AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

The Generalization Spectrum: A Chromatographic Approach to Evaluating Learning Algorithms

Jinghan Zhang, Zerui Cheng, Shiqi Chen, Ge Zhang, Wenhao Huang, Jiashuo Liu, Junxian He, Tianle Cai

ICML 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Data augmentation and generalization bias · 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
train with RL (GRPO) instead of reference-solution SFT at matched memorization (ablation)Qwen3-4B-Thinking · Generalization Spectrum, D0≈0.6near-transfer normalized gain (D1): 21.9% -> 39.0%Improved
train with RL (GRPO) instead of SFT at matched memorization (ablation)Qwen3-4B-Thinking · Generalization Spectrum D3/D4far-transfer normalized gain: -7.1%/-0.5% -> 12.3%/16.8%Improved
use paired ICL demonstration instead of random demonstration (ablation)Qwen3-4B-Thinking · Generalization Spectrum D0-D2pass@1: D2 0.21 -> 0.63Improved
paired ICL at category-matched and unpaired levelsQwen3-4B-Thinking · Generalization Spectrum D3/D4pass@1: D3 0.24 -> 0.17; D4 0.22 -> 0.10Hurt
add pseudocode or key insight to ICL code demonstration (ablation)paired ICL · Generalization Spectrum D0-D2pass@1: D2 64.4% -> 71.1% (pseudocode)Improved
add chain-of-thought trace to ICL code demonstration (ablation)paired ICL · Generalization Spectrum D0-D2pass@1: D2 64.4% -> 63.0%No clear effect
add oracle correspondence hint plus pseudocode to ICL (ablation)paired ICL · Generalization Spectrum D2pass@1: 64.4% -> 80.1%Improved
add oracle hint to ICL at category-matched and unpaired levels (ablation)paired ICL · Generalization Spectrum D3/D4pass@1No clear effect
use RFT with self-generated correct targets instead of demonstration SFT (ablation)SFT on Qwen3-4B-Thinking · Generalization Spectrum D2-D4far-transfer pass@1: compile error 17.6% -> 6.8%Improved
use GPT-OSS-20B off-policy targets instead of Qwen3-4B or QwQ targets for SFT (ablation)SFT · Generalization Spectrum D1-D4pass@1: target perplexity 4.29Hurt
use self-distillation fine-tuning (SDFT) instead of vanilla SFT (ablation)SFT at D0≈0.6 · Generalization Spectrumpass@1 at D1-D2 vs D3-D4Mixed
add training-time hints (pseudocode or key idea) to GRPO (ablation)GRPO · Generalization Spectrum, D0≈0.6steps to reach D0≈0.6: step 170 -> step 60 (KI), step 70 (Pseudo)Improved

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