---
title: "GALA — paper by Xie et al. (2026) — summary and ideas"
description: "GALA replaces this loop with a three-stage curation pipeline built into training: Samples are embedded, clustered with… 76 references."
authors: ["Zhongwei Xie", "Ruihao Liao", "Zimo Wang", "Chong Chen", "Xian-Sheng Hua", "Xiao Luo"]
year: 2026
venue: "Findings of the Association for Computational Linguistics: ACL 2026"
source: "https://aclanthology.org/2026.findings-acl.500/"
doi: "10.18653/v1/2026.findings-acl.500"
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/xie-2026-gala-geometric-data-selection-strategic/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/xie-2026-gala-geometric-data-selection-strategic/)

[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/34/) / GALA: Geometric Data Selection with Strategic Prospecting for Large Language Model Self-training

# GALA: Geometric Data Selection with Strategic Prospecting for Large Language Model Self-training

**Authors:** Zhongwei Xie, Ruihao Liao, Zimo Wang, [Chong Chen](https://md.research.elamur.ai/free/en/authors/chong-chen/), Xian-Sheng Hua and [Xiao Luo](https://md.research.elamur.ai/free/en/authors/xiao-luo/) · **Year:** 2026 · **Venue:** [Findings of the Association for Computational Linguistics: ACL 2026](https://md.research.elamur.ai/free/en/venues/findings/) ([Findings of the Association for Computational Linguistics 2026](https://md.research.elamur.ai/free/en/venues/findings/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.findings-acl.500/) · [PDF](https://aclanthology.org/2026.findings-acl.500.pdf) · [DOI 10.18653/v1/2026.findings-acl.500](https://doi.org/10.18653/v1/2026.findings-acl.500) · [OpenAlex](https://openalex.org/W7166854411)

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

## Summary
Self-training lets an LLM improve by learning from its own solutions. The standard rejection-sampling loop samples many reasoning paths and keeps those that reach the gold answer. This has three weaknesses: it spends compute on near-duplicate solutions, it tends to collapse onto a few dominant solving patterns, and its offline filtering can go stale as the model changes.

GALA replaces this loop with a three-stage curation pipeline built into training:
1. Samples are embedded, clustered with K-Means, and one anchor close to each centroid is kept. This cheaply approximates maximizing embedding-space diversity.
2. For each anchor, the model brainstorms alternative solution strategies and refines them through reflection to add instructional depth.
3. A lightweight check scores the generated paths and admits a mini-batch only if its average quality passes a threshold.

The authors test GALA on math-oriented benchmarks with LLaMA models under both SFT and DPO. It reaches accuracy comparable to full-data training while using far fewer samples. This suggests that smarter curation can make LLM self-improvement substantially cheaper.

## Key points
- Redundancy is measured as the mean pairwise cosine similarity of embeddings. Diversity is measured as a ridge-stabilized log-determinant of the embedding covariance.
- Selection uses all-mpnet-base embeddings and K-Means anchors. Via the Law of Total Variance, this approximates max-volume selection without the O(n³) cost of DPPs.
- Experiments use LLaMA-3.1-8B-Instruct and LLaMA-3.2-1B-Instruct on GSM8K, MATH, ASDIV and MMLU-PRO. Baselines are Full Data, Random Sampling, Uncertainty Sampling and NeMo Curator, each paired with SFT or DPO.
- On Llama-3.1-8B synthetic data, GALA has the lowest Self-BLEU (lower means more diverse) on every set, compared with naive QA and paraphrase augmentation:

  | Dataset | Naive QA | Paraphrase | GALA |
  |---|---|---|---|
  | MATH | 0.72 | 0.74 | 0.67 |
  | GSM8K | 0.77 | 0.82 | 0.74 |
  | MMLU-PRO-MATH | 0.69 | 0.73 | 0.63 |
- Limitations noted by the authors: brainstorming and mini-batch validation can both introduce noise. They propose reward-model guidance as future work, along with applications to medical and geographic foundation models.

*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

- [Hong Kong University of Science and Technology](https://md.research.elamur.ai/free/en/universities/hong-kong-university-of-science-and-technology/) — University, Hong Kong, China · 170 works · 26,341 citations
- [University of Wisconsin–Madison](https://md.research.elamur.ai/free/en/universities/university-of-wisconsin-madison/) — University, Madison, United States · 54 works · 49,089 citations
- [Tongji University](https://md.research.elamur.ai/free/en/universities/tongji-university/) — University, Shanghai, China · 33 works · 2,993 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works

🧭 Topics (1): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,894 works

## Cite

Cite as: Zhongwei Xie, Ruihao Liao, Zimo Wang, Chong Chen, Xian-Sheng Hua and Xiao Luo. 2026. GALA: Geometric Data Selection with Strategic Prospecting for Large Language Model Self-training. Findings of the Association for Computational Linguistics: ACL 2026.

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

```bibtex
@inproceedings{xie2026gala,
  title = {{GALA: Geometric Data Selection with Strategic Prospecting for Large Language Model Self-training}},
  author = {Zhongwei Xie, Ruihao Liao, Zimo Wang, Chong Chen, Xian-Sheng Hua and Xiao Luo},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.500},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.500}
}
```

- [← previous G-LoRA: Global-Local Decoupled Low-Rank Adaptation](https://md.research.elamur.ai/free/en/papers/xiong-2026-g-lora-global-local-decoupled/)
- [next → GaLa: Hypergraph-Guided Visual Language Models for Procedural Planning](https://md.research.elamur.ai/free/en/papers/wang-2026-gala-hypergraph-guided-visual-language/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/34/)

