Multiverse: Your Language Models Secretly Decide How to Parallelize and Merge Generation Paper • 2506.09991 • Published Jun 11 • 56
Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning Paper • 2506.01939 • Published Jun 2 • 173
DeTikZify: Synthesizing Graphics Programs for Scientific Figures and Sketches with TikZ Paper • 2405.15306 • Published May 24, 2024 • 8
LASP-2: Rethinking Sequence Parallelism for Linear Attention and Its Hybrid Paper • 2502.07563 • Published Feb 11 • 24