Search papers, labs, and topics across Lattice.
This paper introduces Echoverse, a framework for creating deep, evolving environments that enhance the training of computer-use agents by integrating stateful applications with a co-evolution loop. By focusing on behavioral depth, targeted interactions, and model improvement, Echoverse significantly boosts the performance of a 9B model from 36.5% to 67.1% across various evaluation splits, demonstrating the importance of environment design in agent training. The research reveals that deeper environments not only improve accuracy but also facilitate knowledge transfer across different interfaces and applications.
Training agents in deep, evolving environments can dramatically enhance their performance, with a 9B model achieving a 30.6% accuracy increase through targeted design.
Computer-use agents learn from what their actions change, so training one needs applications it can act on, break and reset. The applications that matter most are login-gated and stateful, so synthetic environments stand in for them. Recent pipelines generate such environments in bulk, which moves the bottleneck from how many exist to what is inside each one. The returns, we find, come from three properties: how much behavioural depth an environment carries, whether it targets the interaction an agent actually fails, and whether it improves alongside the model. We present Echoverse, which compiles specifications into stateful applications whose tasks are graded against the application's own database, and a co-evolution loop that reads every graded rollout twice: as repairs to the environment, its tasks and its verifier, and as training signal for the model. Trained on twelve such environments, a 9B model improves from $36.5\%$ to $67.1\%$ across fourteen evaluation splits, within fourteen points of the much larger frontier model that taught it. We examine each property in turn. On the same domains, shallow environments push live-site accuracy below the base model ($80.0 \to 75.0$) while deep ones raise it ($80.0 \to 85.0$ and $48.0 \to 65.0$); drilling one interface control across many renderings transfers to held-out widget families and to the open web; and repairing a single environment lifts the model trained on it from $16.2\%$ to $38.5\%$. The same worlds serve as reinforcement-learning environments, where a reward combining the grounded verifier with a dense per-step judge raises held-out score from $58.8\%$ to $68.0\%$. We release four environments as a benchmark, with their applications, seed data and grounded graders. Code: https://aka.ms/echoverse