## References

76 references, 22 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Lakshya A Agrawal, Shangyin Tan, Dilara Soylu, Noah Ziems, Rishi Khare, Krista Opsahl-Ong, Arnav Singhvi, Herumb Shandilya, Michael J Ryan, Meng Jiang, et al. 2025. Gepa: Reflective prompt evolution can outperform reinforcement learning. arXiv preprint arXiv:2507.19457. [find](https://scholar.google.com/scholar?q=Lakshya+A+Agrawal%2C+Shangyin+Tan%2C+Dilara+Soylu%2C+Noah+Ziems%2C+Rishi+Khare%2C+Krista+Opsahl%2DOng%2C+Arnav+Singhvi%2C+Herumb+Shandilya%2C+Michael+J+Ryan%2C+Meng+Jiang%2C+et+al%2E+2025%2E+Gepa%3A+Reflective+prompt+evolution+c)
2. Zeyuan Allen-Zhu and Yuanzhi Li. 2023. [Physics of language models: Part 3.1, knowledge storage and extraction](https://md.research.elamur.ai/free/en/papers/allen-zhu-2023-physics-language-models-part-3/). arXiv preprint arXiv:2309.14316.
3. Chitta Baral, Michael Gelfond, and Nelson Rushton. 2009. Probabilistic reasoning with answer sets. Theory and Practice of Logic Programming, 9(1):57– 144. [find](https://scholar.google.com/scholar?q=Chitta+Baral%2C+Michael+Gelfond%2C+and+Nelson+Rushton%2E+2009%2E+Probabilistic+reasoning+with+answer+sets%2E+Theory+and+Practice+of+Logic+Programming%2C+9%281%29%3A57%E2%80%93+144%2E)
4. Hung-Fu Chang and Tong Li. 2025. A framework for collaborating a large language model tool in brainstorming for triggering creative thoughts. Thinking Skills and Creativity, 56:101755. [find](https://scholar.google.com/scholar?q=Hung%2DFu+Chang+and+Tong+Li%2E+2025%2E+A+framework+for+collaborating+a+large+language+model+tool+in+brainstorming+for+triggering+creative+thoughts%2E+Thinking+Skills+and+Creativity%2C+56%3A101755%2E)
5. Nuo Chen, Yicheng Tong, Jiaying Wu, Minh Duc Duong, Qian Wang, Qingyun Zou, Bryan Hooi, and Bingsheng He. 2025. Beyond brainstorming: What drives high-quality scientific ideas? lessons from multi-agent collaboration. arXiv preprint arXiv:2508.04575. — [arXiv:2508.04575](https://arxiv.org/abs/2508.04575) [find](https://scholar.google.com/scholar?q=Nuo+Chen%2C+Yicheng+Tong%2C+Jiaying+Wu%2C+Minh+Duc+Duong%2C+Qian+Wang%2C+Qingyun+Zou%2C+Bryan+Hooi%2C+and+Bingsheng+He%2E+2025%2E+Beyond+brainstorming%3A+What+drives+high%2Dquality+scientific+ideas%3F+lessons+from+multi%2Dagen)
6. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. 2021. [Training verifiers to solve math word problems](https://md.research.elamur.ai/free/en/papers/cobbe-2021-training-verifiers-solve-math-word/). arXiv preprint arXiv:2110.14168.
7. Jairo Diaz-Rodriguez. 2025. k-llmmeans: scalable, stable, and interpretable text clustering via llm-based centroids. arXiv preprint arXiv:2502.09667. — [arXiv:2502.09667](https://arxiv.org/abs/2502.09667) [find](https://scholar.google.com/scholar?q=Jairo+Diaz%2DRodriguez%2E+2025%2E+k%2Dllmmeans%3A+scalable%2C+stable%2C+and+interpretable+text+clustering+via+llm%2Dbased+centroids%2E+arXiv+preprint+arXiv%3A2502%2E09667%2E+%E2%80%94+%5BarXiv%3A2502%2E09667%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2502%2E09)
8. Mucong Ding, Chenghao Deng, Jocelyn Choo, Zichu Wu, Aakriti Agrawal, Avi Schwarzschild, Tianyi Zhou, Tom Goldstein, John Langford, Animashree Anandkumar, et al. 2024. Easy2hard-bench: Standardized difficulty labels for profiling llm performance and generalization. Advances in Neural Information Proc… [find](https://scholar.google.com/scholar?q=Mucong+Ding%2C+Chenghao+Deng%2C+Jocelyn+Choo%2C+Zichu+Wu%2C+Aakriti+Agrawal%2C+Avi+Schwarzschild%2C+Tianyi+Zhou%2C+Tom+Goldstein%2C+John+Langford%2C+Animashree+Anandkumar%2C+et+al%2E+2024%2E+Easy2hard%2Dbench%3A+Standardized+dif)
9. Jiawei Du, Juncheng Hu, Wenxin Huang, Joey Tianyi Zhou, et al. 2024. Diversity-driven synthesis: Enhancing dataset distillation through directed weight adjustment. Advances in neural information processing systems, 37:119443–119465. [find](https://scholar.google.com/scholar?q=Jiawei+Du%2C+Juncheng+Hu%2C+Wenxin+Huang%2C+Joey+Tianyi+Zhou%2C+et+al%2E+2024%2E+Diversity%2Ddriven+synthesis%3A+Enhancing+dataset+distillation+through+directed+weight+adjustment%2E+Advances+in+neural+information+proce)
10. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. 2024. [The llama 3 herd of models](https://md.research.elamur.ai/free/en/papers/grattafiori-2024-llama-3-herd-models/). arXiv e-prints, pages arXiv–2407.
11. André Freitas and Edward Curry. 2016. Big data curation. In New horizons for a data-driven economy: A roadmap for usage and exploitation of big data in Europe, pages 87–118. Springer International Publishing Cham. [find](https://scholar.google.com/scholar?q=Andr%C3%A9+Freitas+and+Edward+Curry%2E+2016%2E+Big+data+curation%2E+In+New+horizons+for+a+data%2Ddriven+economy%3A+A+roadmap+for+usage+and+exploitation+of+big+data+in+Europe%2C+pages+87%E2%80%93118%2E+Springer+International+Pub)
