AI Paper Atlas

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

Tailoring the Training: Difficulty-Aware Learning Strategy Allocation for Large Language Models

Xiaoling Zhou, Shuaiyu Zhou, Zhemg Lee, Tao Chen, Xirui Li, Peng Chen, Jie Jiang, Wei Ye, Shikun Zhang

ICML 2026 · Large language models › Instruction and fine-tuning · approach: Fine-tuning for reasoning data

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
add SFT as an allocatable strategy (ablation)DALSA · Math + MMLU-Pro overall (6 benchmarks)accuracy: 48.7 -> 54.3Improved
add RL as an allocatable strategy (ablation)DALSA · Math + MMLU-Pro overall (6 benchmarks)accuracy: 50.6 -> 54.3Improved
add DISCARD strategy for mastered samples (ablation)DALSA · Math + MMLU-Pro overall (6 benchmarks)accuracy: 53.8 -> 54.3Improved
add anti-curriculum weighting (ablation)DALSA · Math + MMLU-Pro overall (6 benchmarks)accuracy: 53.7 -> 54.3Improved
add ALS loss instead of standard CE (ablation)DALSA · Math + MMLU-Pro overall (6 benchmarks)accuracy: 53.8 -> 54.3Improved
add anti-curriculum weighting to vanilla RL (ablation)RL (GRPO) · Math + MMLU-Pro overall (6 benchmarks)accuracy: 46.9 -> 48.2Improved
add ALS loss to vanilla SFT (ablation)SFT · Math + MMLU-Pro overall (6 benchmarks)accuracy: 49.0 -> 50.4Improved
per-sample SFT/RL/DISCARD allocation instead of ReLIFT hard-sample SFTReLIFT · Math + MMLU-Pro overall (6 benchmarks)accuracy: 52.6 -> 54.3Improved
DALSA vs standalone RL and SFTRL / SFT · Math + MMLU-Pro overall (6 benchmarks)accuracy: +7.4 and +5.3 percentage pointsImproved
DALSA vs RL on response lengthRL (GRPO) · Math + MMLU-Pro overall (6 benchmarks)response length: 2175 -> 3364Mixed
DALSA vs ReLIFT on Qwen2.5-Math-1.5BReLIFT · Math + MMLU-Pro overall (Qwen2.5-Math-1.5B)accuracy: 36.5 -> 38.2Improved
DALSA vs ReLIFT on Qwen3-8BReLIFT · Math + MMLU-Pro overall (Qwen3-8B)accuracy: 49.5 -> 51.0Improved

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