---
title: "InstructDiff — paper by Su et al. (2026) — summary and ideas"
description: "Fine-tuning LLMs on entire instruction datasets is expensive and yields diminishing returns. 50 references."
authors: ["Junyou Su", "He Zhu", "Xiao Luo", "Liyu Zhang", "Hong-Yu Zhou", "Yun Chen", "Peng Li", "Yang Liu", "Guanhua Chen"]
year: 2026
venue: "Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)"
source: "https://aclanthology.org/2026.acl-long.486/"
doi: "10.18653/v1/2026.acl-long.486"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/su-2026-instructdiff-domain-adaptive-data-selection/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/su-2026-instructdiff-domain-adaptive-data-selection/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/33/) / InstructDiff: Domain-Adaptive Data Selection via Contrastive Entropy for Efficient LLM Fine-Tuning

# InstructDiff: Domain-Adaptive Data Selection via Contrastive Entropy for Efficient LLM Fine-Tuning

**Authors:** Junyou Su, [He Zhu](https://md.research.elamur.ai/free/en/authors/he-zhu/), [Xiao Luo](https://md.research.elamur.ai/free/en/authors/xiao-luo/), Liyu Zhang, Hong-Yu Zhou, [Yun Chen](https://md.research.elamur.ai/free/en/authors/yun-chen/), [Peng Li](https://md.research.elamur.ai/free/en/authors/peng-li/), [Yang Liu](https://md.research.elamur.ai/free/en/authors/yang-liu/) and [Guanhua Chen](https://md.research.elamur.ai/free/en/authors/guanhua-chen/) · **Year:** 2026 · **Venue:** [Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2026](https://md.research.elamur.ai/free/en/venues/acl/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.acl-long.486/) · [PDF](https://aclanthology.org/2026.acl-long.486.pdf) · [DOI 10.18653/v1/2026.acl-long.486](https://doi.org/10.18653/v1/2026.acl-long.486) · [OpenAlex](https://openalex.org/W7166841392)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
Fine-tuning LLMs on entire instruction datasets is expensive and yields diminishing returns. Existing data-selection techniques also tend to work only in the domain they were designed for: methods tuned for general instruction-following transfer poorly to reasoning, and vice versa. InstructDiff addresses this with a single, domain-adaptive criterion. The base model is first lightly fine-tuned on a small random subset, producing a calibration model. For each candidate example, the method then compares the two models on two signals: the change in length-normalized NLL (ΔNLL) and contrastive entropy (ΔH = H_base − H_inst). Samples with extreme ΔNLL in either direction are discarded as redundant or unlearnable. From the remaining pool, the examples with the lowest ΔH are kept.

The same rule behaves differently by domain. On math, it favors samples where tuning raises entropy ("cognitive expansion"). On general and medical data, it favors samples with a small entropy drop ("cognitive compression"). Using just 10% of the data, InstructDiff beats full-data training and prior baselines, so no hand-crafted domain heuristics are needed.

## Key points
- Reported gains relative to full-data training with a 10% subset: about 17% on mathematical reasoning and 52% on general instruction-following. Experiments cover general, medical, math, and code domains. Code is released at github.com/zhuchichi56/Instruct-diff.
- Weak-to-strong calibration works reasonably well. Using Qwen2.5-0.5B to select data for Qwen2.5-7B gives an average of 28.4, compared with 25.80 for random selection and 31.63 for 7B self-calibration. It also cuts warmup compute by up to roughly 14×.
- Iterative re-selection helps mainly in the second round. Math scores go from 31.63 to 32.89 at T=2, and only to 33.24 at T=3. Runs with T=1–2 take 39–64% of full-data training time.