12. 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/). NeurIPS.
13. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Liang Wang, Weizhu Chen, et al. 2022. [Lora: Low-rank adaptation of large language models](https://md.research.elamur.ai/free/en/papers/hu-2021-lora-low-rank-adaptation-large/). Iclr, 1(2):3.
14. Hamish Ivison, Yizhong Wang, Jiacheng Liu, Zeqiu Wu, Valentina Pyatkin, Nathan Lambert, Noah A Smith, Yejin Choi, and Hanna Hajishirzi. 2024. [Unpacking dpo and ppo: Disentangling best practices for learning from preference feedback](https://md.research.elamur.ai/free/en/papers/ivison-2024-unpacking-dpo-ppo-disentangling-best/). Advances in neural information processing systems, 37:36602–36633.
15. Joseph Jennings, Mostofa Patwary, Sandeep Subramanian, Shrimai Prabhumoye, Ayush Dattagupta, Vibhu Jawa, Jiwei Liu, Ryan Wolf, Sarah Yurick, and Varun Singh. 2024. NeMo-Curator: a toolkit for data curation. [find](https://scholar.google.com/scholar?q=Joseph+Jennings%2C+Mostofa+Patwary%2C+Sandeep+Subramanian%2C+Shrimai+Prabhumoye%2C+Ayush+Dattagupta%2C+Vibhu+Jawa%2C+Jiwei+Liu%2C+Ryan+Wolf%2C+Sarah+Yurick%2C+and+Varun+Singh%2E+2024%2E+NeMo%2DCurator%3A+a+toolkit+for+data+cur)
16. Alex Kulesza and Ben Taskar. 2012. Determinantal point processes for machine learning. Foundations and Trends® in Machine Learning, 5(2-3):123–286. [find](https://scholar.google.com/scholar?q=Alex+Kulesza+and+Ben+Taskar%2E+2012%2E+Determinantal+point+processes+for+machine+learning%2E+Foundations+and+Trends%C2%AE+in+Machine+Learning%2C+5%282%2D3%29%3A123%E2%80%93286%2E)
17. Komal Kumar, Tajamul Ashraf, Omkar Thawakar, Rao Muhammad Anwer, Hisham Cholakkal, Mubarak Shah, Ming-Hsuan Yang, Phillip HS Torr, Fahad Shahbaz Khan, and Salman Khan. 2025. Llm post-training: A deep dive into reasoning large language models. arXiv preprint arXiv:2502.21321. — [arXiv:2502.21321](https://arxiv.org/abs/2502.21321) [find](https://scholar.google.com/scholar?q=Komal+Kumar%2C+Tajamul+Ashraf%2C+Omkar+Thawakar%2C+Rao+Muhammad+Anwer%2C+Hisham+Cholakkal%2C+Mubarak+Shah%2C+Ming%2DHsuan+Yang%2C+Phillip+HS+Torr%2C+Fahad+Shahbaz+Khan%2C+and+Salman+Khan%2E+2025%2E+Llm+post%2Dtraining%3A+A+deep+)
18. Nathan Lambert, Jacob Morrison, Valentina Pyatkin, Shengyi Huang, Hamish Ivison, Faeze Brahman, Lester James V Miranda, Alisa Liu, Nouha Dziri, Shane Lyu, et al. 2024. [Tulu 3: Pushing frontiers in open language model post-training](https://md.research.elamur.ai/free/en/papers/lambert-2024-tulu-3-pushing-frontiers-open/). arXiv preprint arXiv:2411.15124.
19. Patrick Lewis, Yuxiang Wu, Linqing Liu, Pasquale Minervini, Heinrich Küttler, Aleksandra Piktus, Pontus Stenetorp, and Sebastian Riedel. 2021. [Paq: 65 million probably-asked questions and what you can do with them](https://md.research.elamur.ai/free/en/papers/lewis-2021-paq-65-million-probably-asked/). Transactions of the Association for Computational Linguistics, 9:1098–1115.
20. Rui Li, Deji Fu, Chunyu Shi, Zhilan Huang, and Gang Lu. 2024. Efficient llms training and inference: An introduction. IEEE Access. [find](https://scholar.google.com/scholar?q=Rui+Li%2C+Deji+Fu%2C+Chunyu+Shi%2C+Zhilan+Huang%2C+and+Gang+Lu%2E+2024%2E+Efficient+llms+training+and+inference%3A+An+introduction%2E+IEEE+Access%2E)
21. Xuefeng Li, Haoyang Zou, and Pengfei Liu. 2025. Limr: Less is more for rl scaling. arXiv preprint arXiv:2502.11886. [find](https://scholar.google.com/scholar?q=Xuefeng+Li%2C+Haoyang+Zou%2C+and+Pengfei+Liu%2E+2025%2E+Limr%3A+Less+is+more+for+rl+scaling%2E+arXiv+preprint+arXiv%3A2502%2E11886%2E)
22. Yisen Li, Lingfeng Yang, Wenxuan Shen, Pan Zhou, Yao Wan, Weiwei Lin, and Dongping Chen. 2026. Crowdselect: Syntheticinstruction data selection with multi-llm wisdom. In Findings of the Association for Computational Linguistics: EACL 2026, pages 1542– 1569. [find](https://scholar.google.com/scholar?q=Yisen+Li%2C+Lingfeng+Yang%2C+Wenxuan+Shen%2C+Pan+Zhou%2C+Yao+Wan%2C+Weiwei+Lin%2C+and+Dongping+Chen%2E+2026%2E+Crowdselect%3A+Syntheticinstruction+data+selection+with+multi%2Dllm+wisdom%2E+In+Findings+of+the+Association+fo)
23. Jessy Lin, Vincent-Pierre Berges, Xilun Chen, Wen-Tau Yih, Gargi Ghosh, and Barlas Oğuz. 2025a. Learning facts at scale with active reading. arXiv preprint arXiv:2508.09494. — [arXiv:2508.09494](https://arxiv.org/abs/2508.09494) [find](https://scholar.google.com/scholar?q=Jessy+Lin%2C+Vincent%2DPierre+Berges%2C+Xilun+Chen%2C+Wen%2DTau+Yih%2C+Gargi+Ghosh%2C+and+Barlas+Og%CC%86uz%2E+2025a%2E+Learning+facts+at+scale+with+active+reading%2E+arXiv+preprint+arXiv%3A2508%2E09494%2E+%E2%80%94+%5BarXiv%3A2508%2E09494%5D%28http)
