Elamur Research
☰
home / papers / 2024 / Generative Verifiers: Reward Modeling as Next-Token Prediction

Generative Verifiers: Reward Modeling as Next-Token Prediction

Written for the Research library (claude-opus-5-5) from the paper text, in our own words: the license "arXiv — non-exclusive licence to distribute" does not allow redistributing the text itself. Original — arxiv.org/pdf/2408.15240v3

Summary

Best-of-N selection lets an LLM sample many candidate solutions and relies on a learned verifier to choose the best one. These verifiers are usually trained as discriminative classifiers that output a single score, which leaves the model's text-generation abilities unused. The authors, from Google DeepMind and partner universities, propose GenRM, which trains the verifier with ordinary next-token prediction. The verifier is asked whether a solution is correct, and its score is the probability it gives to the "Yes" token. The variant GenRM-CoT first writes a step-by-step verification rationale and then gives its verdict. At test time, several rationales are sampled and their "Yes" probabilities are averaged by majority voting. The same model can also be trained on verification and solution generation together. Training rationales are synthetic, generated by Gemini 1.0 Pro and filtered for correctness. On algorithmic string tasks and math reasoning, GenRM outperforms discriminative reward models, DPO-based verifiers, LLM-as-a-Judge and self-consistency. This shows that recasting reward modeling as generation unlocks chain-of-thought reasoning and extra inference compute for verification.

Key points

  • Algorithmic tasks: Gemma-2B verifiers were tested on Last Letter Concat and BBH Word Sorting. Best-of-N success rose from 5% to 45.3%, close to what an oracle verifier achieves.
  • GSM8K: a Gemma2-9B GenRM-CoT verifier ranked solutions produced by Gemini 1.0 Pro. Problems solved rose from 73% to 93.4%, which beats GPT-4 and Gemini 1.5 Pro.
  • Easy-to-hard transfer: verifiers trained only on grade-school math improved Best-of-32 on MATH500 from 28% to 44.6%, and on MMLU abstract algebra from 37.9% to 53.5%.
  • Synthetic rationales are enough: model-generated verification rationales, though imperfect, were sufficient to teach the verifier to catch subtle reasoning errors that discriminative reward models miss.
  • Scaling: generative verifiers improve more than discriminative ones as model size grows. They also gain from more test-time compute through majority voting. The work appeared at ICLR 2025 (arXiv:2408.15240).

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

Fields: Mathematics & statistics

Cited by

141 papers of our library cite it (of the 140673 whose bibliographies we read); 4 of them have a page here, 137 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2408.15240 · arXiv 2408.15240

@misc{zhang2024generative,
  title = {{Generative Verifiers: Reward Modeling as Next-Token Prediction}},
  author = {Zhang, Lunjun and Hosseini, Arian and Bansal, Hritik and Kazemi, Mehran and Kumar, Aviral and Agarwal, Rishabh},
  year = {2024},
  doi = {10.48550/arxiv.2408.15240},
  eprint = {2408.15240},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2408.15240}
}