- Selection depends on model family. Same-family models (e.g., Qwen2.5-7B/14B, or LLaMA3-8B and LLaMA3.1-8B) agree on 60–64% of chosen samples, while cross-family pairs agree on only 15–20%. When training Qwen2.5-7B, Qwen2.5-14B calibration reaches 30.84, versus 25.74–26.31 with LLaMA calibrators.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Tsinghua University](https://md.research.elamur.ai/free/en/universities/tsinghua-university/) — University, Beijing, China · 333 works · 87,019 citations
- [Peking University](https://md.research.elamur.ai/free/en/universities/peking-university/) — University, Beijing, China · 244 works · 39,359 citations
- [University of Wisconsin–Madison](https://md.research.elamur.ai/free/en/universities/university-of-wisconsin-madison/) — University, Madison, United States · 42 works · 48,486 citations
- [Southern University of Science and Technology](https://md.research.elamur.ai/free/en/universities/southern-university-of-science-and-technology/) — University, Shenzhen, China · 19 works · 1,337 citations
- [Shanghai University of Finance and Economics](https://md.research.elamur.ai/free/en/universities/shanghai-university-of-finance-and-economics/) — University, Shanghai, China · 13 works · 825 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (1): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,320 works

## Cite

Cite as: Junyou Su, He Zhu, Xiao Luo, Liyu Zhang, Hong-Yu Zhou, Yun Chen, Peng Li, Yang Liu and Guanhua Chen. 2026. InstructDiff: Domain-Adaptive Data Selection via Contrastive Entropy for Efficient LLM Fine-Tuning. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{su2026instructdiff,
  title = {{InstructDiff: Domain-Adaptive Data Selection via Contrastive Entropy for Efficient LLM Fine-Tuning}},
  author = {Junyou Su, He Zhu, Xiao Luo, Liyu Zhang, Hong-Yu Zhou, Yun Chen, Peng Li, Yang Liu and Guanhua Chen},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.486},
  url = {https://doi.org/10.18653/v1/2026.acl-long.486}
}
```

- [← previous Insights from Transfer Learning Experiments with Word-in-Context and Word Sense Disambiguation Models](https://md.research.elamur.ai/free/en/papers/mujko-2026-insights-transfer-learning-experiments-word/)
- [next → InstructSum: A Benchmark to Evaluate Instruction-Following Capability of Large Language Models in Summarization](https://md.research.elamur.ai/free/en/papers/nishida-2026-instructsum-benchmark-evaluate-instruction-following/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/33/)

## References

50 references, 21 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Anthropic. 2025. Introducing claude 4.5 sonnet. [https://www.anthropic.com/news/](https://www.anthropic.com/news/) claude-sonnet-4-5. [↗](https://www.anthropic.com/news/)
2. Jacob Austin, Augustus Odena, Maxwell Nye, Maarten Bosma, Henryk Michalewski, David Dohan, Ellen Jiang, Carrie Cai, Michael Terry, Quoc Le, and 1 others. 2021. [Program synthesis with large language models](https://md.research.elamur.ai/free/en/papers/austin-2021-program-synthesis-large-language-models/). arXiv preprint arXiv:2108.07732.
3. Jan Betley, Daniel Tan, Niels Warncke, Anna Sztyber-Betley, Xuchan Bao, Martı́n Soto, Nathan Labenz, and Owain Evans. 2025. Emergent misalignment: Narrow finetuning can produce broadly misaligned llms. arXiv preprint arXiv:2502.17424. — [arXiv:2502.17424](https://arxiv.org/abs/2502.17424) [find](https://scholar.google.com/scholar?q=Jan+Betley%2C+Daniel+Tan%2C+Niels+Warncke%2C+Anna+Sztyber%2DBetley%2C+Xuchan+Bao%2C+Mart%C4%B1%CC%81n+Soto%2C+Nathan+Labenz%2C+and+Owain+Evans%2E+2025%2E+Emergent+misalignment%3A+Narrow+finetuning+can+produce+broadly+misaligned+llms)
4. Federico Cassano, John Gouwar, Francesca Lucchetti, Claire Schlesinger, Anders Freeman, Carolyn Jane Anderson, Molly Q Feldman, Michael Greenberg, Abhinav Jangda, and Arjun Guha. 2024. Knowledge transfer from high-resource to low-resource programming languages for code llms. Proceedings of the ACM o… [find](https://scholar.google.com/scholar?q=Federico+Cassano%2C+John+Gouwar%2C+Francesca+Lucchetti%2C+Claire+Schlesinger%2C+Anders+Freeman%2C+Carolyn+Jane+Anderson%2C+Molly+Q+Feldman%2C+Michael+Greenberg%2C+Abhinav+Jangda%2C+and+Arjun+Guha%2E+2024%2E+Knowledge+trans)
5. Lichang Chen, Shiyang Li, Jun Yan, Hai Wang, Kalpa Gunaratna, Vikas Yadav, Zheng Tang, Vijay Srinivasan, Tianyi Zhou, Heng Huang, and 1 others. 2023. [Alpagasus: Training a better alpaca with fewer data](https://md.research.elamur.ai/free/en/papers/chen-2023-alpagasus-training-better-alpaca-fewer/). arXiv preprint arXiv:2307.08701.
6. Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde de Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, and 1 others. 2021. [Evaluating large language models trained on code](https://md.research.elamur.ai/free/en/papers/chen-2021-evaluating-large-language-models-trained/). Preprint, arXiv:2107.03374.
7. Daixuan Cheng, Shaohan Huang, Xuekai Zhu, Bo Dai, Wayne Xin Zhao, Zhenliang Zhang, and Furu Wei. 2025. Reasoning with exploration: An entropy perspective. arXiv preprint arXiv:2506.14758. [find](https://scholar.google.com/scholar?q=Daixuan+Cheng%2C+Shaohan+Huang%2C+Xuekai+Zhu%2C+Bo+Dai%2C+Wayne+Xin+Zhao%2C+Zhenliang+Zhang%2C+and+Furu+Wei%2E+2025%2E+Reasoning+with+exploration%3A+An+entropy+perspective%2E+arXiv+preprint+arXiv%3A2506%2E14758%2E)
8. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, and 1 others. 2024. [The llama 3 herd of models](https://md.research.elamur.ai/free/en/papers/grattafiori-2024-llama-3-herd-models/). arXiv preprint arXiv:2407.21783.
9. Sebastian Farquhar, Jannik Kossen, Lorenz Kuhn, and Yarin Gal. 2024. Detecting hallucinations in large language models using semantic entropy. Nature, 630(8017):625–630. [find](https://scholar.google.com/scholar?q=Sebastian+Farquhar%2C+Jannik+Kossen%2C+Lorenz+Kuhn%2C+and+Yarin+Gal%2E+2024%2E+Detecting+hallucinations+in+large+language+models+using+semantic+entropy%2E+Nature%2C+630%288017%29%3A625%E2%80%93630%2E)
10. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Peiyi Wang, Qihao Zhu, Runxin Xu, Ruoyu Zhang, Shirong Ma, Xiao Bi, and 1 others. 2025. Deepseekr1 incentivizes reasoning in llms through reinforcement learning. Nature, 645(8081):633–638. [find](https://scholar.google.com/scholar?q=Daya+Guo%2C+Dejian+Yang%2C+Haowei+Zhang%2C+Junxiao+Song%2C+Peiyi+Wang%2C+Qihao+Zhu%2C+Runxin+Xu%2C+Ruoyu+Zhang%2C+Shirong+Ma%2C+Xiao+Bi%2C+and+1+others%2E+2025%2E+Deepseekr1+incentivizes+reasoning+in+llms+through+reinforceme)
11. Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2020. [Measuring massive multitask language understanding](https://md.research.elamur.ai/free/en/papers/hendrycks-2020-measuring-massive-multitask-language-understanding/). arXiv preprint arXiv:2009.03300.
12. Liang, and Tatsunori B. Hashimoto. 2023b. Alpacaeval: An automatic evaluator of instruction-following models. [https://github.com/](https://github.com/) tatsu-lab/alpaca_eval. [↗](https://github.com/)
13. Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. 2021. [Measuring mathematical problem solving with the math dataset](https://md.research.elamur.ai/free/en/papers/hendrycks-2021-measuring-mathematical-problem-solving-math/). In Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2).
