AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

STAT: Skill-Targeted Adaptive Training

Yinghui He, Abhishek Panigrahi, Yong Lin, Sanjeev Arora

ICLR 2026 · Large language models › Instruction and fine-tuning · approach: Efficient LLM fine-tuning systems

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 embedding-similarity selection with skill-profile selection (STAT-Sel vs Embed-Sel) (ablation)Llama-3.2-3B-Instruct SFT · MATH testaccuracy: 46.0 -> 51.5Improved
Add skill-targeted STAT-Sel data instead of plain MATH-Train SFTLlama-3.2-3B-Instruct · MATH testaccuracy: 44.8 -> 51.5Improved
Naive SFT on MATH-TrainQwen2.5-3B · MATH testaccuracy: 55.8 -> 50.0Hurt
Restrict SFT to Level 4-5 questions (MATH-Hard) (ablation)Llama-3.2-3B-Instruct · MATH testaccuracy: 44.8 -> 45.6No clear effect
Add synthesized skill-targeted questions (STAT-Syn vs STAT-Sel) (ablation)Llama-3.2-3B-Instruct SFT · MATH^D (difficult subset)accuracy: 26.6 -> 31.7Improved
Add synthetic data, STAT-Syn vs STAT-Sel on MATH (ablation)Llama-3.2-3B-Instruct SFT · MATH testaccuracy: 51.5 -> 50.2No clear effect
Apply STAT-Sel before GRPOLlama-3.2-3B-Instruct · Average over 9 math benchmarksaverage accuracy: 31.8 -> 40.7Improved
Apply STAT-Syn before GRPOQwen2.5-3B · Average over 9 math benchmarksaverage accuracy: 46.0 -> 48.4Improved
Add GRPO on top of STAT-Syn SFTLlama-3.2-3B-Instruct · Average over 9 math benchmarksaverage accuracy: 37.2 -> 41.3Improved
Continual STAT with skill profile from MATH-perturb-hardLlama-3.2-3B-Instruct STAT models · MATH-perturb-hardaccuracy: 3-4% gainImproved
Use embedding-based selection/synthesis instead of skill-based (ablation)Llama-3.2-3B-Instruct · MATH^2accuracy: 21.9 -> 19.5Hurt

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