References

58 references, 27 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkat, et al. Gpt-4 technical report. arXiv preprint arXiv:2303.08774, 2023. find our card
  2. R. Agarwal, A. Singh, L. M. Zhang, B. Bohnet, S. Chan, A. Anand, Z. Abbas, A. Nova, J. D. Co-Reyes, E. Chu, et al. Many-shot in-context learning. arXiv preprint arXiv:2404.11018, 2024. find our card
  3. Z. Ankner, M. Paul, B. Cui, J. D. Chang, and P. Ammanabrolu. Critique-out-loud reward models. arXiv preprint arXiv:2408.11791, 2024. find
  4. Y. Bai, S. Kadavath, S. Kundu, A. Askell, J. Kernion, A. Jones, A. Chen, A. Goldie, A. Mirhoseini, C. McKinnon, et al. Constitutional ai: Harmlessness from ai feedback. arXiv preprint arXiv:2212.08073, 2022. find our card
  5. M. Besta, L. Paleari, A. Kubicek, P. Nyczyk, R. Gerstenberger, P. Iff, T. Lehmann, H. Niewiadomski, and T. Hoefler. Checkembed: Effective verification of llm solutions to open-ended tasks. arXiv preprint arXiv:2406.02524, 2024. — [arXiv:2406.02524](https://arxiv.org/abs/2406.02524) find
  6. S. Borgeaud, A. Mensch, J. Hoffmann, T. Cai, E. Rutherford, K. Millican, G. B. Van Den Driessche, J.-B. Lespiau, B. Damoc, A. Clark, et al. Improving language models by retrieving from trillions of tokens. In International conference on machine learning, pages 2206–2240. PMLR, 2022. find
  7. B. Brown, J. Juravsky, R. Ehrlich, R. Clark, Q. V. Le, C. Ré, and A. Mirhoseini. Large language monkeys: Scaling inference compute with repeated sampling. arXiv preprint arXiv:2407.21787, 2024. find our card
  8. E. Charniak and M. Johnson. Coarse-to-fine n-best parsing and maxent discriminative reranking. In Proceedings of the 43rd Annual Meeting of the Association for Computational Linguistics (ACL’05), pages 173–180, 2005. find our card
  9. A. Chowdhery, S. Narang, J. Devlin, M. Bosma, G. Mishra, A. Roberts, P. Barham, H. W. Chung, C. Sutton, S. Gehrmann, et al. Palm: Scaling language modeling with pathways. Journal of Machine Learning Research, 24(240):1–113, 2023. find
  10. H. W. Chung, L. Hou, S. Longpre, B. Zoph, Y. Tay, W. Fedus, Y. Li, X. Wang, M. Dehghani, S. Brahma, et al. Scaling instruction-finetuned language models. arXiv preprint arXiv:2210.11416, 2022. find
  11. K. Cobbe, V. Kosaraju, M. Bavarian, M. Chen, H. Jun, L. Kaiser, M. Plappert, J. Tworek, J. Hilton, R. Nakano, et al. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168, 2021. find our card
  12. Gemma Team, T. Mesnard, C. Hardin, R. Dadashi, S. Bhupatiraju, S. Pathak, L. Sifre, M. Rivière, M. S. Kale, J. Love, et al. Gemma: Open models based on gemini research and technology. arXiv preprint arXiv:2403.08295, 2024a. find
  13. Gemma Team, M. Riviere, S. Pathak, P. G. Sessa, C. Hardin, S. Bhupatiraju, L. Hussenot, T. Mesnard, B. Shahriari, A. Ramé, et al. Gemma 2: Improving open language models at a practical size. arXiv preprint arXiv:2408.00118, 2024b. find
  14. G. T. Google, R. Anil, S. Borgeaud, Y. Wu, J.-B. Alayrac, J. Yu, R. Soricut, J. Schalkwyk, A. M. Dai, A. Hauth, et al. Gemini: a family of highly capable multimodal models. arXiv preprint arXiv:2312.11805, 2023. find our card
  15. D. Hendrycks, C. Burns, S. Basart, A. Zou, M. Mazeika, D. Song, and J. Steinhardt. Measuring massive multitask language understanding. arXiv preprint arXiv:2009.03300, 2020. find our card
  16. D. Hendrycks, C. Burns, S. Kadavath, A. Arora, S. Basart, E. Tang, D. Song, and J. Steinhardt. Measuring mathematical problem solving with the math dataset. arXiv preprint arXiv:2103.03874, 2021. find our card
  17. A. Hosseini, X. Yuan, N. Malkin, A. Courville, A. Sordoni, and R. Agarwal. V-star: Training verifiers for self-taught reasoners. arXiv preprint arXiv:2402.06457, 2024. find our card
  18. S. Kapoor, N. Gruver, M. Roberts, K. Collins, A. Pal, U. Bhatt, A. Weller, S. Dooley, M. Goldblum, and A. G. Wilson. Large language models must be taught to know what they don’t know. arXiv preprint arXiv:2406.08391, 2024. find
  19. S. Kim, J. Shin, Y. Cho, J. Jang, S. Longpre, H. Lee, S. Yun, S. Shin, S. Kim, J. Thorne, et al. Prometheus: Inducing fine-grained evaluation capability in language models. In The Twelfth International Conference on Learning Representations, 2023. find our card
