AI Paper Atlas

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

Student-Centered Distillation Narrows the Agentic Gap Between Small and Large LLMs

Yuanjie Lyu, Chengyu Wang, Jun Huang, Tong Xu

ICML 2026 · Large language models › Agents, code and math · approach: Long-horizon LLM agents · 0 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
replace full-trajectory behavior cloning with SCoRe-SFT on student-generated, first-error-corrected dataQwen2.5-7B-Instruct · 8 reasoning benchmarks averageaverage score: 43.0 -> 46.1Improved
add short-horizon RL with key-step rewards after SCoRe-SFT (ablation)Qwen2.5-7B-Instruct · 8 reasoning benchmarks averageaverage score: 46.1 -> 52.3Improved
use SCoRe-RL instead of BC + GRPOQwen2.5-7B-Instruct · 8 reasoning benchmarks averageaverage score: 48.2 -> 52.3Improved
use SCoRe-RL instead of BC + GRPOQwen3-8B non-thinking · deep search averageaverage score: 22.2 -> 30.5Improved
add short-horizon rollouts from the verified prefix in RL (ablation)SCoRe-RL · 8 reasoning benchmarks averageaverage score: 48.4 -> 52.3Improved
add key-step rewards in RL (ablation)SCoRe-RL · 8 reasoning benchmarks averageaverage score: 49.7 -> 52.3Improved
train on MPS-generated data instead of the cold-start BC data only (ablation)Qwen2.5-7B-Instruct · 8 reasoning benchmarks averageaverage score: 41.7 -> 46.1Improved
add RL after SFT on MPS data (ablation)Hard-to-Teach held-out subset · math and search hard subsetaccuracy: math 17.3% -> 24.3%; search 20.4% -> 26.3%Improved
add GRPO full-horizon RL to behavior cloning on deep searchQwen3-8B non-thinking · deep search averageaverage score: 22.8 -> 22.2No clear effect
set key-step reward scale to 0.5 versus 0.3 or 0.7 (ablation)SCoRe-RL · math benchmarks (2k-sample search)accuracyImproved

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