Open Models Almanac
Methodspost-trainingreward modelling

general family · filed under post-training

Groupwise Agentic Grading

Groupwise agentic grading is a reward-modelling method for reinforcement learning that provides reward signals beyond binary test cases.1 Instead of assigning the same reward to all solutions that pass the test cases, it compares solutions within each group to distinguish problem-solving quality and behaviors.2

How this definition was written

Summarised by glm-5.3-flash from 1 source; every sentence is checked against the quote it cites.

  1. we introduce groupwise agentic grading to provide more informative re- ward signals beyond binary test cases — MiMo V2 6 technical report
  2. Instead of assigning the same reward to all solutions that pass the test cases, we compare solutions within each group to distinguish problem-solving quality and behaviors — MiMo V2 6 technical report
sources
3
models
2
lab adopt it
1
strongest
core

Documented in

Evidence

4 quotes from 3 sources, one card per model, each opening with its strongest.

How sources treat it
used 2core 2

MiMo-V2.6-Flash

Xiaomiadopted2 sources

we introduce groupwise agentic grading to provide more informative reward signals beyond binary test cases

coreevaluation onlyin MiMo-V2.6 RL trainingXiaomi
2 more from this model

An agentic grader compares rollouts within each group

coretraining objectivein MiMo-V2.6-Flash-RLXiaomi

the compute devoted to groupwise agentic grading.

usedtraining objectivein MiMo-V2.6

MiMo-V2.6-Pro

Xiaomiadopted2 sources

we introduce groupwise agentic grading to provide more informative reward signals beyond binary test cases

coreevaluation onlyin MiMo-V2.6 RL trainingXiaomi
2 more from this model

An agentic grader compares rollouts within each group

usedtraining objectivein MiMo-V2.6-Pro-RLXiaomi

the compute devoted to groupwise agentic grading.

usedtraining objectivein MiMo-V2.6
↑