MMFlash: Faster Speculative Decoding for Vision-Language Models Paper • 2611.18482 • Published 9 days ago • 44
MMFlash Collection MMFlash draft models for multimodal speculative decoding (Qwen3.5-4B/9B, Gemma 4 E4B-it, Qwen3.8-27B) • 5 items • Updated 6 days ago • 9
TRACE: Rollout-Guided Quantization-Aware Training for FP4 Reinforcement Learning of MoE Language Models Paper • 2610.07767 • Published Oct 6 • 43
Open-SWE-Traces: Advancing Dual-Mode Multilingual Distillation for Software Engineering Agents Paper • 2606.16038 • Published Jun 14 • 6
EarlyTom: Early Token Compression Completes Fast Video Understanding Paper • 2605.30010 • Published May 28 • 29
HippoCamp: Benchmarking Contextual Agents on Personal Computers Paper • 2604.01221 • Published Apr 1 • 33