SWE-chat v2 Released with 230K Real-World Coding Agent Prompts

SWE-chat v2, the largest real-world user coding agent interaction dataset, expands 3.5x to 230K prompts across 18K sessions, adding agent skills and subagent trajectories. It is maintained by SALT-NLP and has already supported multiple research studies.

2026-10-03 ~ 2026-10-03 · 2 related posts

• SWE-chat V2 ships 3.5x larger with agent skills and subagent trajectories — Diyi_Yang · 2026-10-03
• SWE-chat v2 expands to 230K real-user coding agent prompts from 18K sessions — Diyi_Yang · 2026-10-03