Skip to content

Self-Supervised Scaling of Terminal Environments for Scientific Domains

Oct 2026 · 2 citations · 41 references
Computer Science

TL;DR

Results indicate that existing scientific software can provide scalable, behaviorally verified supervision for terminal agents, and introduces software-in-the-loop reconstruction, a self-supervised framework that obtains reference outputs and verification targets from existing software workflows, executable programs mapping structured inputs to outputs.

Abstract

Terminal agents are increasingly deployed beyond software engineering in science and other specialized domains. Constructing training environments requires executable reference behavior and a domain-specific verifier that distinguishes semantic correctness from superficially plausible artifacts. Authoring these components for each task requires repeated engineering and limits reuse. We introduce software-in-the-loop reconstruction, a self-supervised framework that obtains reference outputs and verification targets from existing software workflows, executable programs mapping structured inputs to outputs. For each workflow, we execute multiple input configurations and partition cases into public observations and hidden evaluations. Given the instruction, input schema, and public input--output observations, an agent constructs an editable program without access to the source workflow. The candidate is evaluated on hidden configurations against workflow outputs. A hierarchical verifier combines domain-specific semantic comparison, structural validity, and anti-shortcut checks, while public feedback supports iterative revision. The construction admits additional workflows and configurations without authoring a reference solution for each task. We instantiate SWR with 500 workflows and 46 software families across six domains. Across three attempts per task, Qwen3.8-Max solves 838 tasks and produces 1,422 verified trajectories, which we oversample to 3,000 reconstruction-only training examples. Supervised fine-tuning of Qwen3.8-27B improves mean Terminal-Bench 2 performance from 47.94% to 53.56% across three seeds and achieves the highest mean among four matched-token corpus controls on all four reported evaluations. These results indicate that existing scientific software can provide scalable, behaviorally verified supervision for terminal agents.

View source

Similar papers

#artificial intelligence Preprint Oct 2026

POSEIDON: Scaling Training Environments for Scientific Agents from Code Repositories

Training environments remain a major bottleneck for agents in data-driven scientific discovery. Existing collections of executable and verifiable training environments are small, and their construction remains difficult to scale. To address this limitation, we introduce Poseidon, an automated pipeline for generating di...

Qi-Fan Zhang, Yu-Han Li, Nuo Chen et al. · 0 citations

CompoWorld: Compositional Environment Scaling for General Agents

This work introduces Compositional Environment Scaling (\textbf{CompoWorld}), which expands the task space by composing a finite library of reusable services, and guides reinforcement learning toward full task completion by emphasizing criteria with lower pass rates within each rollout group.

Xiao-Wen Yang, Wei-Yi Xu, Wen Da et al. · 1 citation
#artificial intelligence Preprint Oct 2026

Harness Engineering for Software Engineering via Modular Executable Dev-Primitives

Large language models (LLMs) equipped with terminal access have demonstrated strong capabilities in automating software engineering tasks. However, existing agents remain brittle on long-horizon workflows, where they must repeatedly reconstruct program state scattered across source files, configurations, tests, depende...

Hai-Bo Jin, Xin-Jie Li, Peng Kuang et al. · 0 citations
#artificial intelligence Preprint Oct 2026

Davinci-Tool: Scaling Agentic Tool-Use Supervision with Verifiable Executable Environments

Reliable agentic tool use requires training data that pair diverse multi-step workflows with executable evidence that the requested outcomes were achieved. Scaling either trajectory coverage or verification alone is insufficient as new environments introduce distinct states, rules, and dependencies that must also be re...

Xiao-Yang Yuan, Ji Zhao, Xing-Yu Hu et al. · 0 citations
#artificial intelligence Preprint Oct 2026

WebAnything: From the Wild Web to Scalable Agent Environments

This work presents WebAnything, a fully automated framework that reconstructs source-grounded, reusable Web environments from real URLs without per-site manual implementation or repair after initial setup and authorization.

Tian-Ci Xie, Zi-Kun Cai, Chao-Yun Zhang et al. · 0 citations
Preprint Oct 2026

SoftwareGym: Scaling Computer-Use Agent Training with Diverse Synthesized Software Environments

Computer-use agents (CUAs) are rapidly improving, yet synthesizing diverse and reliable software environments at scale remains challenging. Native application environments require substantial construction and integration effort, which limits application coverage and increases the risk of application-specific overfittin...

Tian Jin, Wen-Hao Wang, Jing Kang et al. · 0 citations

Related blog posts

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.