# qiqihezh/agentic-grpo-longhorizon

<https://repo.masters-masters.com/en/repos/qiqihezh/agentic-grpo-longhorizon>

[Home](https://repo.masters-masters.com/en) › [qiqihezh](https://repo.masters-masters.com/en/owners/qiqihezh) › [agentic-grpo-longhorizon](https://repo.masters-masters.com/en/repos/qiqihezh/agentic-grpo-longhorizon)

qiqihezh/agentic-grpo-longhorizon has 266 stars on GitHub. Its RepoMasters card: Alpha tier. Fixing GRPO training collapse in long-horizon multi-tool agents…

| Field | Value |
| --- | --- |
| Owner | [qiqihezh](https://repo.masters-masters.com/en/owners/qiqihezh) |
| Stars | 266 |
| Forks | 17 |
| Language | [Python](https://repo.masters-masters.com/en/languages/python) |
| License | None |
| Created | 2026 |
| Topics | agentic-ai, grpo, long-horizon, process-reward-model, qwen, reinforcement-learning, tau-bench, tool-calling, multi-turn-agents |
| Tier | [Alpha (250 to 699 ★)](https://repo.masters-masters.com/en/ranks/b_movie) |
| View on GitHub | [qiqihezh/agentic-grpo-longhorizon](https://github.com/qiqihezh/agentic-grpo-longhorizon) |

## Repository

Fixing GRPO training collapse in long-horizon multi-tool agents. A lightweight PRM-Lite + LATA joint approach achieves +37% over vanilla GRPO on τ-bench airline (50-task, multi-turn).

The tier for repos with 250 to 699 ★. Published odds per card pulled: 20%. Rarity: Uncommon.

## More repos by qiqihezh

- [qiqihezh/deepresearch-agent · ★ 175 · Python · Prototype](https://repo.masters-masters.com/en/repos/qiqihezh/deepresearch-agent)

## Nearby in Python

- [julianmb/q38rocm · ★ 266 · Python · Alpha](https://repo.masters-masters.com/en/repos/julianmb/q38rocm)
- [Gen-Verse/Recuris · ★ 266 · Python · Alpha](https://repo.masters-masters.com/en/repos/Gen-Verse/Recuris)
- [zjwfufu/SMD · ★ 266 · Python · Alpha](https://repo.masters-masters.com/en/repos/zjwfufu/SMD)
- [kassoulet/soundconverter · ★ 266 · Python · Alpha](https://repo.masters-masters.com/en/repos/kassoulet/soundconverter)
- [erezsh/plyplus · ★ 266 · Python · Alpha](https://repo.masters-masters.com/en/repos/erezsh/plyplus)
- [ClusterLabs/pcs · ★ 266 · Python · Alpha](https://repo.masters-masters.com/en/repos/ClusterLabs/pcs)

---
Data: GitHub · Illustration drawn by RepoMasters from the repo’s ID and its language color.