AI Paper Atlas

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

Expanding the Capability Frontier of LLM Agents with ZPD-Guided Data Synthesis

Xuanzhong Chen, Zile Qiao, Guoxin Chen, Liangcai Su, Zhen Zhang, Xinyu Wang, Pengjun Xie, Fei Huang, Jingren Zhou, Yong Jiang, Ting Chen

ICLR 2026 · Large language models › Agents, code and math · approach: Long-horizon LLM agents

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
RFT on AgentFrontier instead of next-best dataset (MiroVerse)Qwen3-8B · HLE text-onlyaccuracy (%): 15.00 -> 18.80Improved
RFT on AgentFrontier instead of next-best dataset (MiroVerse)Qwen3-32B · HLE text-onlyaccuracy (%): 19.92 -> 23.82Improved
RFT on AgentFrontier instead of next-best dataset (MegaScience)Qwen3-30B-A3B-Thinking-2507 · HLE text-onlyaccuracy (%): 20.15 -> 25.67Improved
RFT on any synthetic agent dataset versus tool-augmented base modelQwen3-30B-A3B-Thinking-2507 · HLE text-onlyaccuracy (%): 10.17 -> 25.67 (AgentFrontier)Improved
add continued pre-training on 50B tokens before RFT (ablation)AgentFrontier-30B-A3B · HLE text-onlyaccuracy (%): 25.7 -> 28.6Improved
add continued pre-training before RFT (ablation)AgentFrontier-30B-A3B · xBench-ScienceQAaccuracy (%): 54.0 -> 61.0Improved
add continued pre-training before RFT (ablation)AgentFrontier-30B-A3B · ZPD Exam-v1accuracy (%): 91.4 -> 93.4Improved
add continued pre-training before RFT (ablation)AgentFrontier-30B-A3B · RBench-Taccuracy (%): 74.4 -> 77.1Improved
RFT on AgentFrontier for tool-use accuracyQwen3-30B-A3B · HLE text-onlyoverall conditional tool accuracy (%): 20.5-21.0 -> 26.3Improved
sample 8 attempts instead of 1RFT Qwen3-30B-A3B · AgentFrontier validation (300)accuracy (%): 21.7 -> 40.7Improved
give tools to base modelQwen3-30B-A3B-Thinking-2507 · HLE text-onlyaccuracy (%): 9.2 -> 10.2No clear effect
give tools to base modelGPT-4o · ZPD Exam-v1accuracy (%): 4.8 -> 51.3Improved

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