24. Xiaotian Lin, Yanlin Qi, Yizhang Zhu, Themis Palpanas, Chengliang Chai, Nan Tang, and Yuyu Luo. 2025b. Lead: iterative data selection for efficient llm instruction tuning. Proceedings of the VLDB Endowment, 19(3):426–439. [find](https://scholar.google.com/scholar?q=Xiaotian+Lin%2C+Yanlin+Qi%2C+Yizhang+Zhu%2C+Themis+Palpanas%2C+Chengliang+Chai%2C+Nan+Tang%2C+and+Yuyu+Luo%2E+2025b%2E+Lead%3A+iterative+data+selection+for+efficient+llm+instruction+tuning%2E+Proceedings+of+the+VLDB+Endo)
25. Shang Liu and Xiaocheng Li. 2023. Understanding uncertainty sampling. arXiv preprint arXiv:2307.02719. — [arXiv:2307.02719](https://arxiv.org/abs/2307.02719) [find](https://scholar.google.com/scholar?q=Shang+Liu+and+Xiaocheng+Li%2E+2023%2E+Understanding+uncertainty+sampling%2E+arXiv+preprint+arXiv%3A2307%2E02719%2E+%E2%80%94+%5BarXiv%3A2307%2E02719%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2307%2E02719%29)
26. Zijun Liu, Yanzhe Zhang, Peng Li, Yang Liu, and Diyi Yang. 2023. [Dynamic llm-agent network: An llm-agent collaboration framework with agent team optimization](https://md.research.elamur.ai/free/en/papers/liu-2023-dynamic-llm-powered-agent-network/). arXiv preprint arXiv:2310.02170.
27. Keer Lu, Xiaonan Nie, Zheng Liang, Da Pan, Shusen Zhang, Keshi Zhao, Weipeng Chen, Zenan Zhou, Guosheng Dong, Bin Cui, et al. 2024. Datasculpt: Crafting data landscapes for long-context llms through multi-objective partitioning. arXiv preprint arXiv:2409.00997. — [arXiv:2409.00997](https://arxiv.org/abs/2409.00997) [find](https://scholar.google.com/scholar?q=Keer+Lu%2C+Xiaonan+Nie%2C+Zheng+Liang%2C+Da+Pan%2C+Shusen+Zhang%2C+Keshi+Zhao%2C+Weipeng+Chen%2C+Zenan+Zhou%2C+Guosheng+Dong%2C+Bin+Cui%2C+et+al%2E+2024%2E+Datasculpt%3A+Crafting+data+landscapes+for+long%2Dcontext+llms+through+m)
28. Junyu Luo, Xiao Luo, Xiusi Chen, Zhiping Xiao, Wei Ju, and Ming Zhang. 2024. Semi-supervised fine-tuning for large language models. arXiv preprint arXiv:2410.14745. [find](https://scholar.google.com/scholar?q=Junyu+Luo%2C+Xiao+Luo%2C+Xiusi+Chen%2C+Zhiping+Xiao%2C+Wei+Ju%2C+and+Ming+Zhang%2E+2024%2E+Semi%2Dsupervised+fine%2Dtuning+for+large+language+models%2E+arXiv+preprint+arXiv%3A2410%2E14745%2E)
29. Junyu Luo, Xiao Luo, Xiusi Chen, Zhiping Xiao, Wei Ju, and Ming Zhang. 2025. Semi-supervised fine-tuning for large language models. In Findings of the Association for Computational Linguistics: NAACL 2025, pages 2795–2808. [find](https://scholar.google.com/scholar?q=Junyu+Luo%2C+Xiao+Luo%2C+Xiusi+Chen%2C+Zhiping+Xiao%2C+Wei+Ju%2C+and+Ming+Zhang%2E+2025%2E+Semi%2Dsupervised+fine%2Dtuning+for+large+language+models%2E+In+Findings+of+the+Association+for+Computational+Linguistics%3A+NAACL+)
30. Xiao Luo, Wei Ju, Yiyang Gu, Yifang Qin, Siyu Yi, Daqing Wu, Luchen Liu, and Ming Zhang. 2023. Toward effective semi-supervised node classification with hybrid curriculum pseudo-labeling. ACM Transactions on Multimedia Computing, Communications and Applications, 20(3):1–19. [find](https://scholar.google.com/scholar?q=Xiao+Luo%2C+Wei+Ju%2C+Yiyang+Gu%2C+Yifang+Qin%2C+Siyu+Yi%2C+Daqing+Wu%2C+Luchen+Liu%2C+and+Ming+Zhang%2E+2023%2E+Toward+effective+semi%2Dsupervised+node+classification+with+hybrid+curriculum+pseudo%2Dlabeling%2E+ACM+Transact)
31. Trung Quoc Luong, Xinbo Zhang, Zhanming Jie, Peng Sun, Xiaoran Jin, and Hang Li. 2024. [Reft: Reasoning with reinforced fine-tuning](https://md.research.elamur.ai/free/en/papers/trung-2024-reft-reasoning-reinforced-fine-tuning/). arXiv preprint arXiv:2401.08967.
32. Shen-yun Miao, Chao-Chun Liang, and Keh-Yih Su. 2020. A diverse corpus for evaluating and developing english math word problem solvers. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 975–984. [find](https://scholar.google.com/scholar?q=Shen%2Dyun+Miao%2C+Chao%2DChun+Liang%2C+and+Keh%2DYih+Su%2E+2020%2E+A+diverse+corpus+for+evaluating+and+developing+english+math+word+problem+solvers%2E+In+Proceedings+of+the+58th+Annual+Meeting+of+the+Association+for)
33. Iman Mirzadeh, Keivan Alizadeh, Hooman Shahrokhi, Oncel Tuzel, Samy Bengio, and Mehrdad Farajtabar. 2024. [Gsm-symbolic: Understanding the limitations of mathematical reasoning in large language models](https://md.research.elamur.ai/free/en/papers/mirzadeh-2024-gsm-symbolic-understanding-limitations-mathematical/). arXiv preprint arXiv:2410.05229.
34. Philipp Mondorf and Barbara Plank. 2024. [Beyond accuracy: evaluating the reasoning behavior of large language models–a survey](https://md.research.elamur.ai/free/en/papers/mondorf-2024-accuracy-evaluating-reasoning-behavior-large/). arXiv preprint arXiv:2404.01869.
35. Sania Nayab, Giulio Rossolini, Marco Simoni, Andrea Saracino, Giorgio Buttazzo, Nicolamaria Manes, and Fabrizio Giacomelli. 2024. Concise thoughts: Impact of output length on llm reasoning and cost. arXiv preprint arXiv:2407.19825. — [arXiv:2407.19825](https://arxiv.org/abs/2407.19825) [find](https://scholar.google.com/scholar?q=Sania+Nayab%2C+Giulio+Rossolini%2C+Marco+Simoni%2C+Andrea+Saracino%2C+Giorgio+Buttazzo%2C+Nicolamaria+Manes%2C+and+Fabrizio+Giacomelli%2E+2024%2E+Concise+thoughts%3A+Impact+of+output+length+on+llm+reasoning+and+cost%2E+a)
36. Vu-Linh Nguyen, Mohammad Hossein Shaker, and Eyke Hüllermeier. 2022. How to measure uncertainty in uncertainty sampling for active learning. Machine Learning, 111(1):89–122. [find](https://scholar.google.com/scholar?q=Vu%2DLinh+Nguyen%2C+Mohammad+Hossein+Shaker%2C+and+Eyke+H%C3%BCllermeier%2E+2022%2E+How+to+measure+uncertainty+in+uncertainty+sampling+for+active+learning%2E+Machine+Learning%2C+111%281%29%3A89%E2%80%93122%2E)
37. Liangming Pan, Michael Saxon, Wenda Xu, Deepak Nathani, Xinyi Wang, and William Yang Wang. 2023. Automatically correcting large language models: Surveying the landscape of diverse self-correction strategies. arXiv preprint arXiv:2308.03188. [find](https://scholar.google.com/scholar?q=Liangming+Pan%2C+Michael+Saxon%2C+Wenda+Xu%2C+Deepak+Nathani%2C+Xinyi+Wang%2C+and+William+Yang+Wang%2E+2023%2E+Automatically+correcting+large+language+models%3A+Surveying+the+landscape+of+diverse+self%2Dcorrection+stra)
38. Xinyu Pan, Di Wang, and Fugee Tsung. 2025. Empowering intelligent quality control with large models: A comprehensive survey of methods, challenges, and perspectives. Authorea Preprints. [find](https://scholar.google.com/scholar?q=Xinyu+Pan%2C+Di+Wang%2C+and+Fugee+Tsung%2E+2025%2E+Empowering+intelligent+quality+control+with+large+models%3A+A+comprehensive+survey+of+methods%2C+challenges%2C+and+perspectives%2E+Authorea+Preprints%2E)
39. Chong Peng, Zhao Kang, Huiqing Li, and Qiang Cheng. 2015. Subspace clustering using log-determinant rank approximation. In Proceedings of the 21th ACM SIGKDD international conference on Knowledge Discovery and Data Mining, pages 925–934. [find](https://scholar.google.com/scholar?q=Chong+Peng%2C+Zhao+Kang%2C+Huiqing+Li%2C+and+Qiang+Cheng%2E+2015%2E+Subspace+clustering+using+log%2Ddeterminant+rank+approximation%2E+In+Proceedings+of+the+21th+ACM+SIGKDD+international+conference+on+Knowledge+Disc)
40. Yuxiao Qu, Matthew YR Yang, Amrith Setlur, Lewis Tunstall, Edward Emanuel Beeching, Ruslan Salakhutdinov, and Aviral Kumar. 2025. Optimizing test-time compute via meta reinforcement fine-tuning. arXiv preprint arXiv:2503.07572. [find](https://scholar.google.com/scholar?q=Yuxiao+Qu%2C+Matthew+YR+Yang%2C+Amrith+Setlur%2C+Lewis+Tunstall%2C+Edward+Emanuel+Beeching%2C+Ruslan+Salakhutdinov%2C+and+Aviral+Kumar%2E+2025%2E+Optimizing+test%2Dtime+compute+via+meta+reinforcement+fine%2Dtuning%2E+arXiv)
41. Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. 2023. [Direct preference optimization: Your language model is secretly a reward model](https://md.research.elamur.ai/free/en/papers/rafailov-2023-direct-preference-optimization-language-model/). Advances in neural information processing systems, 36:53728–53741.
42. Nils Reimers and Iryna Gurevych. 2019. [Sentence-bert: Sentence embeddings using siamese bert-networks](https://md.research.elamur.ai/free/en/papers/reimers-2019-sentence-bert-sentence-embeddings-siamese/). arXiv preprint arXiv:1908.10084.
43. Chuck Rosenberg, Martial Hebert, and Henry Schneider-man. 2005. Semi-supervised self-training of object detection models. [find](https://scholar.google.com/scholar?q=Chuck+Rosenberg%2C+Martial+Hebert%2C+and+Henry+Schneider%2Dman%2E+2005%2E+Semi%2Dsupervised+self%2Dtraining+of+object+detection+models%2E)
44. Noveen Sachdeva, Benjamin Coleman, Wang-Cheng Kang, Jianmo Ni, Lichan Hong, Ed H Chi, James Caverlee, Julian McAuley, and Derek Zhiyuan Cheng. 2024. How to train data-efficient llms. arXiv preprint arXiv:2402.09668. [find](https://scholar.google.com/scholar?q=Noveen+Sachdeva%2C+Benjamin+Coleman%2C+Wang%2DCheng+Kang%2C+Jianmo+Ni%2C+Lichan+Hong%2C+Ed+H+Chi%2C+James+Caverlee%2C+Julian+McAuley%2C+and+Derek+Zhiyuan+Cheng%2E+2024%2E+How+to+train+data%2Defficient+llms%2E+arXiv+preprint+ar)
45. Amir Saeidi, Shivanshu Verma, Md Nayem Uddin, and Chitta Baral. 2024. Insights into alignment: Evaluating dpo and its variants across multiple tasks. arXiv preprint arXiv:2404.14723. [find](https://scholar.google.com/scholar?q=Amir+Saeidi%2C+Shivanshu+Verma%2C+Md+Nayem+Uddin%2C+and+Chitta+Baral%2E+2024%2E+Insights+into+alignment%3A+Evaluating+dpo+and+its+variants+across+multiple+tasks%2E+arXiv+preprint+arXiv%3A2404%2E14723%2E)