  20. D. P. Kingma. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014. find
  21. N. Lambert, V. Pyatkin, J. Morrison, L. Miranda, B. Y. Lin, K. Chandu, N. Dziri, S. Kumar, T. Zick, Y. Choi, et al. Rewardbench: Evaluating reward models for language modeling. arXiv preprint arXiv:2403.13787, 2024. find
  22. H. Lightman, V. Kosaraju, Y. Burda, H. Edwards, B. Baker, T. Lee, J. Leike, J. Schulman, I. Sutskever, and K. Cobbe. Let’s verify step by step. arXiv preprint arXiv:2305.20050, 2023. find our card
  23. Z. Lin, D. Pathak, B. Li, J. Li, X. Xia, G. Neubig, P. Zhang, and D. Ramanan. Evaluating text-to-visual generation with image-to-text generation. arXiv preprint arXiv:2404.01291, 2024. — [arXiv:2404.01291](https://arxiv.org/abs/2404.01291) find
  24. Z. Ling, Y. Fang, X. Li, Z. Huang, M. Lee, R. Memisevic, and H. Su. Deductive verification of chain-of-thought reasoning. Advances in Neural Information Processing Systems, 36, 2024. find our card
  25. Y. Liu, A. Singh, C. D. Freeman, J. D. Co-Reyes, and P. J. Liu. Improving large language model fine-tuning for solving math problems. arXiv preprint arXiv:2310.10047, 2023. — [arXiv:2310.10047](https://arxiv.org/abs/2310.10047) find
  26. I. Loshchilov and F. Hutter. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101, 2017. find
  27. L. Luo, Y. Liu, R. Liu, S. Phatale, H. Lara, Y. Li, L. Shu, Y. Zhu, L. Meng, J. Sun, et al. Improve mathematical reasoning in language models by automated process supervision. arXiv preprint arXiv:2406.06592, 2024. find our card
  28. N. McAleese, R. M. Pokorny, J. F. C. Uribe, E. Nitishinskaya, M. Trebacz, and J. Leike. Llm critics help catch llm bugs. arXiv preprint arXiv:2407.00215, 2024. — [arXiv:2407.00215](https://arxiv.org/abs/2407.00215) find
  29. A. Meurer, C. P. Smith, M. Paprocki, O. Čertík, S. B. Kirpichev, M. Rocklin, A. Kumar, S. Ivanov, J. K. Moore, S. Singh, et al. Sympy: symbolic computing in python. PeerJ Computer Science, 3:e103, 2017. find
  30. R. Nakano, J. Hilton, S. Balaji, J. Wu, L. Ouyang, C. Kim, C. Hesse, S. Jain, V. Kosaraju, W. Saunders, et al. Webgpt: Browser-assisted question-answering with human feedback. arXiv preprint arXiv:2112.09332, 2021. find our card
  31. A. Pal, D. Karkhanis, S. Dooley, M. Roberts, S. Naidu, and C. White. Smaug: Fixing failure modes of preference optimisation with dpo-positive. arXiv preprint arXiv:2402.13228, 2024. find
  32. R. Y. Pang, W. Yuan, K. Cho, H. He, S. Sukhbaatar, and J. Weston. Iterative reasoning preference optimization. arXiv preprint arXiv:2404.19733, 2024. find our card
  33. R. Rafailov, A. Sharma, E. Mitchell, C. D. Manning, S. Ermon, and C. Finn. Direct preference optimization: Your language model is secretly a reward model. Advances in Neural Information Processing Systems, 36, 2024. find
  34. A. Roberts, H. W. Chung, A. Levskaya, G. Mishra, J. Bradbury, D. Andor, S. Narang, B. Lester, C. Gaffney, A. Mohiuddin, C. Hawthorne, A. Lewkowycz, A. Salcianu, M. van Zee, J. Austin, S. Goodman, L. B. Soares, H. Hu, S. Tsvyashchenko, A. Chowdhery, J. Bastings, J. Bulian, X. Garcia, J. Ni, A. Chen,… find
  35. W. Saunders, C. Yeh, J. Wu, S. Bills, L. Ouyang, J. Ward, and J. Leike. Self-critiquing models for assisting human evaluators. arXiv preprint arXiv:2206.05802, 2022. find our card
  36. T. Schick, J. Dwivedi-Yu, R. Dessì, R. Raileanu, M. Lomeli, E. Hambro, L. Zettlemoyer, N. Cancedda, and T. Scialom. Toolformer: Language models can teach themselves to use tools. Advances in Neural Information Processing Systems, 36, 2024. find our card
  37. A. Setlur, S. Garg, X. Geng, N. Garg, V. Smith, and A. Kumar. Rl on incorrect synthetic data scales the efficiency of llm math reasoning by eight-fold. arXiv preprint arXiv:2406.14532, 2024. find
  38. A. Singh, J. D. Co-Reyes, R. Agarwal, A. Anand, P. Patil, P. J. Liu, J. Harrison, J. Lee, K. Xu, A. Parisi, et al. Beyond human data: Scaling self-training for problem-solving with language models. arXiv preprint arXiv:2312.06585, 2023. find our card
