GFT: From Imitation to Reward Fine-Tuning with Unbiased Group Advantages and Dynamic Coefficient Rectification
Paper • 2604.14258 • Published • 22
xuwang
wx91726
AI & ML interests
None yet
Recent Activity
upvoted a paper 3 months ago
Visual Generation Unlocks Human-Like Reasoning through Multimodal World Models upvoted a paper 4 months ago
N3D-VLM: Native 3D Grounding Enables Accurate Spatial Reasoning in Vision-Language ModelsOrganizations
None yet