Tag: Group Relative Self-Distillation
Browse our exclusive articles!
SnapFlow: Fast One-Step Action Generation for VLAs
Discover SnapFlow, a one-step action generation method for flow-matching VLAs that boosts efficiency and speed without sacrificing performance.
π² Boosts Long-Context Reasoning in Large Language Models
Discover how π² uses structured reasoning data to enhance long-context reasoning accuracy in large language models with open-source tools.
SPARE: Efficient Self-Distillation for Model Unlearning
Discover SPARE, a two-stage self-distillation method for efficient unlearning in image generation models, balancing forgetting and retention effectively.
UI-Voyager: Self-Evolving GUI Agent Learning from Failures
Discover UI-Voyager, a self-evolving GUI agent that improves mobile automation by learning from failed experiences with 81% success rate.