2 papers
cs.CL2026
Are Dense Rewards Worth Their Tokens in GRPO?
Vikash Singh, Weicong Chen, Debargha Ganguly +16
RLVR trains language models to reason by rewarding correct final answers. A widely used algorithm, group-relative policy optimization (GRPO), gives every token of a solution the sa…
math.NT2026
On the Space Defined by the Modular Transformation Law for Multiple Eisenstein Series
Katsumi Kina
Motivated by a modular transformation law for multiple Eisenstein series established by Yu, we introduce spaces of holomorphic functions satisfying this law together with an additi…