Elamur Research⌕
☰
home / papers / 2025 / Generalizing Verifiable Instruction Following

Generalizing Verifiable Instruction Following

Paper · 2025 · Valentina Pyatkin, Saumya Malik, Victoria Graf et al.

Summary

Language models are commonly tested on whether they obey checkable output rules, such as word counts or required formats. Most of this testing relies on IFEval and its 25 constraint types. The authors argue that high IFEval scores mostly reflect overfitting to that narrow set rather than real skill. To measure transfer, they build IFBench: 58 new hand-crafted constraints, each checked automatically by a Python function. These are attached to held-out WildChat prompts, and humans vetted each pairing for compatibility. Leading systems such as GPT-4.1 and Claude 3.7 Sonnet score below 50% on it.

The authors also release IFTrain, 29 additional training constraints. They study how to train for generalization with reinforcement learning from verifiable rewards (IF-RLVR) using GRPO. Their training prompts stack several constraints and widen parameter ranges. This produces large gains on both benchmarks for Tülu, Qwen and OLMo models. However, RLVR-tuned models tend to favor the constraint over the underlying task, which motivates adding a preference reward. The result is a contamination-resistant evaluation and a practical training recipe.

Key points

  • IFBench design:
    • 58 unseen verifiable constraints across 7 categories: count, ratio, words, sentence, format, custom and copy.
    • 300 test prompts, each carrying one or two constraints.
    • Evaluated in single-turn and multi-turn (rewrite) settings, with strict and loose accuracy.
  • IFTrain: 29 new training constraints with verifiers, which more than doubles the available training constraint types. Training prompts (about 60k–100k) combine Tülu-3-SFT instructions with constraints from IFTrain and IFEval. A conflict dictionary prevents contradictory constraint pairs.
  • Tülu-3-8B with IF-RLVR: IFEval rises from 82.4 to 92.2, and IFBench from 28.9 to 45.9.
  • Qwen-2.5-7B base with IF-RLVR: reaches 87.8 on IFEval and 54.7 on IFBench. The approach also works for OLMo models.
  • Trade-off: a GPT-4.1 judge rated base-policy answers higher than IF-RLVR answers (about 7 vs. 6.4 out of 10) when constraints were ignored, which signals over-optimization toward constraints. Code and data are available at github.com/allenai/IFBench.

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

🎓University of WashingtonSeattle · United States353 works · 200,043 citations🔬Allen Institute for Artificial IntelligenceSeattle · United States275 works · 97,634 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

Cite

Cite as Valentina Pyatkin, Saumya Malik, Victoria Graf, Hamish Ivison, Shengyi Huang, Pradeep Dasigi, Nathan Lambert and Hannaneh Hajishirzi. 2025. Generalizing Verifiable Instruction Following. arXiv:2507.02833.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2507.02833 · arXiv 2507.02833

@misc{pyatkin2025generalizing,
  title = {{Generalizing Verifiable Instruction Following}},
  author = {Valentina Pyatkin, Saumya Malik, Victoria Graf, Hamish Ivison, Shengyi Huang, Pradeep Dasigi, Nathan Lambert and Hannaneh Hajishirzi},
  year = {2025},
  doi = {10.48550/arxiv.2507.02833},
  eprint = {2507.02833},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2507.02833}
}

