AI Paper Atlas

Research map / Large language models / Agents, code and math

DESIGNER: Design-Logic-Guided Multidisciplinary Data Synthesis for LLM Reasoning

Weize Liu, Yongchi Zhao, Yijia Luo, Mingyu Xu, Jiaheng Liu, Yanan Li, Xiguo Hu, ZhiqiBai, Yuchi Xu, Wenbo Su, Bo Zheng

ICLR 2026 · Large language models › Agents, code and math · approach: Mathematical and structural reasoning benchmarks · 2 citations (OpenAlex)

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 changedWhereMeasuredEffect
SFT base model on DLR-Web+Book synthetic data vs official instruct modelLlama-3.1-8B · GPQA-Diamondaccuracy: 23.18 -> 65.45Improved
SFT Qwen3-8B-Base on DLR-Web+Book vs official thinking-mode modelQwen3-8B · GPQA-Diamondaccuracy: 59.44 -> 71.01Improved
SFT Qwen3-4B-Base on DLR-Web only vs official thinking modelQwen3-4B · GPQA-Diamondaccuracy: 54.70 -> 53.74Mixed
DLR data vs other synthetic datasets at equal sizeQwen3-8B-Base · GPQA-Diamondaccuracy: 56.67 -> 60.35 (NaturalReasoning vs DLR-Book)Improved
combine Web and Book dataLlama-3.2-3B · GPQA-Diamondaccuracy: 22.42 (Web) / 38.33 (Book) -> 42.27Improved
use Design Logics instead of exemplar questions (ablation)DESIGNER · GPQA-Diamond (304,181 questions)accuracy: 58.89 -> 60.35Improved
use Design Logics instead of exemplar questions on full 3.04M corpus (ablation)DESIGNER · GPQA-Diamondaccuracy: 64.10 -> 69.39Improved
add coarse retrieval ranking (ablation)DESIGNER · MMLU-Proaccuracy: 73.76 -> 74.98Improved
add fine LLM re-ranking (ablation)DESIGNER · GPQA-Diamondaccuracy: 59.34 -> 60.35Improved
synthesize from book corpus instead of web corpus (ablation)DESIGNER · multidisciplinary benchmarksaccuracyImproved
scale synthesized data size (ablation)DESIGNER · multidisciplinary benchmarksaccuracyImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy