# Group-in-Group Policy Optimization for LLM Agent Training

Introduces Group-in-Group Policy Optimization (GiGPO), an RL method for assigning step-level credit in multi-turn agents, evaluated on WebShop among other tasks.

- Catalog URL: https://awesomecua.com/resources/arxiv-2505-10978/
- Original source: https://arxiv.org/abs/2505.10978
- Kind: paper
- Review: curated
- Scope: adjacent
- Environments: unspecified
- Topics: reinforcement-learning, reward-models
- Evidence checked: 2026-11-05T23:01:45Z
- ID: arxiv:2505.10978

## Limitations

The annotation and scope were reviewed against the complete retained arXiv abstract (v3). WebShop is one of several agent benchmarks used to evaluate the general training method. Experiments, reported performance and artifact availability have not been independently reproduced.

- Record JSON: https://awesomecua.com/resources/arxiv-2505-10978/record.json

## Authors

Lang Feng, Zhenghai Xue, Tingcong Liu, Bo An

## Citation exports

- BibTeX: https://awesomecua.com/resources/arxiv-2505-10978/citation.bib
- CSL JSON: https://awesomecua.com/resources/arxiv-2505-10978/citation.csl.json

## Recorded arXiv snapshot

- Version: 3
- Version source: https://arxiv.org/abs/2505.10978v3
- First submission: 2025-05-16T08:26:59Z
- Version updated: 2025-10-28T15:11:36Z
- Abstract SHA-256: a3faa5753692553b6ef3309532d37c2607e3cd9733475979ff456f588fc2a542
- Abstract hash normalization: sha256-utf8-whitespace-collapsed-v1

## License declarations

- code: not recorded
- data: not recorded
- document: not recorded
- weights: not recorded

## Primary evidence

- https://export.arxiv.org/api/query?id_list=2501.12485%2C2502.01456%2C2502.03214%2C2502.03971%2C2502.06776%2C2502.12561%2C2502.13417%2C2502.17419%2C2502.20694%2C2503.02878%2C2503.16365%2C2503.20036%2C2504.07491%2C2504.12682%2C2505.03733%2C2505.05262%2C2505.05895%2C2505.07062%2C2505.10978%2C2505.17629%2C2505.20075%2C2505.23678%2C2505.24298%2C2505.24876%2C2506.00520%2C2506.01952%2C2506.06287%2C2506.14079%2C2506.19998%2C2506.24045%2C2507.01006%2C2507.04370%2C2507.06261%2C2507.10202%2C2507.10644%2C2507.11662%2C2507.23773%2C2508.09736%2C2508.16271%2C2508.16763%2C2508.18265%2C2508.19828%2C2508.20018%2C2508.21475%2C2509.02547%2C2509.08826%2C2509.08827%2C2509.11067%2C2509.13313%2C2509.21733&max_results=50 (retrieved 2026-11-05T21:13:58Z; SHA-256 62b85cd4bee4c668046d5f6eb67dedef915ba699d6765891016403c863691a73)
- https://export.arxiv.org/api/query?id_list=2505.05262%2C2505.05849%2C2505.05895%2C2505.06507%2C2505.07003%2C2505.07062%2C2505.07473%2C2505.10468%2C2505.10887%2C2505.10978%2C2505.11717%2C2505.12053%2C2505.12490%2C2505.12728%2C2505.13291%2C2505.13887%2C2505.14030%2C2505.14454%2C2505.14646%2C2505.14963%2C2505.15801%2C2505.15816%2C2505.15859%2C2505.17020%2C2505.17629%2C2505.17735%2C2505.18746%2C2505.19713%2C2505.20023%2C2505.20075%2C2505.20732%2C2505.20795%2C2505.21055%2C2505.22304%2C2505.22429%2C2505.22648%2C2505.23450%2C2505.23559%2C2505.23678%2C2505.24298%2C2505.24876%2C2506.00520%2C2506.00641%2C2506.01423%2C2506.01520%2C2506.01952%2C2506.02153%2C2506.03231%2C2506.03569%2C2506.04301&max_results=50 (retrieved 2026-11-05T23:01:45Z; SHA-256 72d781966615fe4da127a9c089d5d6ab85838feb4205ef1ebd446ef97189bf70)