hermes-ai.netis an unofficial, independent community guide to Hermes Agent, with localized docs, release notes, desktop notes, and practical setup paths.

paperReasoning Post Training

AC2 Beats GRPO on Math Proofs Using 2.5x Fewer Decoding FLOPs

A new actor-critic recipe called AC2 trains LLMs on long reasoning tasks without rolling every trajectory to completion, hitting GRPO's peak with 2.5x fewer decoding FLOPs.

$AC2 Beats GRPO on Math Proofs Using 2.5x Fewer Decoding FLOPs$

Source

AlphaSignal

Published

Sep 30, 2026, 8:13 AM

Author

AlphaSignal Newsroom

Read

1 min read

A new actor-critic recipe called AC2 trains LLMs on long reasoning tasks without rolling every trajectory to completion, hitting GRPO's peak with 2.5x fewer decoding FLOPs.

Reporting is indexed from AlphaSignal. Rights remain with the original publisher and cited sources.

Read original report

Next reads

Harvard's Projection Sampling Makes Standard Fine-Tuning Match Reinforcement Learning

AlphaSignal · paper

AC2 Beats GRPO on Math Proofs Using 2.5x Fewer Decoding FLOPs

Next reads

Harvard's Projection Sampling Makes Standard Fine-Tuning Match Reinforcement Learning

Apple's RLTL;DR Teaches AI to Learn From Its Own Failures

UC Berkeley's EasyPPO Stops AI Training Collapses With Three Critic Fixes