Lune

EMNLP Findings2024

Data Diversity Matters for Robust Instruction Tuning

Alexander Bukharin, Shiyang Li, Zhengyang Wang, Jingfeng Yang, Bing Yin, Xian Li, Chao Zhang, Tuo Zhao, Haoming Jiang

Abstract

Recent works have shown that by curating high quality and diverse instruction tuning datasets, we can significantly improve instruction-following capabilities. However, creating such datasets is difficult and most works rely on manual curation or proprietary language models. Automatic data curation is difficult as it is still not clear how we can define diversity for instruction tuning, how diversity and quality depend on one other, and how we can optimize dataset quality and diversity. To resolve these issue, we propose a new algorithm, Quality-Diversity Instruction Tuning (QDIT). QDIT provides a simple method to simultaneously control dataset diversity and quality, allowing us to conduct an in-depth study on the effect of diversity and quality on instruction tuning performance. From this study we draw two key insights (1) there is a natural tradeoff between data diversity and quality and (2) increasing data diversity significantly improves the worst case instruction following performance, therefore improving robustness. We validate the performance of QDIT on several large scale instruction tuning datasets, where we find it can substantially improve worst and average case performance compared to qualitydriven data selection.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

37 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • Measuring Massive Multitask Language Understanding cites Language Models are Few-Shot Learners
  • Prismatic Synthesis: Gradient-based Data Diversification Boosts Generalization in LLM Reasoning cites Measuring Massive Multitask Language Understanding
  • Prismatic Synthesis: Gradient-based Data Diversification Boosts Generalization in LLM Reasoning cites Deberta: decoding-Enhanced Bert with Disentangled Attention
  • Prismatic Synthesis: Gradient-based Data Diversification Boosts Generalization in LLM Reasoning cites LESS: Selecting Influential Data for Targeted Instruction Tuning
  • Deberta: decoding-Enhanced Bert with Disentangled Attention cites Language Models are Few-Shot Learners
  • Superfiltering: Weak-to-Strong Data Filtering for Fast Instruction-Tuning cites Measuring Massive Multitask Language Understanding
  • Superfiltering: Weak-to-Strong Data Filtering for Fast Instruction-Tuning cites Finetuned Language Models are Zero-Shot Learners
  • TSDS: Data Selection for Task-Specific Model Finetuning cites Measuring Massive Multitask Language Understanding
  • LESS: Selecting Influential Data for Targeted Instruction Tuning cites Measuring Massive Multitask Language Understanding
  • LESS: Selecting Influential Data for Targeted Instruction Tuning cites Finetuned Language Models are Zero-Shot Learners
  • LESS: Selecting Influential Data for Targeted Instruction Tuning cites Training language models to follow instructions with human feedback
  • LEAD: Iterative Data Selection for Efficient LLM Instruction Tuning cites Measuring Massive Multitask Language Understanding
  • LEAD: Iterative Data Selection for Efficient LLM Instruction Tuning cites Superfiltering: Weak-to-Strong Data Filtering for Fast Instruction-Tuning
  • LEAD: Iterative Data Selection for Efficient LLM Instruction Tuning cites LESS: Selecting Influential Data for Targeted Instruction Tuning
  • Finetuned Language Models are Zero-Shot Learners cites Language Models are Few-Shot Learners
  • Finetuned Language Models are Zero-Shot Learners cites Training language models to follow instructions with human feedback
  • Training language models to follow instructions with human feedback cites Finetuned Language Models are Zero-Shot Learners
  • Training language models to follow instructions with human feedback cites Language Models are Few-Shot Learners
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers37

  1. LESS: Selecting Influential Data for Targeted Instruction TuningMengzhou Xia, Sadhika Malladi, Suchin Gururangan, Sanjeev Arora et al.ICML 2024 · 460 citations
  2. Prismatic Synthesis: Gradient-based Data Diversification Boosts Generalization in LLM ReasoningJaehun Jung, Seungju Han, Ximing Lu, Skyler Hallinan et al.NeurIPS 2025 · 50 citations
  3. TSDS: Data Selection for Task-Specific Model FinetuningZifan Liu, Amin Karbasi, Theodoros RekatsinasNeurIPS 2024 · 35 citations
  4. Superfiltering: Weak-to-Strong Data Filtering for Fast Instruction-TuningMing Li, Yong Zhang, Shwai He, Zhitao Li et al.ACL 2024 · 16 citations
  5. LEAD: Iterative Data Selection for Efficient LLM Instruction TuningXiaotian Lin, Yanlin Qi, Yizhang Zhu, Themis Palpanas et al.VLDB 2026 · 16 citations

Builds on20

  1. Language Models are Few-Shot LearnersTom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah et al.NeurIPS 2020 · 64,255 citations
  2. Training language models to follow instructions with human feedbackLong Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida et al.NeurIPS 2022 · 24,707 citations
  3. Measuring Massive Multitask Language UnderstandingDan Hendrycks, Collin Burns, Steven Basart, Andy Zou et al.ICLR 2021 · 7,905 citations
  4. Finetuned Language Models are Zero-Shot LearnersJason Wei, Maarten Bosma, Vincent Y. Zhao, Kelvin Guu et al.ICLR 2022 · 4,966 citations
  5. Deberta: decoding-Enhanced Bert with Disentangled AttentionPengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu ChenICLR 2021 · 3,729 citations

Related papers

  1. From Selection to Refinement: Iterative Optimization for Instruction DataHang Hu, Ziyan Liu, Rujie Wen, Ruihui Hou et al.ACL 2026
  2. MIG: Automatic Data Selection for Instruction Tuning by Maximizing Information Gain in Semantic SpaceYicheng Chen, Yining Li, Kai Hu, Zerun Ma et al.ACL Findings 2025 · 2 citations
  3. SelectIT: Selective Instruction Tuning for LLMs via Uncertainty-Aware Self-ReflectionLiangxin Liu, Xuebo Liu, Derek F. Wong, Dongfang Li et al.NeurIPS 2024 · 49 citations
  4. Cure-SFT: Diagnostic-Guided Data Curation for Instruction TuningYuankang Fu, Xinrong Gong, Chen Gong, Tong Zhang et al.ICML 2026
  5. Learning Instructions with Unlabeled Data for Zero-Shot Cross-Task GeneralizationYuxian Gu, Pei Ke, Xiaoyan Zhu, Minlie HuangEMNLP 2022 · 3 citations
Dusk over the sea between two cliffs drawn in fine vertical lines