Matched-Context Group Policy Optimization for LLM Agent Training
We propose Matched-Context Group Policy Optimization (MCGPO), a procedure for constructing weighted action preferences from already collected trajectories.
ProofPaper ↗
Key points
- Training long-horizon language model agents requires assigning credit to intermediate decisions from sparse terminal feedback.
- Within a finite group of sibling rollouts, useful action comparisons depend on their context, evidence for both alternatives, and how that evidence enters learning.
- Two-action support selects among ordered exact and coarse matching rules.
- Fixed folds separate candidate decisions from the terminal outcomes used to orient their preferences, and support-dependent evidence weights scale an auxiliary pairwise objective alongside shared GraphGPO task learning.
Sources (1)
- [1]Matched-Context Group Policy Optimization for LLM Agent TrainingarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 03:41 PM
We propose Matched-Context Group Policy Optimization (MCGPO), a procedure for constructing weighted action preferences from already collected trajectories.
Training long-horizon language model agents requires assigning credit to intermediate decisions from sparse terminal feedback.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 9, 2026AdaC-GRPO: Towards Mastering Long-Horizon Visual Sudoku Puzzles via Adaptive Curriculum Learning
- Oct 9, 2026Qwen/Qwen-Image-2.1-Turbo
- Oct 9, 2026Introducing Clef-omni with full multimodality, plus a faster Clef and a cheaper Clef-flash
- Oct 8, 2026ConwayResearch/Underdog-Saluki-27B-1.0
- Oct 8, 2026SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference
- Oct 8, 2026REMORY: Learning Residual Memory for Context Compaction
