Research map / Large language models / Agents, code and math
DeepAnalyze: Agentic Large Language Models for Autonomous Data Science
Shaolei Zhang, Ju Fan, Meihao Fan, Yizhe Liu, Yuxin Zhang, Xiaoyong Du
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add the separate <Understand> action (ablation) | DeepAnalyze-8B · DABStep | Accuracy: 31.78 -> 38.88 | Improved |
| add the separate <Understand> action (ablation) | DeepAnalyze-8B · WikiTQ | Exact match: 80.78 -> 83.24 | Improved |
| add the separate <Understand> action (ablation) | DeepAnalyze-8B · DS-1000 | Pass@1: 61.20 -> 61.70 | No clear effect |
| add multi-ability agentic training after single-ability fine-tuning (ablation) | DeepAnalyze-8B · DABStep | Accuracy: 15.34 -> 38.88 | Improved |
| add single-ability fine-tuning before agentic training (ablation) | DeepAnalyze-8B · DABStep | Accuracy: 30.66 -> 38.88 | Improved |
| use two-stage curriculum instead of one-stage mixed training (ablation) | DeepAnalyze-8B · DS-1000 | Pass@1: 54.80 -> 61.70 | Improved |
| use two-stage curriculum instead of one-stage mixed training (ablation) | DeepAnalyze-8B · DABStep | Accuracy: 36.89 -> 38.88 | Improved |
| add distillation of reasoning trajectories (ablation) | Reasoning-Table data, trained LLM · WikiTQ | Exact match: 75.54 -> 78.80 | Improved |
| add keyword-guided refinement on top of distillation (ablation) | Reasoning-Table data, trained LLM · WikiTQ | Exact match: 78.80 -> 80.25 | Improved |
| add keyword-guided refinement on top of distillation (ablation) | Reasoning-Table data, trained LLM · HybridQA | Exact match: 36.12 -> 38.84 | Improved |
| add keyword-guided refinement on top of distillation (ablation) | Reasoning-Table data, trained LLM · HiTab | Exact match: 74.44 -> 75.86 | Improved |
| add agentic training stage on top of single-ability fine-tuning (ablation) | DeepAnalyze-8B · TableQA (7-benchmark average) | Exact match (AVG): 60.64 -> 64.47 | Improved |
Later papers that compared with this method
1 later accepted paper reports DeepAnalyze as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- VeriGraph: Towards Verifiable Data-Analytic Agents (NeurIPS 2026 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DS-Agent: Automated Data Science by Empowering Large Language Models with Case-Based Reasoning (ICML 2024 · similarity 0.89)
- Scaling Generalist Data-Analytic Agents (ICLR 2026 · similarity 0.89)
- Agentic-imodels: Evolving agentic interpretability tools via autoresearch (NeurIPS 2026 · similarity 0.88)
- DSBench: How Far Are Data Science Agents from Becoming Data Science Experts? (ICLR 2025 · similarity 0.87)
- Spider2-V: How Far Are Multimodal Agents From Automating Data Science and Engineering Workflows? (NeurIPS 2024 · similarity 0.86)
- Star-Agents: Automatic Data Optimization with LLM Agents for Instruction Tuning (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
