Qwen releases QwenGyre, an RL framework for extreme-long-horizon agent training
ResearchCrafty1804 · reddit · 2026-10-12
Qwen open-sourced QwenGyre, an elastic RL framework for training agents on extreme-long-horizon tasks. Its Elastic Scheduler reallocates GPUs between rollout and training without interrupting live executions (1.85x faster than Colocate, 1.78x vs Async), while a Trajectory Processor turns branching executions (compaction, sub-agents, retries) into prefix-shared trajectory trees with deduplicated samples. Training Qwen3.8-2.4T-A95B with 700K-token rollouts lifted NL2RepoBench from 52.5% to 58.5% in 48 steps, cutting end-to-end training time 44% vs Async. Paper on arXiv.
More from coding & agent
- Running coding agents in YOLO mode: a practical 5-layer guardrail setup — Arindam_1729 · 2026-10-13
- Cloudflare Agents SDK adds web_search (Exa/Linkup/ceramic) and Markdown web_fetch tools — irvinebroque · 2026-10-13
- 45 minutes with AI: failure cases, BQ datasets, and a generative UI — rseroter · 2026-10-13
- DeepSeek desktop agent app reviewed: model promising, harness trails Cursor in 5 ways — casper_hansen_ · 2026-10-13
- Hone, founded by ex-Cognition and Mercor staff, builds business-running agents — nmasc_ · 2026-10-13
- How tldraw built a spatial harness for coding agents on the canvas — max__drake · 2026-10-13