Hi Augment Code Team,
I’m a PhD student at UIUC and I'm working on an empirical study of coding agents. I recently read your blog post (https://www.augmentcode.com/blog/auggie-tops-swe-bench-pro?utm_source=chatgpt.com)reporting Auggie’s performance on SWE-bench Pro and was impressed to see it outperform Cursor, Claude Code, and Codex.
I would be very interested in including Auggie in my study. Would it be possible to publicly release—or privately share for research purposes—the agent trajectories from your evaluation, including those of Auggie and the baseline agents? Full interaction logs, tool calls, and generated patches would be especially valuable, though any available trajectory data would be greatly appreciated.
Thank you for your consideration!
Best,
Changshu Liu
Hi Augment Code Team,
I’m a PhD student at UIUC and I'm working on an empirical study of coding agents. I recently read your blog post (https://www.augmentcode.com/blog/auggie-tops-swe-bench-pro?utm_source=chatgpt.com)reporting Auggie’s performance on SWE-bench Pro and was impressed to see it outperform Cursor, Claude Code, and Codex.
I would be very interested in including Auggie in my study. Would it be possible to publicly release—or privately share for research purposes—the agent trajectories from your evaluation, including those of Auggie and the baseline agents? Full interaction logs, tool calls, and generated patches would be especially valuable, though any available trajectory data would be greatly appreciated.
Thank you for your consideration!
Best,
Changshu Liu