14. Yunshui Li, Binyuan Hui, Xiaobo Xia, Jiaxi Yang, Min Yang, Lei Zhang, Shuzheng Si, Ling-Hao Chen, Junhao Liu, Tongliang Liu, and 1 others. 2024c. [One-shot learning as instruction data prospector for large language models](https://md.research.elamur.ai/free/en/papers/li-2024-one-shot-learning-as-instruction/). In Proceedings of the 62nd Annual Meeting of the Association for Computational…
15. Di Jin, Eileen Pan, Nassim Oufattole, Wei-Hung Weng, Hanyi Fang, and Peter Szolovits. 2021. [What disease does this patient have? a large-scale open domain question answering dataset from medical exams](https://md.research.elamur.ai/free/en/papers/jin-2020-disease-this-patient-have-large/). Applied Sciences, 11(14):6421.
16. James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, and 1 others. 2017. Overcoming catastrophic forgetting in neural networks. Proceedings of the national academy of sciences, 114(1… [find](https://scholar.google.com/scholar?q=James+Kirkpatrick%2C+Razvan+Pascanu%2C+Neil+Rabinowitz%2C+Joel+Veness%2C+Guillaume+Desjardins%2C+Andrei+A+Rusu%2C+Kieran+Milan%2C+John+Quan%2C+Tiago+Ramalho%2C+Agnieszka+Grabska%2DBarwinska%2C+and+1+others%2E+2017%2E+Overcomin)
17. Aitor Lewkowycz, Anders Andreassen, David Dohan, Ethan Dyer, Henryk Michalewski, Vinay Ramasesh, Ambrose Slone, Cem Anil, Imanol Schlag, Theo Gutman-Solo, and 1 others. 2022. [Solving quantitative reasoning problems with language models](https://md.research.elamur.ai/free/en/papers/lewkowycz-2022-solving-quantitative-reasoning-problems-language/). Advances in neural information processing systems, 35:3843–3857…
18. Jia LI, Edward Beeching, Lewis Tunstall, Ben Lipkin, Roman Soletskyi, Shengyi Costa Huang, Kashif Rasul, Longhui Yu, Albert Jiang, Ziju Shen, Zihan Qin, Bin Dong, Li Zhou, Yann Fleureau, Guillaume Lample, and Stanislas Polu. 2024. Numinamath. [https://huggingface.co/](https://huggingface.co/) datasets/AI-MO/NuminaMath-CoT. [↗](https://huggingface.co/)
19. Ming Li, Yong Zhang, Shwai He, Zhitao Li, Hongyu Zhao, Jianzong Wang, Ning Cheng, and Tianyi Zhou. 2024a. [Superfiltering: Weak-to-strong data filtering for fast instruction-tuning](https://md.research.elamur.ai/free/en/papers/li-2024-superfiltering-weak-strong-data-filtering/). arXiv preprint arXiv:2402.00530.
20. Ming Li, Yong Zhang, Zhitao Li, Jiuhai Chen, Lichang Chen, Ning Cheng, Jianzong Wang, Tianyi Zhou, and Jing Xiao. 2024b. [From quantity to quality: Boosting llm performance with self-guided data selection for instruction tuning](https://md.research.elamur.ai/free/en/papers/li-2023-quantity-quality-boosting-llm-performance/). In Proceedings of the 2024 Conference of the North American Chapter of t…
21. Xian Li, Ping Yu, Chunting Zhou, Timo Schick, Omer Levy, Luke Zettlemoyer, Jason Weston, and Mike Lewis. 2023a. [Self-alignment with instruction back-translation](https://md.research.elamur.ai/free/en/papers/li-2023-self-alignment-instruction-backtranslation/). arXiv preprint arXiv:2308.06259.
