Not All Steps are Informative: On the Linearity of LLMs' RLVR Training
Paper • 2601.04537 • Published
RLVR Linearity
RL training and evaluation datasets, and checkpoints in 'Linear Dynamics in the RLVR Training of Large Language Models'
RL training and evaluation datasets, and checkpoints in 'Linear Dynamics in the RLVR Training of Large Language Models'