  39. N. Stiennon, L. Ouyang, J. Wu, D. Ziegler, R. Lowe, C. Voss, A. Radford, D. Amodei, and P. F. Christiano. Learning to summarize with human feedback. Advances in Neural Information Processing Systems, 33: 3008–3021, 2020. find
  40. Z. Sun, L. Yu, Y. Shen, W. Liu, Y. Yang, S. Welleck, and C. Gan. Easy-to-hard generalization: Scalable alignment beyond human supervision. arXiv preprint arXiv:2403.09472, 2024. find
  41. M. Suzgun, N. Scales, N. Schärli, S. Gehrmann, Y. Tay, H. W. Chung, A. Chowdhery, Q. V. Le, E. H. Chi, D. Zhou, et al. Challenging big-bench tasks and whether chain-of-thought can solve them. arXiv preprint arXiv:2210.09261, 2022. find our card
  42. G. Team, M. Reid, N. Savinov, D. Teplyashin, T. Lillicrap, J.-b. Alayrac, R. Soricut, A. Lazaridou, O. Firat, J. Schrittwieser, et al. Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context. arXiv e-prints, pages arXiv–2403, 2024. find our card
  43. J. Uesato, N. Kushman, R. Kumar, F. Song, N. Siegel, L. Wang, A. Creswell, G. Irving, and I. Higgins. Solving math word problems with process-and outcome-based feedback. arXiv preprint arXiv:2211.14275, 2022. find our card
  44. P. Wang, L. Li, Z. Shao, R. Xu, D. Dai, Y. Li, D. Chen, Y. Wu, and Z. Sui. Math-shepherd: A label-free step-by-step verifier for llms in mathematical reasoning. arXiv preprint arXiv:2312.08935, 2023. find our card
  45. T. Wang, I. Kulikov, O. Golovneva, P. Yu, W. Yuan, J. Dwivedi-Yu, R. Y. Pang, M. Fazel-Zarandi, J. Weston, and X. Li. Self-taught evaluators. arXiv preprint arXiv:2408.02666, 2024. find
  46. X. Wang, J. Wei, D. Schuurmans, Q. Le, E. Chi, S. Narang, A. Chowdhery, and D. Zhou. Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171, 2022. find our card
  47. J. Wei, X. Wang, D. Schuurmans, M. Bosma, F. Xia, E. Chi, Q. V. Le, D. Zhou, et al. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824–24837, 2022. find our card
  48. M. Wortsman, P. J. Liu, L. Xiao, K. Everett, A. Alemi, B. Adlam, J. D. Co-Reyes, I. Gur, A. Kumar, R. Novak, et al. Small-scale proxies for large-scale transformer training instabilities. arXiv preprint arXiv:2309.14322, 2023. find
  49. T. Wu, W. Yuan, O. Golovneva, J. Xu, Y. Tian, J. Jiao, J. Weston, and S. Sukhbaatar. Meta-rewarding language models: Self-improving alignment with llm-as-a-meta-judge. arXiv preprint arXiv:2407.19594, 2024. find
  50. R. Yang, R. Ding, Y. Lin, H. Zhang, and T. Zhang. Regularizing hidden states enables learning generalizable reward model for llms. arXiv preprint arXiv:2406.10216, 2024. — [arXiv:2406.10216](https://arxiv.org/abs/2406.10216) find
  51. S. Yao, D. Yu, J. Zhao, I. Shafran, T. Griffiths, Y. Cao, and K. Narasimhan. Tree of thoughts: Deliberate problem solving with large language models. Advances in Neural Information Processing Systems, 36, 2024. find
  52. Z. Ye, F. Greenlee-Scott, M. Bartolo, P. Blunsom, J. A. Campos, and M. Gallé. Improving reward models with synthetic critiques. arXiv preprint arXiv:2405.20850, 2024. find
  53. F. Yu, A. Gao, and B. Wang. Ovm, outcome-supervised value models for planning in mathematical reasoning. In Findings of the Association for Computational Linguistics: NAACL 2024, pages 858–875, 2024. find
  54. W. Yuan, R. Y. Pang, K. Cho, S. Sukhbaatar, J. Xu, and J. Weston. Self-rewarding language models. arXiv preprint arXiv:2401.10020, 2024. find
  55. E. Zelikman, Y. Wu, J. Mu, and N. Goodman. Star: Bootstrapping reasoning with reasoning. Advances in Neural Information Processing Systems, 35:15476–15488, 2022. find our card
  56. E. Zhang, V. Zhu, N. Saphra, A. Kleiman, B. L. Edelman, M. Tambe, S. M. Kakade, and E. Malach. Transcendence: Generative models can outperform the experts that train them. arXiv preprint arXiv:2406.11741, 2024. find
  57. Y. Zhang, Y. Li, L. Cui, D. Cai, L. Liu, T. Fu, X. Huang, E. Zhao, Y. Zhang, Y. Chen, et al. Siren’s song in the ai ocean: a survey on hallucination in large language models. arXiv preprint arXiv:2309.01219, 2023. find our card
  58. L. Zheng, W.-L. Chiang, Y. Sheng, S. Zhuang, Z. Wu, Y. Zhuang, Z. Lin, Z. Li, D. Li, E. Xing, et al. Judging llm-as-a-judge with mt-bench and chatbot arena. Advances in Neural Information Processing Systems, 36, 2024. find