Trending Research
FastVLM: Efficient Vision Encoding for Vision Language Models
apple/ml-fastvlm
•
•
At different operational resolutions, the vision encoder of a VLM can be optimized along two axes: reducing encoding latency and minimizing the number of visual tokens passed to the LLM, thereby lowering overall latency.
Group-in-Group Policy Optimization for LLM Agent Training
In this work, we propose Group-in-Group Policy Optimization (GiGPO), a novel RL algorithm that achieves fine-grained credit assignment for LLM agents while preserving the appealing properties of group-based RL: critic-free, low memory, and stable convergence.
VisionReasoner: Unified Visual Perception and Reasoning via Reinforcement Learning
dvlab-research/VisionReasoner
•
•
Large vision-language models exhibit inherent capabilities to handle diverse visual perception tasks.
SOAP: Style-Omniscient Animatable Portraits
Creating animatable 3D avatars from a single image remains challenging due to style limitations (realistic, cartoon, anime) and difficulties in handling accessories or hairstyles.
MTVCrafter: 4D Motion Tokenization for Open-World Human Image Animation
To tackle this problem, we propose MTVCrafter (Motion Tokenization Video Crafter), the first framework that directly models raw 3D motion sequences (i. e., 4D motion) for human image animation.
AdaptThink: Reasoning Models Can Learn When to Think
Recently, large reasoning models have achieved impressive performance on various tasks by employing human-like deep thinking.
Spherical Channels for Modeling Atomic Interactions
We propose the Spherical Channel Network (SCN) to model atomic energies and forces.
Grouping First, Attending Smartly: Training-Free Acceleration for Diffusion Transformers
We validate GRAT on pretrained Flux and HunyuanVideo for image and video generation, respectively.
KORGym: A Dynamic Game Platform for LLM Reasoning Evaluation
Recent advancements in large language models (LLMs) underscore the need for more comprehensive evaluation methods to accurately assess their reasoning capabilities.
Absolute Zero: Reinforced Self-play Reasoning with Zero Data
LeapLabTHU/Absolute-Zero-Reasoner
•
•
Reinforcement learning with verifiable rewards (RLVR) has shown promise in enhancing the reasoning capabilities of large language models by learning directly from outcome-based rewards.
