Papers5

#Best-of-N

Specificity-aware reinforcement learning for fine-grained open-world classification

Samuele Angheben, Davide Berasi et al.Mar 3arXiv

This paper teaches AI to name things in pictures very specifically (like “golden retriever” instead of just “dog”) without making more mistakes.

#open-world classification#fine-grained recognition#large multimodal models

From Scale to Speed: Adaptive Test-Time Scaling for Image Editing

Intermediate

Xiangyan Qu, Zhenlong Yuan et al.Feb 24arXiv

This paper speeds up and improves AI image editing by giving hard edits more attention and easy edits less, just like a smart coach.

#adaptive test-time scaling#image chain-of-thought#image editing

Sci-CoE: Co-evolving Scientific Reasoning LLMs via Geometric Consensus with Sparse Supervision

Intermediate

Xiaohan He, Shiyang Feng et al.Feb 12arXiv

Sci-CoE is a two-stage training method that helps one language model learn to both solve science problems and check those solutions with very little labeled data.

#scientific reasoning#co-evolution#solver-verifier

WorldCompass: Reinforcement Learning for Long-Horizon World Models

Beginner

Zehan Wang, Tengfei Wang et al.Feb 9arXiv

WorldCompass teaches video world models to follow actions better and keep pictures pretty by using reinforcement learning after pretraining.

#world models#reinforcement learning#clip-level rollout

RE-TRAC: REcursive TRAjectory Compression for Deep Search Agents

Intermediate

Jialiang Zhu, Gongrui Zhang et al.Feb 2arXiv

Re-TRAC is a new way for AI search agents to learn from each try, write a clean summary of what happened, and then use that summary to do better on the next try.

#Re-TRAC#trajectory compression#deep research agents