References

36 references, 17 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Bo Adler, Niket Agarwal, Ashwath Aithal, Dong H Anh, Pallab Bhattacharya, Annika Brundyn, Jared Casper, Bryan Catanzaro, Sharon Clay, Jonathan Cohen, et al. Nemotron-4 340b technical report. arXiv preprint arXiv:2406.11704, 2024.
  2. Noga Amit, Shafi Goldwasser, Orr Paradise, and Guy Rothblum. Models that prove their own correctness. arXiv preprint arXiv:2405.15722, 2024. — [arXiv:2405.15722](https://arxiv.org/abs/2405.15722) find
  3. Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Yunxuan Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, et al. Scaling instruction-finetuned language models. J. Mach. Learn. Res., 2024. find
  4. Ricardo Dominguez-Olmedo, Florian E Dorner, and Moritz Hardt. Training on the test task confounds evaluation and emergence. arXiv preprint arXiv:2407.07890, 2024. — [arXiv:2407.07890](https://arxiv.org/abs/2407.07890) find
  5. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. The llama 3 herd of models. arXiv preprint arXiv:2407.21783, 2024.
  6. Shahriar Golchin and Mihai Surdeanu. Time travel in llms: Tracing data contamination in large language models. In The Twelfth International Conference on Learning Representations.
  7. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, et al. The llama 3 herd of models. arXiv preprint arXiv:2407.21783, 2024.
  8. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948, 2025.
  9. Andreas Hochlehnert, Hardik Bhatnagar, Vishaal Udandarao, Samuel Albanie, Ameya Prabhu, and Matthias Bethge. A sober look at progress in language model reasoning: Pitfalls and paths to reproducibility. 2025. find
  10. Matthew Douglas Hoffman, Du Phan, David Dohan, Sholto Douglas, Tuan Anh Le, Aaron Parisi, Pavel Sountsov, Charles Sutton, Sharad Vikram, and Rif A Saurous. Training chain-of-thought via latent-variable inference. In NeurIPS, 2023. find
  11. Hamish Ivison, Yizhong Wang, Valentina Pyatkin, Nathan Lambert, Matthew E. Peters, Pradeep Dasigi, Joel Jang, David Wadden, Noah A. Smith, Iz Beltagy, and Hanna Hajishirzi. Camels in a changing climate: Enhancing lm adaptation with tulu 2. ArXiv, abs/2311.10702, 2023.
  12. Yuxin Jiang, Yufei Wang, Xingshan Zeng, Wanjun Zhong, Liangyou Li, Fei Mi, Lifeng Shang, Xin Jiang, Qun Liu, and Wei Wang. Followbench: A multi-level fine-grained constraints following benchmark for large language models. CoRR, 2023.
  13. Amirhossein Kazemnejad, Milad Aghajohari, Eva Portelance, Alessandro Sordoni, Siva Reddy, Aaron Courville, and Nicolas Le Roux. Vineppo: Unlocking rl potential for llm reasoning through refined credit assignment. 2024. find
  14. Joongwon Kim, Anirudh Goyal, Aston Zhang, Bo Xiong, Rui Hou, Melanie Kambadur, Dhruv Mahajan, Hannaneh Hajishirzi, and Liang Tan. A systematic examination of preference learning through the lens of instruction-following. In Luis Chiruzzo, Alan Ritter, and Lu Wang, editors, Proceedings of the 2025 Co… find
  15. Nathan Lambert, Jacob Morrison, Valentina Pyatkin, Shengyi Huang, Hamish Ivison, Faeze Brahman, Lester James V Miranda, Alisa Liu, Nouha Dziri, Shane Lyu, et al. T\\" ulu 3: Pushing frontiers in open language model post-training. arXiv preprint arXiv:2411.15124, 2024.
  16. Gili Lior, Asaf Yehudai, Ariel Gera, and Liat Ein-Dor. Wildifeval: Instruction following in the wild. arXiv preprint arXiv:2503.06573, 2025. find
  17. Seyed Iman Mirzadeh, Keivan Alizadeh, Hooman Shahrokhi, Oncel Tuzel, Samy Bengio, and Mehrdad Farajtabar. Gsm-symbolic: Understanding the limitations of mathematical reasoning in large language models. In The Thirteenth International Conference on Learning Representations.
  18. Team OLMo, Pete Walsh, Luca Soldaini, Dirk Groeneveld, Kyle Lo, Shane Arora, Akshita Bhagia, Yuling Gu, Shengyi Huang, Matt Jordan, et al. 2 olmo 2 furious. arXiv preprint arXiv:2501.00656, 2024. find
