TBNAI
DataPraxis: Scaling Analytics Tasks via Verified Workspace Evolution
Publication: cs.AI updates on arXiv.org
Author: Nan Jia, Wenxuan Shi, Jingqing Ruan, Zhizhao Zeng, Tianyi Zhai, Zhongda Su, Feng Hong, Zhengyu Chen, Wenhao Yan, Haojin Yang, Siqi Dai, Weipeng Zhang, Ke Zeng, Xunliang Cai
Source published: 21/10/2026, 04:00:00 UTC
Briefing created: 21/10/2026, 04:25:44 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 21/10/2026, 04:03:24 UTC.
Briefings
DataPraxis is a framework for synthesizing analytics tasks using verified workspace evolution, generating a 5,000 trajectory dataset to train large language models. The system constructs tasks from validated findings and their dependencies, achieving performance metrics of 55.11 on DABStep and 51.9 on JobBench, which match or surpass frontier models like GLM-5.2.
Key points
- DataPraxis-5K is a dataset containing 5,000 verified trajectories spanning diverse domains and output formats.
- DataPraxis-27B achieves 55.11 on DABStep and 51.9 on JobBench, matching or surpassing frontier models such as GLM-5.2.
Context
This research addresses the need for diverse, challenging tasks in general-purpose agents by introducing DataPraxis, which uses verified workspace evolution to uncover deeper analytical questions and construct tasks from validated findings.
Limitations
The evidence is limited to the arXiv preprint v1, which may not reflect final experimental results or full-scale deployment capabilities. Performance metrics are reported on specific benchmarks without details on sample sizes or statistical significance.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 21/10/2026, 04:25:44 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeiey7g746q5m5h7umilypfl3uz7yeg2dmryrv6j3qq2brjjrhd4vbm