Research map / Large language models / Instruction and fine-tuning
Cure-SFT: Diagnostic-Guided Data Curation for Instruction Tuning
Yuankang Fu, Xinrong Gong, Chen Gong, Tong Zhang, Kaixiang Yang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add stratified semantic-geometric sampling (Stage I) (ablation) | Cure-SFT · Alpaca 10% | average winning score: 1.72 -> 1.78 | Improved |
| add diagnostic triage (Stage II) and remediation on top of sampling (ablation) | Cure-SFT · Alpaca 10% | average winning score: 1.05 -> 1.19 | Improved |
| add targeted remediation of fixable samples (Stage III) (ablation) | Cure-SFT · Alpaca 10% | average winning score: 1.19 -> 1.78 | Improved |
| random sampling instead of semantic-geometric sampling (ablation) | Cure-SFT · Alpaca 10%, Vicuna | winning score: 1.93 vs 1.91 | No clear effect |
| targeted remediation instead of unified rewriting | Cure-SFT vs Rewrite · Alpaca 10%, AlpacaEval 2.0 LC win | LC win rate: 12.98 -> 11.45 | Hurt |
| targeted remediation instead of unified rewriting | Cure-SFT vs Rewrite · Alpaca 10% | average winning score: 1.54 -> 1.78 | Improved |
| use selection-based methods (IFD, Deita) on higher-quality data | IFD, Deita · WizardLM 10% | average winning score: IFD 0.67, Deita 0.62 vs Random 0.86 | Hurt |
| increase data budget from 10% to 15% | Cure-SFT · Alpaca, WizardLM test | winning score: 1.85 -> 1.78 | No clear effect |
| use 1% data budget | Cure-SFT · Alpaca, WizardLM test | winning score: 1.60 at 1% | Improved |
| use smaller teacher Qwen2.5-14B instead of 72B (ablation) | Cure-SFT · Alpaca 10% | winning score: 1.48 to 1.90 vs 72B slightly higher | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- T-SHIRT — T-SHIRT: Token-Selective Hierarchical Data Selection for Instruction Tuning (NeurIPS 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SHED: Shapley-Based Automated Dataset Refinement for Instruction Fine-Tuning (NeurIPS 2024 · similarity 0.88)
- What Makes Good Data for Alignment? A Comprehensive Study of Automatic Data Selection in Instruction Tuning (ICLR 2024 · similarity 0.87)
- MAmmoTH2: Scaling Instructions from the Web (NeurIPS 2024 · similarity 0.87)
- Improving Data Efficiency via Curating LLM-Driven Rating Systems (ICLR 2025 · similarity 0.86)
- Adapt-$\infty$: Scalable Continual Multimodal Instruction Tuning via Dynamic Data Selection (ICLR 2025 · similarity 0.86)
- T-SHIRT: Token-Selective Hierarchical Data Selection for Instruction Tuning (NeurIPS 2025 · similarity 0.86)
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.
