Papers2

#long-context training

Molmo2: Open Weights and Data for Vision-Language Models with Video Understanding and Grounding

Christopher Clark, Jieyu Zhang et al.Jan 15arXiv

Molmo2 is a family of vision-language models that can watch videos, understand them, and point to or track things over time using fully open weights, data, and code.

#vision-language model#video grounding#pointing and tracking

Not triaged yet

Distribution-Aligned Sequence Distillation for Superior Long-CoT Reasoning

Intermediate

Shaotian Yan, Kaiyuan Liu et al.Jan 14arXiv

The paper introduces DASD-4B-Thinking, a small (4B) open-source reasoning model that scores like much larger models on hard math, science, and coding tests.

#sequence-level distillation#divergence-aware sampling#temperature-scheduled learning

Not triaged yet