerating Auto-Regressive Decoding with Constrained Acceptance Speculative Sampling
Paper • 2604.04987 • Published • 3
Initialisation Determines the Basin: Efficient Codebook Optimisation for Extreme LLM Quantization
Paper • 2604.08118 • Published • 2
MixFlow: Mixed Source Distributions Improve Rectified Flows
Paper • 2604.09181 • Published • 3
Robust Reasoning Benchmark
Paper • 2604.08571 • Published • 4
On Semiotic-Grounded Interpretive Evaluation of Generative Art
Paper • 2604.08641 • Published • 4