ToolSpace: Tool-centric synthesis of long-horizon Agent environments
Start here
On November 3, 2026, Hugging Face Daily Papers included the paper ToolSpace, which proposes a tool-first framework: mine execution traces from real MCP services and open agent corpora, use deterministic rules to extract data, state and retry dependencies from adjacent calls, cluster a weighted tool-combination graph into environments, then generate tasks via state-aware random walks with rollback and decompile them into natural-language tasks with assertions, retaining only samples that can be successfully replayed from the initial state, for synthesizing a multi-environment tool-use benchmark and training data.
Generated by AI from reports · Updated 15 days ago
Attention and developments
Attention, hourly
Now 32, 0% in 24 hours·Peak 33 (Nov 15 08:00)
Storyline
- · 1 reports· Attention32
ToolSpace: Tool-centric synthesis of long-horizon Agent environmentsThis event
- · 1 reports· Attention18Renmin University of China, Fudan University and others propose the ToolSpace agent environment synthesis framework
Progress timeline
- XiaohongshuXiaohongshu · Wang Shuo (SJTU)· DiscussionToolSpace: How does AI learn to get things done across tools?(opens the original post in a new tab)
- Hugging Face Daily PapersToolSpace: Tool-Centric Environment Synthesis Scaling for Long-Horizon Agents
The paper proposes ToolSpace, a tool-first framework that synthesizes multi-environment tool-use benchmarks and training data by mining execution traces from real MCP services and open agent corpora. In contrast to data-centric approaches that first synthesize databases and then derive tools, it mines data, state, and retry dependencies from adjacent calls using deterministic rules, clusters a weighted tool-combination graph into environments, and generates tasks via state-aware random walks with rollback, decompiling them into natural-language tasks with assertion checks and retaining only samples that can be successfully replayed from the initial state. On AutomationBench and tau3-Bench Banking, this pipeline generates harder Pro versions, with 7 frontier models dropping by up to 38.2 points. On seven long-horizon tool-use benchmarks, ToolSpace raises the average score of Qwen3.6-35B-A3B from 23.0 to 41.0 (43.2 after RL) and GLM-5.2 from 44.6 to 56.6, comparable to Claude Opus 5.
Related discussion
- Older than 72 hours · Not included in the 72-hour figures
- XiaohongshuXiaohongshu · Wang Shuo (SJTU)ToolSpace: How does AI learn to get things done across tools?(opens the original post in a new tab)
Attention trend
New sources per day
- Nov 15: 1 new sources (0 outlets, 1 accounts)
- Nov 16: 0 new sources (0 outlets, 0 accounts)
- Nov 17: 0 new sources (0 outlets, 0 accounts)
- Nov 18: 0 new sources (0 outlets, 0 accounts)
- Nov 19: 0 new sources (0 outlets, 0 accounts)
- Nov 20: 0 new sources (0 outlets, 0 accounts)