  19. Yiwei Qin, Kaiqiang Song, Yebowen Hu, Wenlin Yao, Sangwoo Cho, Xiaoyang Wang, Xuansheng Wu, Fei Liu, Pengfei Liu, and Dong Yu. Infobench: Evaluating instruction following ability in large language models. In Findings of the Association for Computational Linguistics ACL 2024, pages 13025–13048, 2024.
  20. Manley Roberts, Himanshu Thakur, Christine Herlihy, Colin White, and Samuel Dooley. To the cutoff... and beyond? a longitudinal perspective on llm data contamination. In The Twelfth International Conference on Learning Representations. find
  21. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Y Wu, et al. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300, 2024.
  22. Alessandro Stolfo, Vidhisha Balachandran, Safoora Yousefi, Eric Horvitz, and Besmira Nushi. Improving instruction-following in language models through activation steering. arXiv preprint arXiv:2410.12877, 2024. — [arXiv:2410.12877](https://arxiv.org/abs/2410.12877) find
  23. Jiao Sun, Yufei Tian, Wangchunshu Zhou, Nan Xu, Qian Hu, Rahul Gupta, John Frederick Wieting, Nanyun Peng, and Xuezhe Ma. Evaluating large language models on controlled generation tasks. In The 2023 Conference on Empirical Methods in Natural Language Processing. find
  24. Zhi Rui Tam, Cheng-Kuang Wu, Yi-Lin Tsai, Chieh-Yen Lin, Hung-yi Lee, and Yun-Nung Chen. Let me speak freely? a study on the impact of format restrictions on performance of large language models. arXiv preprint arXiv:2408.02442, 2024. — [arXiv:2408.02442](https://arxiv.org/abs/2408.02442) find
  25. Gemma Team, Aishwarya Kamath, Johan Ferret, Shreya Pathak, Nino Vieillard, Ramona Merhej, Sarah Perrin, Tatiana Matejovicova, Alexandre Ramé, Morgane Rivière, et al. Gemma 3 technical report. arXiv preprint arXiv:2503.19786, 2025. find
  26. Qwen Team. Qwen2 technical report. arXiv preprint arXiv:2412.15115, 2024. find
  27. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023.
  28. Eric Wallace, Kai Xiao, Reimar Leike, Lilian Weng, Johannes Heidecke, and Alex Beutel. The instruction hierarchy: Training llms to prioritize privileged instructions. arXiv preprint arXiv:2404.13208, 2024. find
  29. Yiping Wang, Qing Yang, Zhiyuan Zeng, Liliang Ren, Liyuan Liu, Baolin Peng, Hao Cheng, Xuehai He, Kuan Wang, Jianfeng Gao, et al. Reinforcement learning for reasoning in large language models with one training example. arXiv preprint arXiv:2504.20571, 2025. find
  30. Zhaoyang Wang, Jinqi Jiang, Huichi Zhou, Wenhao Zheng, Xuchao Zhang, Chetan Bansal, and Huaxiu Yao. Verifiable format control for large language model generations. In Findings of the Association for Computational Linguistics: NAACL 2025, pages 3499–3513, 2025. find
  31. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, et al. Qwen3 technical report. arXiv preprint arXiv:2505.09388, 2025.
  32. An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, et al. Qwen2. 5 technical report. arXiv preprint arXiv:2412.15115, 2024. find
  33. Eric Zelikman, Yuhuai Wu, Jesse Mu, and Noah Goodman. Star: Bootstrapping reasoning with reasoning. Advances in Neural Information Processing Systems, 35:15476–15488, 2022.
  34. Wenting Zhao, Xiang Ren, Jack Hessel, Claire Cardie, Yejin Choi, and Yuntian Deng. Wildchat: 1m chatgpt interaction logs in the wild. In The Twelfth International Conference on Learning Representations.
  35. Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric Xing, et al. Judging llm-as-a-judge with mt-bench and chatbot arena. Advances in Neural Information Processing Systems, 36:46595–46623, 2023.
  36. Jeffrey Zhou, Tianjian Lu, Swaroop Mishra, Siddhartha Brahma, Sujoy Basu, Yi Luan, Denny Zhou, and Le Hou. Instruction-following evaluation for large language models. arXiv preprint arXiv:2311.07911, 2023.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/pyatkin-2025-generalizing-verifiable-instruction-following/