Meta's AutoInteract trains coding agents with simulated users who push back

Meta researchers built AutoInteract, a framework that trains coding agents on conversations with simulated users who clarify vague requests, change requirements midway and correct mistakes, and report that Qwen3.5-4B rose from 6.0% to 20.6% on SWE-Together after training on those conversations. Meta's research blog states the framework keeps executable tests as the reward while synthesizing interaction as training data, departing from the usual post-training recipe of fully specified tasks with an automatic verifier and no human in the loop. At 27B, the faster configuration required slightly more corrections.

Meta's AutoInteract trains coding agents with simulated users who push back Meta reports Qwen3.5-4B rose from 6.0% to 20.6% on SWE-Together after training on conversations where users clarify, revise and correct. By RuntimeWire Staff https://runtimewire.com/author/runtimewire-staff · Published · Updated Primary source: RuntimeWire original reporting https://facebookresearch.github.io/RAM/blogs/autointeract/ runtimewire-report-autointeract-2026-11-06 Why it matters Coding agents are usually trained and graded on fully specified problems, but real users clarify and change requirements as they see output. AutoInteract's reported gains suggest interaction itself can be synthesized as training data while keeping executable tests as the reward. At 27B, the faster configuration required slightly more corrections. Meta researchers have built AutoInteract, a framework that trains coding agents on conversations with simulated users who clarify vague requests, change requirements midway and correct mistakes. In Meta's research blog https://facebookresearch.github.io/RAM/blogs/autointeract/?ref=runtimewire , the researchers say the usual post-training recipe uses fully specified tasks with an automatic verifier and no human in the loop, while real users often start with a goal that is not fully articulated and revise it after seeing output.… When you purchase through links in our articles, we may earn a small commission https://runtimewire.com/affiliate-monetization-standards . This doesn’t affect our editorial independence.