46. Nabeel Seedat, Nicolas Huynh, Boris van Breugel, and Mihaela van der Schaar. 2024. Curated LLM: Synergy of LLMs and data curation for tabular augmentation in low-data regimes. In Forty-first International Conference on Machine Learning. [find](https://scholar.google.com/scholar?q=Nabeel+Seedat%2C+Nicolas+Huynh%2C+Boris+van+Breugel%2C+and+Mihaela+van+der+Schaar%2E+2024%2E+Curated+LLM%3A+Synergy+of+LLMs+and+data+curation+for+tabular+augmentation+in+low%2Ddata+regimes%2E+In+Forty%2Dfirst+Internati)
47. Ozan Sener and Silvio Savarese. 2017. Active learning for convolutional neural networks: A core-set approach. arXiv preprint arXiv:1708.00489. [find](https://scholar.google.com/scholar?q=Ozan+Sener+and+Silvio+Savarese%2E+2017%2E+Active+learning+for+convolutional+neural+networks%3A+A+core%2Dset+approach%2E+arXiv+preprint+arXiv%3A1708%2E00489%2E)
48. Avi Singh, John D Co-Reyes, Rishabh Agarwal, Ankesh Anand, Piyush Patil, Xavier Garcia, Peter J Liu, James Harrison, Jaehoon Lee, Kelvin Xu, et al. 2023. [Beyond human data: Scaling self-training for problem-solving with language models](https://md.research.elamur.ai/free/en/papers/singh-2023-human-data-scaling-self-training/). arXiv preprint arXiv:2312.06585.
49. 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 differential entropy for efficient llm fine-tuning. arXiv preprint arXiv:2601.23006. — [arXiv:2601.23006](https://arxiv.org/abs/2601.23006) [find](https://scholar.google.com/scholar?q=Junyou+Su%2C+He+Zhu%2C+Xiao+Luo%2C+Liyu+Zhang%2C+Hong%2DYu+Zhou%2C+Yun+Chen%2C+Peng+Li%2C+Yang+Liu%2C+and+Guanhua+Chen%2E+2026%2E+Instructdiff%3A+Domain%2Dadaptive+data+selection+via+differential+entropy+for+efficient+llm+fine)
50. Yifan Sun, Jingyan Shen, Yibin Wang, Tianyu Chen, Zhendong Wang, Mingyuan Zhou, and Huan Zhang. 2025a. Improving data efficiency for llm reinforcement fine-tuning through difficulty-targeted online data selection and rollout replay. arXiv preprint arXiv:2506.05316. — [arXiv:2506.05316](https://arxiv.org/abs/2506.05316) [find](https://scholar.google.com/scholar?q=Yifan+Sun%2C+Jingyan+Shen%2C+Yibin+Wang%2C+Tianyu+Chen%2C+Zhendong+Wang%2C+Mingyuan+Zhou%2C+and+Huan+Zhang%2E+2025a%2E+Improving+data+efficiency+for+llm+reinforcement+fine%2Dtuning+through+difficulty%2Dtargeted+online+da)
51. Zexu Sun, Yongcheng Zeng, Erxue Min, Heyang Gao, Bokai Ji, and Xu Chen. 2025b. Cog-rethinker: Hierarchical metacognitive reinforcement learning for llm reasoning. arXiv preprint arXiv:2510.15979. — [arXiv:2510.15979](https://arxiv.org/abs/2510.15979) [find](https://scholar.google.com/scholar?q=Zexu+Sun%2C+Yongcheng+Zeng%2C+Erxue+Min%2C+Heyang+Gao%2C+Bokai+Ji%2C+and+Xu+Chen%2E+2025b%2E+Cog%2Drethinker%3A+Hierarchical+metacognitive+reinforcement+learning+for+llm+reasoning%2E+arXiv+preprint+arXiv%3A2510%2E15979%2E+%E2%80%94+%5Ba)
52. Zhiqing Sun, Longhui Yu, Yikang Shen, Weiyang Liu, Yiming Yang, Sean Welleck, and Chuang Gan. 2024. Easy-to-hard generalization: Scalable alignment beyond human supervision. Advances in Neural Information Processing Systems, 37:51118–51168. [find](https://scholar.google.com/scholar?q=Zhiqing+Sun%2C+Longhui+Yu%2C+Yikang+Shen%2C+Weiyang+Liu%2C+Yiming+Yang%2C+Sean+Welleck%2C+and+Chuang+Gan%2E+2024%2E+Easy%2Dto%2Dhard+generalization%3A+Scalable+alignment+beyond+human+supervision%2E+Advances+in+Neural+Informa)
53. Daniel Truhn, Jorge S Reis-Filho, and Jakob Nikolas Kather. 2023. Large language models should be used as scientific reasoning engines, not knowledge databases. Nature medicine, 29(12):2983–2984. [find](https://scholar.google.com/scholar?q=Daniel+Truhn%2C+Jorge+S+Reis%2DFilho%2C+and+Jakob+Nikolas+Kather%2E+2023%2E+Large+language+models+should+be+used+as+scientific+reasoning+engines%2C+not+knowledge+databases%2E+Nature+medicine%2C+29%2812%29%3A2983%E2%80%932984%2E)
54. Bo Wang, Mingda Chen, Ming Deng, Youfang Lin, Mark Harman, Mike Papadakis, and Jie M Zhang. 2026. A comprehensive study on large language models for mutation testing. ACM Transactions on Software Engineering and Methodology. [find](https://scholar.google.com/scholar?q=Bo+Wang%2C+Mingda+Chen%2C+Ming+Deng%2C+Youfang+Lin%2C+Mark+Harman%2C+Mike+Papadakis%2C+and+Jie+M+Zhang%2E+2026%2E+A+comprehensive+study+on+large+language+models+for+mutation+testing%2E+ACM+Transactions+on+Software+Engi)
55. Shenzhi Wang, Le Yu, Chang Gao, Chujie Zheng, Shixuan Liu, Rui Lu, Kai Dang, Xionghui Chen, Jianxin Yang, Zhenru Zhang, et al. 2025a. [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.
