Fudan Releases Sci Computing Agent Env
jiqizhixin · x · 2026-07-10
The Fudan University NLP team introduced SciAgentGym, an interactive environment for scientific reasoning featuring 1,780 domain-specific tools. The team also built the SciAgentBench benchmark, finding that top models perform poorly on long-horizon complex tasks.
To address this, researchers proposed the SciForge method, which treats tool operations as a dependency graph to generate smarter training data. The resulting SciAgent-8B model outperformed the much larger Qwen3-VL-235B-Instruct and demonstrated strong cross-domain tool-use generalization.
More from coding & agent
- A better path to agent autonomy is running waves, finding friction, and iterating — JnBrymn · 2026-07-22
- Coding agents are heading toward an AI-writes, AI-reviews, human-approves workflow — aftahi_ai · 2026-07-22
- oMLX 0.5.2 adds Mac menu-bar stats, low-bit decode kernels, and faster downloads — awnihannun · 2026-07-22
- GitHub review bot hits its PR limit and forces a 39-minute cooldown — DanielLockyer · 2026-07-22
- Max reasoning effort appears to be mobile-only in Codex Remote, not desktop — GabGarrett · 2026-07-22
- A Reddit demo argues online stores should expose carts and pricing through MCP — gelembjuk · 2026-07-22