Research map / Large language models / Agents, code and math
Hunt Instead of Wait: Evaluating Deep Data Research on Large Language Models
Wei Liu, Peijie Yu, Michele Orini, Yali Du, Yulan He
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 |
|---|---|---|---|
| give explicit goal-directed queries instead of query-free exploration (ablation) | Qwen3-Next-80B-A3B · DDR-Bench average | accuracy [%]: 32.59 -> 43.21 | Improved |
| give explicit goal-directed queries (ablation) | Qwen3-Next-80B-A3B · DDR-Bench GLOBEM | accuracy [%]: 35.40 -> 31.95 | Hurt |
| raise reasoning budget (ablation) | Qwen3-Next-80B-A3B · DDR-Bench 10-K (message-wise) | accuracy [%]: 45.58 -> 36.40 | Hurt |
| raise reasoning budget (ablation) | Qwen3-Next-80B-A3B · DDR-Bench 10-K (trajectory-wise) | accuracy [%]: 31.10 -> 37.34 | Improved |
| raise reasoning budget (ablation) | Qwen3-Next-80B-A3B · DDR-Bench | interactive turns (10-K): 27.93 -> 11.89 | Mixed |
| add long-short-term memory mechanism (ablation) | Qwen3-Next-80B-A3B · DDR-Bench 10-K (message-wise) | accuracy [%]: 45.58 -> 37.34 | Hurt |
| add memory mechanism (ablation) | Qwen3-Next-80B-A3B · DDR-Bench GLOBEM | accuracy [%]: 35.40 -> 35.86 | Improved |
| add memory mechanism (ablation) | Qwen3-Next-80B-A3B · DDR-Bench MIMIC (trajectory-wise) | accuracy [%]: 20.80 -> 14.34 | Hurt |
| novelty ranking vs checklist accuracy ranking | DDR-Bench · 10-K | rank correlation: 0.94 | Improved |
| increase interaction turns | frontier models · DDR-Bench | accuracy | Improved |
Closest papers by meaning
- Large Language Models Think Too Fast To Explore Effectively (NeurIPS 2025 · similarity 0.89)
- From Intent to Evidence: A Categorical Approach for Structural Evaluation of Deep Research Agents (NeurIPS 2026 · similarity 0.88)
- InnovatorBench: Evaluating Agents’ Ability to Conduct Innovative AI Research (ICLR 2026 · similarity 0.88)
- WideSearch: Benchmarking Agentic Broad Info-Seeking (ICLR 2026 · similarity 0.88)
- \$OneMillion-Bench: How Far are Language Agents from Human Experts? (NeurIPS 2026 · similarity 0.87)
- Agentic-imodels: Evolving agentic interpretability tools via autoresearch (NeurIPS 2026 · similarity 0.87)
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.