56. Xiaoxuan Wang, Yihe Deng, Mingyu Derek Ma, and Wei Wang. 2025b. Entropy-based adaptive weighting for self-training. arXiv preprint arXiv:2503.23913. — [arXiv:2503.23913](https://arxiv.org/abs/2503.23913) [find](https://scholar.google.com/scholar?q=Xiaoxuan+Wang%2C+Yihe+Deng%2C+Mingyu+Derek+Ma%2C+and+Wei+Wang%2E+2025b%2E+Entropy%2Dbased+adaptive+weighting+for+self%2Dtraining%2E+arXiv+preprint+arXiv%3A2503%2E23913%2E+%E2%80%94+%5BarXiv%3A2503%2E23913%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2503%2E239)
57. Yizhong Wang, Yeganeh Kordi, Swaroop Mishra, Alisa Liu, Noah A Smith, Daniel Khashabi, and Hannaneh Hajishirzi. 2022. Self-instruct: Aligning language models with self-generated instructions. arXiv preprint arXiv:2212.10560. [find](https://scholar.google.com/scholar?q=Yizhong+Wang%2C+Yeganeh+Kordi%2C+Swaroop+Mishra%2C+Alisa+Liu%2C+Noah+A+Smith%2C+Daniel+Khashabi%2C+and+Hannaneh+Hajishirzi%2E+2022%2E+Self%2Dinstruct%3A+Aligning+language+models+with+self%2Dgenerated+instructions%2E+arXiv+pr)
58. Yubo Wang, Xueguang Ma, Ge Zhang, Yuansheng Ni, Abhranil Chandra, Shiguang Guo, Weiming Ren, Aaran Arulraj, Xuan He, Ziyan Jiang, et al. 2024. [Mmlu-pro: A more robust and challenging multi-task language understanding benchmark](https://md.research.elamur.ai/free/en/papers/wang-2024-mmlu-pro-more-robust-challenging/). Advances in Neural Information Processing Systems, 37:95266– 95290.
59. Yu Xia, Subhojyoti Mukherjee, Zhouhang Xie, Junda Wu, Xintong Li, Ryan Aponte, Hanjia Lyu, Joe Barrow, Hongjie Chen, Franck Dernoncourt, et al. 2025. [From selection to generation: A survey of llm-based active learning](https://md.research.elamur.ai/free/en/papers/xia-2025-selection-generation-survey-llm-based/). arXiv preprint arXiv:2502.11767.
60. Zhiping Xiao, Yusheng Zhao, Qixin Zhang, Jiaye Xie, Wanjia Zhao, Weizhi Zhang, Xiao Luo, Philip S Yu, and Ming Zhang. 2026. Sample lottery: Unsupervised discovery of critical instances for llm reasoning. In The Fourteenth International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Zhiping+Xiao%2C+Yusheng+Zhao%2C+Qixin+Zhang%2C+Jiaye+Xie%2C+Wanjia+Zhao%2C+Weizhi+Zhang%2C+Xiao+Luo%2C+Philip+S+Yu%2C+and+Ming+Zhang%2E+2026%2E+Sample+lottery%3A+Unsupervised+discovery+of+critical+instances+for+llm+reasoni)
61. Ling Yang, Zhaochen Yu, Tianjun Zhang, Minkai Xu, Joseph E Gonzalez, Bin Cui, and Shuicheng Yan. 2024. Supercorrect: Advancing small llm reasoning with thought template distillation and self-correction. arXiv preprint arXiv:2410.09008. — [arXiv:2410.09008](https://arxiv.org/abs/2410.09008) [find](https://scholar.google.com/scholar?q=Ling+Yang%2C+Zhaochen+Yu%2C+Tianjun+Zhang%2C+Minkai+Xu%2C+Joseph+E+Gonzalez%2C+Bin+Cui%2C+and+Shuicheng+Yan%2E+2024%2E+Supercorrect%3A+Advancing+small+llm+reasoning+with+thought+template+distillation+and+self%2Dcorrectio)
62. Wang Yang, Zirui Liu, Hongye Jin, Qingyu Yin, Vipin Chaudhary, and Xiaotian Han. 2025. Longer context, deeper thinking: Uncovering the role of long-context ability in reasoning. arXiv preprint arXiv:2505.17315. — [arXiv:2505.17315](https://arxiv.org/abs/2505.17315) [find](https://scholar.google.com/scholar?q=Wang+Yang%2C+Zirui+Liu%2C+Hongye+Jin%2C+Qingyu+Yin%2C+Vipin+Chaudhary%2C+and+Xiaotian+Han%2E+2025%2E+Longer+context%2C+deeper+thinking%3A+Uncovering+the+role+of+long%2Dcontext+ability+in+reasoning%2E+arXiv+preprint+arXiv%3A2)
63. 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.
64. Eric Zelikman, Yuhuai Wu, and Noah D Goodman. 2022. Star: Self-taught reasoner. In Proceedings of the NIPS, volume 22. [find](https://scholar.google.com/scholar?q=Eric+Zelikman%2C+Yuhuai+Wu%2C+and+Noah+D+Goodman%2E+2022%2E+Star%3A+Self%2Dtaught+reasoner%2E+In+Proceedings+of+the+NIPS%2C+volume+22%2E)
65. Jenny Zhang, Shengran Hu, Cong Lu, Robert T. Lange, and Jeff Clune. 2025a. Darwin godel machine: Open-ended evolution of self-improving agents. CoRR, abs/2505.22954. [find](https://scholar.google.com/scholar?q=Jenny+Zhang%2C+Shengran+Hu%2C+Cong+Lu%2C+Robert+T%2E+Lange%2C+and+Jeff+Clune%2E+2025a%2E+Darwin+godel+machine%3A+Open%2Dended+evolution+of+self%2Dimproving+agents%2E+CoRR%2C+abs%2F2505%2E22954%2E)
66. Juzheng Zhang, Jiacheng You, Ashwinee Panda, and Tom Goldstein. 2025b. Lori: Reducing cross-task interference in multi-task low-rank adaptation. arXiv preprint arXiv:2504.07448. — [arXiv:2504.07448](https://arxiv.org/abs/2504.07448) [find](https://scholar.google.com/scholar?q=Juzheng+Zhang%2C+Jiacheng+You%2C+Ashwinee+Panda%2C+and+Tom+Goldstein%2E+2025b%2E+Lori%3A+Reducing+cross%2Dtask+interference+in+multi%2Dtask+low%2Drank+adaptation%2E+arXiv+preprint+arXiv%3A2504%2E07448%2E+%E2%80%94+%5BarXiv%3A2504%2E07448%5D%28h)
