SynthAPI
A controllable data-generation pipeline for synthesizing training environments for tool-calling agents, enabling explicit control of diversity along three axes: tools, tasks, and databases.
Timeline (last 90 days)
The research proposes the SynthAPI pipeline, which can explicitly control the diversity of generated environments along three axes: tools, tasks, and databases.
The research team trained 849 models of different scales (four sizes in total) on 458,752 tasks to derive scaling laws.
Scaling data according to the research findings, the trained agents outperform models trained on data from prior work on all four evaluation benchmarks.
Scaling laws can not only predict performance but also inversely guide the design of data synthesis pipelines.
All Facts (4)
The research proposes the SynthAPI pipeline, which can explicitly control the diversity of generated environments along three axes: tools, tasks, and databases.
50%UnverifiedThe research team trained 849 models of different scales (four sizes in total) on 458,752 tasks to derive scaling laws.
50%UnverifiedScaling data according to the research findings, the trained agents outperform models trained on data from prior work on all four evaluation benchmarks.
50%UnverifiedScaling laws can not only predict performance but also inversely guide the design of data synthesis pipelines.
50%