Reproduce “Scaling Laws for Tool-Use Agents via Controllable Environment Synthesis” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Andrew Szot, Samira Abnar, Omar Attia, Abbas Kazerouni, Michael Kirchhof, Cheng-Yu Hsieh, Raviteja Vemulapalli, Alexander Toshev
SynthAPI lets researchers test how tool, task, and database-state diversity affect training for agents that use stateful tools. It varies apps, tasks per app, starting database profiles, and distractor data across training budgets. On held-out high-compute runs, its composition-aware scaling law reaches a 0.0285 MAE, versus 0.1706 for task count alone, with extrapolation up to 16×; validation covers one synthesis pipeline.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.