67. MMLU-Pro-MATH (Wang et al., 2024): MMLU-Pro-MATH is the math subset of MMLU-Pro dataset for data enhancement. MMLU-Pro dataset features 12,000 complex questions spanning various disciplines, including 1,356 mathematical problems. We randomly sample 400 problems for training and use the rest for eval… [find](https://scholar.google.com/scholar?q=MMLU%2DPro%2DMATH+%28Wang+et+al%2E%2C+2024%29%3A+MMLU%2DPro%2DMATH+is+the+math+subset+of+MMLU%2DPro+dataset+for+data+enhancement%2E+MMLU%2DPro+dataset+features+12%2C000+complex+questions+spanning+various+disciplines%2C+including)
68. Yang Zhang, Amr Mohamed, Hadi Abdine, Guokan Shang, and Michalis Vazirgiannis. 2026. [Beyond random sampling: Efficient language model pretraining via curriculum learning](https://md.research.elamur.ai/free/en/papers/zhang-2026-beyond-random-sampling-efficient-language/). In Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Pa…
69. Wanru Zhao, Hongxiang Fan, Shell Xu Hu, Wangchunshu Zhou, Bofan Chen, and Nicholas D Lane. 2025. Clues: Collaborative high-quality data selection for llms via training dynamics. arXiv preprint arXiv:2507.03004. — [arXiv:2507.03004](https://arxiv.org/abs/2507.03004) [find](https://scholar.google.com/scholar?q=Wanru+Zhao%2C+Hongxiang+Fan%2C+Shell+Xu+Hu%2C+Wangchunshu+Zhou%2C+Bofan+Chen%2C+and+Nicholas+D+Lane%2E+2025%2E+Clues%3A+Collaborative+high%2Dquality+data+selection+for+llms+via+training+dynamics%2E+arXiv+preprint+arXiv%3A2)
70. Han Zhong, Zikang Shan, Guhao Feng, Wei Xiong, Xinle Cheng, Li Zhao, Di He, Jiang Bian, and Liwei Wang. 2024. Dpo meets ppo: Reinforced token optimization for rlhf. arXiv preprint arXiv:2404.18922. — [arXiv:2404.18922](https://arxiv.org/abs/2404.18922) [find](https://scholar.google.com/scholar?q=Han+Zhong%2C+Zikang+Shan%2C+Guhao+Feng%2C+Wei+Xiong%2C+Xinle+Cheng%2C+Li+Zhao%2C+Di+He%2C+Jiang+Bian%2C+and+Liwei+Wang%2E+2024%2E+Dpo+meets+ppo%3A+Reinforced+token+optimization+for+rlhf%2E+arXiv+preprint+arXiv%3A2404%2E18922%2E+%E2%80%94+)
71. Zhi Zhou, Tan Yuhao, Zenan Li, Yuan Yao, Lan-Zhe Guo, Xiaoxing Ma, and Yu-Feng Li. 2025. Bridging internal probability and self-consistency for effective and efficient llm reasoning. arXiv preprint arXiv:2502.00511. — [arXiv:2502.00511](https://arxiv.org/abs/2502.00511) [find](https://scholar.google.com/scholar?q=Zhi+Zhou%2C+Tan+Yuhao%2C+Zenan+Li%2C+Yuan+Yao%2C+Lan%2DZhe+Guo%2C+Xiaoxing+Ma%2C+and+Yu%2DFeng+Li%2E+2025%2E+Bridging+internal+probability+and+self%2Dconsistency+for+effective+and+efficient+llm+reasoning%2E+arXiv+preprint+ar)
72. ASDiv (Miao et al., 2020): ASDiv dataset features 2,305 complex mathematical questions. We randomly sample 230 problems for training and use the rest for evaluation. [find](https://scholar.google.com/scholar?q=ASDiv+%28Miao+et+al%2E%2C+2020%29%3A+ASDiv+dataset+features+2%2C305+complex+mathematical+questions%2E+We+randomly+sample+230+problems+for+training+and+use+the+rest+for+evaluation%2E)
73. B This appendix provides concrete examples of the prompt templates used for the data augmentation strategies described in Section 3.3 and Teacher Clarity Score Prompt. [find](https://scholar.google.com/scholar?q=B+This+appendix+provides+concrete+examples+of+the+prompt+templates+used+for+the+data+augmentation+strategies+described+in+Section+3%2E3+and+Teacher+Clarity+Score+Prompt%2E)
74. Yaoming Zhu, Sidi Lu, Lei Zheng, Jiaxian Guo, Weinan Zhang, Jun Wang, and Yong Yu. 2018. [Texygen: A benchmarking platform for text generation models](https://md.research.elamur.ai/free/en/papers/zhu-2018-texygen-benchmarking-platform-text-generation/). In The 41st international ACM SIGIR conference on research & development in information retrieval, pages 1097–1100.
75. Yuchang Zhu, Huizhe Zhang, Bingzhe Wu, Jintang Li, Zibin Zheng, Peilin Zhao, Liang Chen, and Yatao Bian. 2025. Measuring diversity in synthetic datasets. arXiv preprint arXiv:2502.08512. — [arXiv:2502.08512](https://arxiv.org/abs/2502.08512) [find](https://scholar.google.com/scholar?q=Yuchang+Zhu%2C+Huizhe+Zhang%2C+Bingzhe+Wu%2C+Jintang+Li%2C+Zibin+Zheng%2C+Peilin+Zhao%2C+Liang+Chen%2C+and+Yatao+Bian%2E+2025%2E+Measuring+diversity+in+synthetic+datasets%2E+arXiv+preprint+arXiv%3A2502%2E08512%2E+%E2%80%94+%5BarXiv%3A2502)
76. Yuxin Zuo, Kaiyan Zhang, Li Sheng, Shang Qu, Ganqu Cui, Xuekai Zhu, Haozhan Li, Yuchen Zhang, Xinwei Long, Ermo Hua, et al. 2025. [Ttrl: Test-time reinforcement learning](https://md.research.elamur.ai/free/en/papers/zuo-2025-ttrl-test-time-reinforcement-learning/). arXiv preprint arXiv:2504.16084.

*Updated 2 November 2026*