Google Research·· 21 d ago
ToolGrad: Efficient tool-use dataset generation with textual "gradients"
ToolGrad: Efficient tool-use dataset generation with textual "gradients"
AI summary
At ACL 2026, Google Research proposed ToolGrad, an answer-first, question-later approach that generates real tool-call chains before deriving user queries, replacing traditional DFS trial-and-error annotation.
Selection record
Threshold 60Official, first-handFirst 42Second 42
Not admittedSum of both 84 < twice the threshold 120
- Source tier
- Official, first-hand; this tier's threshold is 60
- Pre-filter
- passed：AI智能体工具使用数据集生成研究
A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.
Source: Google Research · research.google