[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86068-en":3,"doc-seo-86068-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},86068,687197207057,"Sage","https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0",8,"Research & Report","SETA Scaling Environments for Terminal Agents","Large language models are shifting toward agents that accomplish tasks through diverse interfaces, including web and graphical user interfaces. Terminal command line offers a general-purpose, text-based interface spanning system operations and data science workflows. Scaling terminal-agent training remains difficult because it needs diverse, coherent task instructions, executable environments, and reliable verification, while naturally grounded supervision data is limited. SETA introduces verifiable RL environments via SETA-Synth and SETA-Evol, plus SETA-Env with 4,500+ environments.","arXiv :2607 . 1089 1v 1 [ cs .AI] 12 Jul 2026  \nSETA: Scaling Environments for Terminal Agents  \nQijia Shen 1 ,2 , Zhiqi Huang3 ,4 , Vamsidhar Kamanuru5 , Aznaur Aliev6 Jay Rainton5 , Ahmed Awelkair1 ,2 ,6 , Zhichen Zeng7 , Jiajun Li7  \nShi Dong7 , Yueming Yuan7 , Boyuan Ma5 , Qizheng Zhang8 Jiwei Fu5 , Yuzhen Mao8 , Wendong Fan 1 ,2 , Ping Nie 10  \nPhilip Torr9 , Bernard Ghanem6 , Changran Hu5  \nJonathan Lingjie Li5 , Urmish Thakker5 , Guohao Li 1 ,2  \n[1](1 CAMEL-AI.org 2 Eigent.AI 3 Imperial College London 4 University College London)[ CAMEL-AI.org](1 CAMEL-AI.org 2 Eigent.AI 3 Imperial College London 4 University College London)[ 2](1 CAMEL-AI.org 2 Eigent.AI 3 Imperial College London 4 University College London)[ Eigent.AI](1 CAMEL-AI.org 2 Eigent.AI 3 Imperial College London 4 University College London)[ 3](1 CAMEL-AI.org 2 Eigent.AI 3 Imperial College London 4 University College London)[ Imperial College London](1 CAMEL-AI.org 2 Eigent.AI 3 Imperial College London 4 University College London)[ 4](1 CAMEL-AI.org 2 Eigent.AI 3 Imperial College London 4 University College London)[ University College London](1 CAMEL-AI.org 2 Eigent.AI 3 Imperial College London 4 University College London)  \n5 SambaNova 6 KAUST 7RadixArk 8 Stanford University  \n9University of Oxford 10University of Waterloo  \n§ [https://github.com/camel-ai/seta](https://github.com/camel-ai/seta)  \n [https://huggingface.co/datasets/camel-ai/SETA-Env](https://huggingface.co/datasets/camel-ai/SETA-Env)  \n [https://huggingface.co/camel-ai/Qwen3-8B-SETA-Env-RL](https://huggingface.co/camel-ai/Qwen3-8B-SETA-Env-RL)[ ](https://huggingface.co/camel-ai/Qwen3-8B-SETA-Env-RL)Original blog: SETA-Scaling-Environments-for-Terminal-Agents* .  \nAbstract  \nLarge language models (LLMs) are rapidly shifting toward agents that solve tasks through diverse interfaces, including web and graphical user interfaces (GUIs) .  \nAmong these, the terminal command line provides a text-based, general-purpose interface, covering tasks from system operations to data science and machine learning. However, scaling terminal-agent training remains challenging, as it requires diverse and coherent task instructions, executable environments, and reliable verification, while lacking naturally grounded supervision data. In this work, we propose SETA, a scalable framework for generating verifiable terminal environments for reinforcement learning (RL) . The framework consists of two pipelines sharing a unified verification mechanism: SETA-Synth converts diverse sources into standardized RL environments, and SETA-Evol further expands from existing environments with adaptive control of difficulty and diversity. Together, we construct and release SETA-Env, the largest open-source verifiable terminal RL dataset to date, containing over 4 ,500 environments. We evaluate our dataset by training Qwen3-8B with GRPO on SETA-Env, achieving 12% pass rate on Terminal-Bench 2.0, the best reported result for an RL-trained model at the 8B scale. We further observe gains on DeepSeek-V4-Flash under the same terminal agent harness, with pass@1 on Terminal-Bench 2.0 improving from 40% to 43% and pass@5 improving from 54% to 58% . These results demonstrate that SETAEnv provides high-quality training environments for terminal agents and serves asa valuable resource for advancing research on terminal-based agent learning.  \n1 Introduction  \nLarge language models (LLMs) have rapidly shifted from pure text generation towards agents that execute tools to finish tasks in interactive environments. ReAct [Yao et al., 2023] introduced a general agent framework interleaving reasoning and acting, followed by numerous works adding scaffolding and interfaces to enhance different aspects of agent capabilities. Representative efforts  \n*This work extends our January blog post, SETA-Scaling-Environments-for-Terminal-Agents.  \nPreprint.  \n\n| Dataset | Domain | Size | Grounded?† | Executable?‡ | Adaptive Difficulty?§ | RL-Validated?¶ |\n| --","cbCaicaeisjQOEE9","https://ap.wps.com/l/cbCaicaeisjQOEE9","pdf",1809043,4,1,32,"English","en",105,"# Abstract\n# Introduction\n## Related Work: Agent Interfaces and Terminal Challenges\n## Dataset Comparison and Generation Pipelines","[{\"question\":\"What is the main goal of SETA?\",\"answer\":\"SETA aims to build scalable, verifiable terminal environments for reinforcement learning, enabling reliable training of terminal agents.\"},{\"question\":\"How do SETA-Synth and SETA-Evol work together?\",\"answer\":\"SETA-Synth standardizes diverse sources into RL-ready terminal environments, while SETA-Evol expands existing environments by adaptively controlling difficulty and diversity.\"},{\"question\":\"What results are reported when training on SETA-Env?\",\"answer\":\"Training Qwen3-8B with GRPO on SETA-Env achieves a 12% pass rate on Terminal-Bench 2.0, and DeepSeek-V4-Flash shows pass@1 improving from 40% to 43% and pass@5 from 54% to 58% on the same benchmark.\"}]",1784208276,81,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"seta-scaling-environments-for-terminal-agents","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":20},"https://docshare.wps.com/document/seta-scaling-environments-for-terminal-agents/86068/",{"url":52,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-25","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What is the main goal of SETA?","Question",{"text":75,"@type":76},"SETA aims to build scalable, verifiable terminal environments for reinforcement learning, enabling reliable training of terminal agents.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How do SETA-Synth and SETA-Evol work together?",{"text":80,"@type":76},"SETA-Synth standardizes diverse sources into RL-ready terminal environments, while SETA-Evol expands existing environments by adaptively controlling difficulty and diversity.",{"name":82,"@type":73,"acceptedAnswer":83},"What results are reported when training on SETA-Env?",{"text":84,"@type":76},"Training Qwen3-8B with GRPO on SETA-Env achieves a 12% pass rate on Terminal-Bench 2.0, and DeepSeek-V4-Flash shows pass@1 improving from 40% to 43% and pass@5 from 54% to 58% on the same benchmark.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]