[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85828-en":3,"doc-seo-85828-105":29,"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":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85828,8796095461564,"Liam","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","IdeaTrail Full Process Agent Trajectories for Scientific Ideation","Scientific research is a complex, multi-stage workflow that unfolds through literature search, paper reading, tool use, evidence checking, cross-paper synthesis, brainstorming, iterative writing, and careful rejection of weak directions. IdeaTrail introduces a multi-turn process-trajectory dataset for scientific ideation and proposal generation. Each instance records evidence gathering through tool interactions and intermediate artifact evolution to an idea-level or proposal-level endpoint. A Generator–Advisor synthesis loop constrains grounded, causally ordered, and leakage-free trajectories under a cutoff-aware tool horizon, providing both data and a recipe for process-supervision.","arXiv :2607 . 10 144v 1 [ cs .AI] 11 Jul 2026  \nIdeaTrail: Full-Process Agent Trajectories for Scientific  \nIdeation  \nHengquan Guo  \nShanghaiTech University  \n[guohq@shanghaitech. edu. cn](guohq@shanghaitech. edu. cn)  \nAbstract  \nScientific research is a complex, multi-stage workflow rather than a single act of text generation.  \nThe ideation process typically emerges through literature search, paper reading, tool use, claim checking, cross-paper synthesis, brainstorming, rejection of weak directions, and iterative writing.  \nExisting resources capture individual components of this process, but datasets that jointly record tool use, evidence acquisition, intermediate artifact evolution, and idea-or proposal-level endpoints remain limited. This report introduces IdeaTrail, a multi-turn process-trajectory dataset for scientific ideation and proposal generation. Each instance records a research process from evidence gathering to either idea selection or proposal construction. Rather than freely fabricating trajectories, IdeaTrail starts from human-selected high-quality research papersand proposal artifacts and uses a Generator–Advisor synthesis loop. The Generator produces the visible trajectory through actions, observations, and artifact edits, while the Advisor has access to the full generation context and checks grounding, causal order, naturalness, and leakage from hidden targets. This reverse-to-forward procedure produces multi-turn research data that remains aligned with real scientific artifacts while approximating the uncertainty, evidence use, and staged convergence of research practice. IdeaTrail provides both a dataset and a general recipe for synthesizing process-supervision data for scientific research agents.  \nDataset: IdeaTrail  \n1 Introduction  \nScientific research is a long-horizon agentic workflow, not a single step of text generation. Research ideas and proposals emerge through multiple rounds of literature search, tool use, paper reading, evidence checking, cross-paper synthesis, brainstorming, idea selection, and proposal writing. Recent AI scientist systems increasingly automate substantial portions of this workflow, from hypothesis generation to experimentation and manuscript production [3, 4 , 5] . In parallel, datasets and training frameworks have begun to formalize scientific ideation as idea evaluation, literature-grounded generation, or decomposed discovery tasks [1, 2 , 6 , 7] . However, public supervision for long-horizon scientific ideation trajectories remains limited. Existing resources typically emphasize final ideas, selected intermediate decisions, or system outputs rather than the coupled sequence of evidence discovery, tool interaction, intermediate reasoning, artifact evolution, idea selection, and proposal construction.  \nThis report introduces IdeaTrail, a dataset of reverse-synthesized process supervision for scientific ideation. Each IdeaTrail instance records a multi-turn research trajectory in which an agent moves from a broad research query to either an idea-level or proposal-level endpoint.  \nThe trajectory is represented as a multi-turn message stream with tool invocations, evidencegathering steps, reasoning turns, and artifact updates. For proposal-extension trajectories, the research proposal serves as the endpoint of scientific ideation and the starting point for downstream implementation. Such a proposal should be concrete enough to support coding, experimentation, and verification by subsequent agents. It should specify the research question, the core mechanism or method, the expected contribution, the evaluation setup, and the risks.  \nTrajectories are generated through a Generator–Advisor review loop. The Advisor has access to the full generation context, including the selected paper, the proposal endpoint, and hidden constraints. The Generator sees only the visible query, the current trajectory prefix, the current artifacts, and a set of available tools. It proposes actio","cbCainmY767Thfew","https://ap.wps.com/l/cbCainmY767Thfew","pdf",985798,1,15,"English","en",105,"# Introduction\n## Dataset Overview\n# Related Work","[{\"question\":\"What does IdeaTrail record in each dataset instance?\",\"answer\":\"Each IdeaTrail instance records a multi-turn research trajectory from broad queries through evidence gathering, tool interactions, reasoning turns, artifact updates, and ends at either an idea-level or proposal-level endpoint.\"},{\"question\":\"How does the Generator–Advisor loop produce the trajectories?\",\"answer\":\"The Generator produces visible steps forward from the current trajectory and artifacts, while the Advisor reviews using full generation context to check grounding, causal order, naturalness, and prevents leakage from hidden targets.\"},{\"question\":\"Why is cutoff-aware tool use important for IdeaTrail?\",\"answer\":\"Trajectories are generated under a specified information horizon so the agent cannot use evidence that would be unavailable under the intended data cutoff, reducing temporal leakage and better aligning the process with the target research artifact.\"}]",1784206513,38,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"ideatrail-full-process-agent-trajectories-for-scientific-ideation","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/ideatrail-full-process-agent-trajectories-for-scientific-ideation/85828/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","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 does IdeaTrail record in each dataset instance?","Question",{"text":75,"@type":76},"Each IdeaTrail instance records a multi-turn research trajectory from broad queries through evidence gathering, tool interactions, reasoning turns, artifact updates, and ends at either an idea-level or proposal-level endpoint.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the Generator–Advisor loop produce the trajectories?",{"text":80,"@type":76},"The Generator produces visible steps forward from the current trajectory and artifacts, while the Advisor reviews using full generation context to check grounding, causal order, naturalness, and prevents leakage from hidden targets.",{"name":82,"@type":73,"acceptedAnswer":83},"Why is cutoff-aware tool use important for IdeaTrail?",{"text":84,"@type":76},"Trajectories are generated under a specified information horizon so the agent cannot use evidence that would be unavailable under the intended data cutoff, reducing temporal leakage and better aligning the process with the target research artifact.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]