Data World Engine Generates RL Training Environments for Data Analysis
Verified-by-construction environments sidestep the usual reward-hacking problem in RL for open-ended analysis tasks, a reusable recipe for teams training agents beyond fixed-answer benchmarks. The diagnostic finding that frontier models compute correctly but conclude wrongly matters for anyone building data-analysis agents who should audit the reasoning step, not just the arithmetic. The paper is available on arXiv (2610.21656) with a Chat with Paper interface at academy.dair.ai.
A new paper introduces an engine that builds data-analysis training environments from scratch. Data World first encodes the mechanisms that generate a dataset, then produces the files, tasks, and grading rubrics, so open-ended findings can be verified and audits confirm each answer is recoverable from the files. RL training of Qwen3.6-35B-A3B on 2,135 such tasks, without distilled trajectories, adds 20.9 points on DABstep-Hard, 14.28 on DAComp, and 12.13 on JobBench. The paper also reports that ten frontier models compute correct numbers more often than they draw correct conclusions from them.
Nice paper on scaling data environments from scratch. They built an engine that creates data-analysis environments from scratch. They apply RL, which adds 20.9 points on DABstep-Hard. Data World first writes down the mechanisms that generate a dataset, then produces the files, tasks, and grading rubrics. Because the generating process is known, open-ended findings can be checked, and independent audits confirm each answer can be recovered from the files. Training Qwen3.6-35B-A3B with RL on 2,135 of these tasks, with no distilled trajectories, also adds 14.28 points on DAComp and 12.13 on JobBench. The benchmark finding is just as useful. Ten frontier models compute the right numbers more often than they draw the right conclusion from them. arxiv.org/abs/2610.21656 Chat with Paper: academy.dair.ai/papers/data-wo… 💬 7 🔄 2 ❤️ 18 👀 1577 📊 9 ⚡