22. Xuechen Li, Tianyi Zhang, Yann Dubois, Rohan Taori, Ishaan Gulrajani, Carlos Guestrin, Percy Hunter Lightman, Vineet Kosaraju, Yuri Burda, Harrison Edwards, Bowen Baker, Teddy Lee, Jan Leike, John Schulman, Ilya Sutskever, and Karl Cobbe. 2023. [Let’s verify step by step](https://md.research.elamur.ai/free/en/papers/lightman-2023-let-s-verify-step-step/). In The Twelfth International…
23. Liangxin Liu, Xuebo Liu, Derek F Wong, Dongfang Li, Ziyi Wang, Baotian Hu, and Min Zhang. 2024. Selectit: Selective instruction tuning for llms via uncertainty-aware self-reflection. Advances in Neural Information Processing Systems, 37:97800– 97825. [find](https://scholar.google.com/scholar?q=Liangxin+Liu%2C+Xuebo+Liu%2C+Derek+F+Wong%2C+Dongfang+Li%2C+Ziyi+Wang%2C+Baotian+Hu%2C+and+Min+Zhang%2E+2024%2E+Selectit%3A+Selective+instruction+tuning+for+llms+via+uncertainty%2Daware+self%2Dreflection%2E+Advances+in+Neura)
24. Wei Liu, Weihao Zeng, Keqing He, Yong Jiang, and Junxian He. 2023. [What makes good data for alignment? a comprehensive study of automatic data selection in instruction tuning](https://md.research.elamur.ai/free/en/papers/liu-2023-makes-good-data-alignment-comprehensive/). arXiv preprint arXiv:2312.15685.
25. Zihan Liu, Yang Chen, Mohammad Shoeybi, Bryan Catanzaro, and Wei Ping. 2025. Acemath: Advancing frontier math reasoning with post-training and reward modeling. In Findings of the Association for Computational Linguistics: ACL 2025, pages 3993– 4015. [find](https://scholar.google.com/scholar?q=Zihan+Liu%2C+Yang+Chen%2C+Mohammad+Shoeybi%2C+Bryan+Catanzaro%2C+and+Wei+Ping%2E+2025%2E+Acemath%3A+Advancing+frontier+math+reasoning+with+post%2Dtraining+and+reward+modeling%2E+In+Findings+of+the+Association+for+Compu)
26. Keming Lu, Hongyi Yuan, Zheng Yuan, Runji Lin, Junyang Lin, Chuanqi Tan, Chang Zhou, and Jingren Zhou. 2023. Instag: Instruction tagging for analyzing supervised fine-tuning of large language models. Preprint, arXiv:2308.07074. [find](https://scholar.google.com/scholar?q=Keming+Lu%2C+Hongyi+Yuan%2C+Zheng+Yuan%2C+Runji+Lin%2C+Junyang+Lin%2C+Chuanqi+Tan%2C+Chang+Zhou%2C+and+Jingren+Zhou%2E+2023%2E+Instag%3A+Instruction+tagging+for+analyzing+supervised+fine%2Dtuning+of+large+language+models%2E+)
27. Zhicun Lyu, Xinye Li, Zheng Xie, and Ming Li. 2025. Top pass: improve code generation by pass@ k-maximized code ranking. Frontiers of Computer Science, 19(8):198341. [find](https://scholar.google.com/scholar?q=Zhicun+Lyu%2C+Xinye+Li%2C+Zheng+Xie%2C+and+Ming+Li%2E+2025%2E+Top+pass%3A+improve+code+generation+by+pass%40+k%2Dmaximized+code+ranking%2E+Frontiers+of+Computer+Science%2C+19%288%29%3A198341%2E)
28. Mathematical Association of America. 2023. 2023 competition problems. [find](https://scholar.google.com/scholar?q=Mathematical+Association+of+America%2E+2023%2E+2023+competition+problems%2E)
29. Amc Andrew Mutton, Mark Dras, Stephen Wan, and Robert Dale. 2007. GLEU: Automatic evaluation of sentence-level fluency. In Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics, pages 344–351, Prague, Czech Republic. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Amc+Andrew+Mutton%2C+Mark+Dras%2C+Stephen+Wan%2C+and+Robert+Dale%2E+2007%2E+GLEU%3A+Automatic+evaluation+of+sentence%2Dlevel+fluency%2E+In+Proceedings+of+the+45th+Annual+Meeting+of+the+Association+of+Computational+Li)
30. OpenAI. 2025. Openai o3 and o4-mini system card, january 2025. [https://openai.com/index/](https://openai.com/index/) o3-o4-mini-system-card. [↗](https://openai.com/index/)
31. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, and 1 others. 2022. [Training language models to follow instructions with human feedback](https://md.research.elamur.ai/free/en/papers/ouyang-2022-training-language-models-follow-instructions/). Advances in neural information processing systems, 35:27730– 27744.
