Peterande / D-FINE

D-FINE: Redefine Regression Task of DETRs as Fine-grained Distribution Refinement [ICLR 2025 Spotlight]

haha-yuki-haha / AUGUstus

🚗 AUGUstus: Autonomous Driving VLA Model with Unified Generation and Understanding. 🏋️ 512 H800 GPUs · ~2 days · 318K+ video hours · 513B+ training tokens. ⚡ 12 Hz on-vehicle deployment. SFT + GRPO + NAVSIM.

AllSpark-Research / Vela

Vela: User Behavior as Supervision Sparks LLM Intelligence — official paper, task templates, and prompts from AllSpark at Xiaohongshu.