32. Ankit Pal, Logesh Kumar Umapathi, and Malaikannan Sankarasubbu. 2022. [Medmcqa: A large-scale multi-subject multi-choice dataset for medical domain question answering](https://md.research.elamur.ai/free/en/papers/pal-2022-medmcqa-large-scale-multi-subject/). In Conference on health, inference, and learning, pages 248–260. PMLR.
33. Qwen, :, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, and 1 others. 2025. Qwen2.5 technical report. Preprint, arXiv:2412.15115. [find](https://scholar.google.com/scholar?q=Qwen%2C+%3A%2C+An+Yang%2C+Baosong+Yang%2C+Beichen+Zhang%2C+Binyuan+Hui%2C+Bo+Zheng%2C+Bowen+Yu%2C+Chengyuan+Li%2C+Dayiheng+Liu%2C+Fei+Huang%2C+and+1+others%2E+2025%2E+Qwen2%2E5+technical+report%2E+Preprint%2C+arXiv%3A2412%2E15115%2E)
34. Abdullah Vanlioglu. 2025. Entropy-guided sequence weighting for efficient exploration in rl-based llm fine-tuning. arXiv preprint arXiv:2503.22456. — [arXiv:2503.22456](https://arxiv.org/abs/2503.22456) [find](https://scholar.google.com/scholar?q=Abdullah+Vanlioglu%2E+2025%2E+Entropy%2Dguided+sequence+weighting+for+efficient+exploration+in+rl%2Dbased+llm+fine%2Dtuning%2E+arXiv+preprint+arXiv%3A2503%2E22456%2E+%E2%80%94+%5BarXiv%3A2503%2E22456%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2503%2E2245)
35. Christian Walder and Deep Karkhanis. 2025. Pass@ k policy optimization: Solving harder reinforcement learning problems. arXiv preprint arXiv:2505.15201. — [arXiv:2505.15201](https://arxiv.org/abs/2505.15201) [find](https://scholar.google.com/scholar?q=Christian+Walder+and+Deep+Karkhanis%2E+2025%2E+Pass%40+k+policy+optimization%3A+Solving+harder+reinforcement+learning+problems%2E+arXiv+preprint+arXiv%3A2505%2E15201%2E+%E2%80%94+%5BarXiv%3A2505%2E15201%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2505)
36. Shenzhi Wang, Le Yu, Chang Gao, Chujie Zheng, Shixuan Liu, Rui Lu, Kai Dang, Xionghui Chen, Jianxin Yang, Zhenru Zhang, and 1 others. 2025. [Beyond the 80/20 rule: High-entropy minority tokens drive effective reinforcement learning for llm reasoning](https://md.research.elamur.ai/free/en/papers/wang-2025-80-20-rule-high-entropy/). arXiv preprint arXiv:2506.01939.
37. Yejie Wang, Keqing He, Dayuan Fu, Zhuoma Gongque, Heyang Xu, Yanxu Chen, Zhexu Wang, Yujia Fu, Guanting Dong, Muxi Diao, Jingang Wang, Mengdi Zhang, Xunliang Cai, and Weiran Xu. 2024. How do your code llms perform? empowering code instruction tuning with high-quality data. Preprint, arXiv:2409.03810… — [arXiv:2409.03810](https://arxiv.org/abs/2409.03810) [find](https://scholar.google.com/scholar?q=Yejie+Wang%2C+Keqing+He%2C+Dayuan+Fu%2C+Zhuoma+Gongque%2C+Heyang+Xu%2C+Yanxu+Chen%2C+Zhexu+Wang%2C+Yujia+Fu%2C+Guanting+Dong%2C+Muxi+Diao%2C+Jingang+Wang%2C+Mengdi+Zhang%2C+Xunliang+Cai%2C+and+Weiran+Xu%2E+2024%2E+How+do+your+code)
38. Yizhong Wang, Yeganeh Kordi, Swaroop Mishra, Alisa Liu, Noah A. Smith, Daniel Khashabi, and Hannaneh Hajishirzi. 2023. Self-instruct: Aligning language models with self-generated instructions. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long… [find](https://scholar.google.com/scholar?q=Yizhong+Wang%2C+Yeganeh+Kordi%2C+Swaroop+Mishra%2C+Alisa+Liu%2C+Noah+A%2E+Smith%2C+Daniel+Khashabi%2C+and+Hannaneh+Hajishirzi%2E+2023%2E+Self%2Dinstruct%3A+Aligning+language+models+with+self%2Dgenerated+instructions%2E+In+Proc)
39. Alexander Wettig, Aatmik Gupta, Saumya Malik, and Danqi Chen. 2024. Qurating: Selecting high-quality data for training language models. arXiv preprint arXiv:2402.09739. [find](https://scholar.google.com/scholar?q=Alexander+Wettig%2C+Aatmik+Gupta%2C+Saumya+Malik%2C+and+Danqi+Chen%2E+2024%2E+Qurating%3A+Selecting+high%2Dquality+data+for+training+language+models%2E+arXiv+preprint+arXiv%3A2402%2E09739%2E)
40. Mengzhou Xia, Sadhika Malladi, Suchin Gururangan, Sanjeev Arora, and Danqi Chen. 2024a. [Less: Selecting influential data for targeted instruction tuning](https://md.research.elamur.ai/free/en/papers/xia-2024-less-selecting-influential-data-targeted/). arXiv preprint arXiv:2402.04333.
41. Tingyu Xia, Bowen Yu, Kai Dang, An Yang, Yuan Wu, Yuan Tian, Yi Chang, and Junyang Lin. 2024b. Rethinking data selection at scale: Random selection is almost all you need. Preprint, arXiv:2410.09335. [find](https://scholar.google.com/scholar?q=Tingyu+Xia%2C+Bowen+Yu%2C+Kai+Dang%2C+An+Yang%2C+Yuan+Wu%2C+Yuan+Tian%2C+Yi+Chang%2C+and+Junyang+Lin%2E+2024b%2E+Rethinking+data+selection+at+scale%3A+Random+selection+is+almost+all+you+need%2E+Preprint%2C+arXiv%3A2410%2E09335%2E)
42. Yang Xiao, Mohan Jiang, Jie Sun, Keyu Li, Jifan Lin, Yumin Zhuang, Ji Zeng, Shijie Xia, Qishuo Hua, Xuefeng Li, and 1 others. 2025. Limi: Less is more for agency. arXiv preprint arXiv:2509.17567. — [arXiv:2509.17567](https://arxiv.org/abs/2509.17567) [find](https://scholar.google.com/scholar?q=Yang+Xiao%2C+Mohan+Jiang%2C+Jie+Sun%2C+Keyu+Li%2C+Jifan+Lin%2C+Yumin+Zhuang%2C+Ji+Zeng%2C+Shijie+Xia%2C+Qishuo+Hua%2C+Xuefeng+Li%2C+and+1+others%2E+2025%2E+Limi%3A+Less+is+more+for+agency%2E+arXiv+preprint+arXiv%3A2509%2E17567%2E+%E2%80%94+%5Ba)
43. Zhangchen Xu, Fengqing Jiang, Luyao Niu, Yuntian Deng, Radha Poovendran, Yejin Choi, and Bill Yuchen Lin. 2024. [Magpie: Alignment data synthesis from scratch by prompting aligned llms with nothing](https://md.research.elamur.ai/free/en/papers/xu-2024-magpie-alignment-data-synthesis-scratch/). arXiv preprint arXiv:2406.08464.
44. Yixin Ye, Zhen Huang, Yang Xiao, Ethan Chern, Shijie Xia, and Pengfei Liu. 2025. [Limo: Less is more for reasoning](https://md.research.elamur.ai/free/en/papers/ye-2025-limo-less-more-reasoning/). arXiv preprint arXiv:2502.03387.
45. Mingjia Yin, Chuhan Wu, Yufei Wang, Hao Wang, Wei Guo, Yasheng Wang, Yong Liu, Ruiming Tang, Defu Lian, and Enhong Chen. 2024. Entropy law: The story behind data compression and llm performance. arXiv preprint arXiv:2407.06645. [find](https://scholar.google.com/scholar?q=Mingjia+Yin%2C+Chuhan+Wu%2C+Yufei+Wang%2C+Hao+Wang%2C+Wei+Guo%2C+Yasheng+Wang%2C+Yong+Liu%2C+Ruiming+Tang%2C+Defu+Lian%2C+and+Enhong+Chen%2E+2024%2E+Entropy+law%3A+The+story+behind+data+compression+and+llm+performance%2E+arXiv)
46. Chunting Zhou, Pengfei Liu, Puxin Xu, Srinivasan Iyer, Jianfeng Sun, Yuning Mao, Xuezhe Ma, Avia Efrat, Ping Yu, Lili Yu, and 1 others. 2023. Lima: Less is more for alignment. In Advances in Neural Information Processing Systems, volume 36. [find](https://scholar.google.com/scholar?q=Chunting+Zhou%2C+Pengfei+Liu%2C+Puxin+Xu%2C+Srinivasan+Iyer%2C+Jianfeng+Sun%2C+Yuning+Mao%2C+Xuezhe+Ma%2C+Avia+Efrat%2C+Ping+Yu%2C+Lili+Yu%2C+and+1+others%2E+2023%2E+Lima%3A+Less+is+more+for+alignment%2E+In+Advances+in+Neural+In)
47. He Zhu, Zhiwen Ruan, Junyou Su, Xingwei He, Yun Chen, Wenjia Zhang, and Guanhua Chen. 2025a. Tag-instruct: Controlled instruction complexity enhancement through structure-based augmentation. Preprint, arXiv:2505.18557. [find](https://scholar.google.com/scholar?q=He+Zhu%2C+Zhiwen+Ruan%2C+Junyou+Su%2C+Xingwei+He%2C+Yun+Chen%2C+Wenjia+Zhang%2C+and+Guanhua+Chen%2E+2025a%2E+Tag%2Dinstruct%3A+Controlled+instruction+complexity+enhancement+through+structure%2Dbased+augmentation%2E+Preprint%2C)
48. He Zhu, Junyou Su, Tianle Lun, Yicheng Tao, Wenjia Zhang, Zipei Fan, and Guanhua Chen. 2024. Fanno: Augmenting high-quality instruction data with open-sourced llms only. Preprint, arXiv:2408.01323. [find](https://scholar.google.com/scholar?q=He+Zhu%2C+Junyou+Su%2C+Tianle+Lun%2C+Yicheng+Tao%2C+Wenjia+Zhang%2C+Zipei+Fan%2C+and+Guanhua+Chen%2E+2024%2E+Fanno%3A+Augmenting+high%2Dquality+instruction+data+with+open%2Dsourced+llms+only%2E+Preprint%2C+arXiv%3A2408%2E01323%2E)
49. Yongfu Zhu, Lin Sun, Guangxiang Zhao, Weihong Lin, and Xiangzheng Zhang. 2025b. Uncertainty under the curve: A sequence-level entropy area metric for reasoning llm. arXiv preprint arXiv:2508.20384. — [arXiv:2508.20384](https://arxiv.org/abs/2508.20384) [find](https://scholar.google.com/scholar?q=Yongfu+Zhu%2C+Lin+Sun%2C+Guangxiang+Zhao%2C+Weihong+Lin%2C+and+Xiangzheng+Zhang%2E+2025b%2E+Uncertainty+under+the+curve%3A+A+sequence%2Dlevel+entropy+area+metric+for+reasoning+llm%2E+arXiv+preprint+arXiv%3A2508%2E20384%2E+%E2%80%94+)
50. Terry Yue Zhuo, Minh Chien Vu, Jenny Chim, Han Hu, Wenhao Yu, Ratnadira Widyasari, Imam Nur Bani Yusuf, Haolan Zhan, Junda He, Indraneil Paul, and 1 others. 2024. [Bigcodebench: Benchmarking code generation with diverse function calls and complex instructions](https://md.research.elamur.ai/free/en/papers/zhuo-2024-bigcodebench-benchmarking-code-generation-diverse/). arXiv preprint arXiv:2406.15877.

*Updated 1 November 2026*