Elamur Research
☰
home / papers / 2023 / Fine-Grained Human Feedback Gives Better Rewards for Language Model Training

Fine-Grained Human Feedback Gives Better Rewards for Language Model Training

Written for the Research library (claude-opus-5-5) from the paper text, in our own words: the license "arXiv — non-exclusive license to distribute" does not allow redistributing the text itself. Original — arxiv.org/pdf/2306.01693

Summary

Standard RLHF turns a human's overall preference between whole outputs into one scalar reward at the end of the sequence. For long generations this signal is sparse and hard to interpret, because it doesn't say which part of the text failed or why. The authors propose FINE-GRAINED RLHF, which makes rewards finer in two ways. First, rewards are dense: they are given per segment, such as a sentence or sub-sentence. Second, they are split by error type, with a separate reward model for each kind of problem. The weighted per-segment rewards and a KL penalty are combined and fed into PPO. The framework is tested on two tasks. For detoxification, PERSPECTIVE scores each sentence. For long-form QA, the authors built a new dataset, QA-FEEDBACK, in which annotators marked spans with specific error categories. Both tasks improve over a holistic reward, according to automatic and human evaluation. Toxicity also drops with fewer training steps while fluency is preserved. Changing the reward-model weights lets users steer the model's behaviour. This makes RLHF easier to interpret and to control.

Key points

  • Feedback in QA-FEEDBACK comes at three levels: irrelevance, repetition or incoherence (sub-sentence); false or unverifiable facts (sentence); and missing information (whole answer).
  • The detoxification experiments use REALTOXICITYPROMPTS (about 100K web prompts) with the PERSPECTIVE API called after each generated sentence.
  • Annotation costs about the same as preference labels: roughly 6 minutes per long-form QA sample for either kind of feedback.
  • Main costs: more compute (several API calls or reward models per output), manual work to design feedback types for each task, and reliance on carefully controlled annotation quality.
  • Data, human feedback and code are released at FineGrainedRLHF.github.io (NeurIPS 2023).

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: Other

Cited by

178 papers of our library cite it (of the 140673 whose bibliographies we read); 11 of them have a page here, 167 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2306.01693 · arXiv 2306.01693

@misc{wu2023fine,
  title = {{Fine-Grained Human Feedback Gives Better Rewards for Language Model Training}},
  author = {Wu, Zeqiu and Hu, Yushi and Shi, Weijia and Dziri, Nouha and Suhr, Alane and Ammanabrolu, Prithviraj and Smith, Noah A. and Ostendorf, Mari and Hajishirzi, Hannaneh},
  year = {2023},
  doi = {10.48550/arxiv.2306.01693},
  eprint = {2306.01693},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2306.01693}
}

References

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

  1. Perspective API, https://github.com/conversationai/perspectiveapi. ↗
  2. Yuntao Bai, Andy Jones, Kamal Ndousse, Amanda Askell, Anna Chen, Nova DasSarma, Dawn Drain, Stanislav Fort, Deep Ganguli, Tom Henighan, et al. Training a helpful and harmless assistant with reinforcement learning from human feedback. arXiv preprint arXiv:2204.05862, 2022. find
  3. Yuntao Bai, Saurav Kadavath, Sandipan Kundu, Amanda Askell, Jackson Kernion, Andy Jones, Anna Chen, Anna Goldie, Azalia Mirhoseini, Cameron McKinnon, et al. Constitutional ai: Harmlessness from ai feedback. arXiv preprint arXiv:2212.08073, 2022. find our card
  4. Iz Beltagy, Matthew E Peters, and Arman Cohan. Longformer: The long-document transformer. arXiv preprint arXiv:2004.05150, 2020. find
  5. Angelica Chen, Jérémy Scheurer, Tomasz Korbak, Jon Ander Campos, Jun Shern Chan, Samuel R Bowman, Kyunghyun Cho, and Ethan Perez. Improving code generation by training with natural language feedback. arXiv preprint arXiv:2303.16749, 2023. find
  6. Bhavana Dalvi Mishra, Oyvind Tafjord, and Peter Clark. Towards teachable reasoning systems: Using a dynamic memory of user feedback for continual system improvement. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 9465–9480, Abu Dhabi, United Arab Emi… find
  7. Sumanth Dathathri, Andrea Madotto, Janice Lan, Jane Hung, Eric Frank, Piero Molino, Jason Yosinski, and Rosanne Liu. Plug and play language models: A simple approach to controlled text generation. In International Conference on Learning Representations, 2020. find our card
  8. Yao Dou, Maxwell Forbes, Rik Koncel-Kedziorski, Noah A. Smith, and Yejin Choi. Is GPT-3 text indistinguishable from human text? scarecrow: A framework for scrutinizing machine text. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pa… find
  9. Yao Fu, Litu Ou, Mingyu Chen, Yuhao Wan, Hao Peng, and Tushar Khot. Chain-of-thought hub: A continuous effort to measure large language models’ reasoning performance. arXiv preprint arXiv:2305.17306, 2023. find
  10. Deep Ganguli, Amanda Askell, Nicholas Schiefer, Thomas Liao, Kamilė Lukošiūtė, Anna Chen, Anna Goldie, Azalia Mirhoseini, Catherine Olsson, Danny Hernandez, et al. The capacity for moral self-correction in large language models. arXiv preprint arXiv:2302.07459, 2023. find our card
  11. Tianyu Gao, Howard Yen, Jiatong Yu, and Danqi Chen. Enabling large language models to generate text with citations. arXiv preprint arXiv:2305.14627, 2023. find our card
  12. Samuel Gehman, Suchin Gururangan, Maarten Sap, Yejin Choi, and Noah A. Smith. Real-ToxicityPrompts: Evaluating neural toxic degeneration in language models. In Findings of the Association for Computational Linguistics: EMNLP 2020, pages 3356–3369, Online, November 2020. Association for Computational… find
  13. Amelia Glaese, Nat McAleese, Maja Tr˛ebacz, John Aslanides, Vlad Firoiu, Timo Ewalds, Maribeth Rauh, Laura Weidinger, Martin Chadwick, Phoebe Thacker, et al. Improving alignment of dialogue agents via targeted human judgements. arXiv preprint arXiv:2209.14375, 2022. find our card
  14. Braden Hancock, Antoine Bordes, Pierre-Emmanuel Mazare, and Jason Weston. Learning from dialogue after deployment: Feed yourself, chatbot! In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3667–3684, Florence, Italy, July 2019. Association for Computat… find
  15. Matthew Honnibal, Ines Montani, Sofie Van Landeghem, and Adriane Boyd. spaCy: Industrial-strength Natural Language Processing in Python. 2020. find
  16. Tomasz Korbak, Kejian Shi, Angelica Chen, Rasika Bhalerao, Christopher L Buckley, Jason Phang, Samuel R Bowman, and Ethan Perez. Pretraining language models with human preferences. arXiv preprint arXiv:2302.08582, 2023. find
  17. Ben Krause, Akhilesh Deepak Gotmare, Bryan McCann, Nitish Shirish Keskar, Shafiq Joty, Richard Socher, and Nazneen Fatema Rajani. GeDi: Generative discriminator guided sequence generation. In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 4929–4952, Punta Cana, Dominica… find our card
  18. Kalpesh Krishna, Erin Bransom, Bailey Kuehl, Mohit Iyyer, Pradeep Dasigi, Arman Cohan, and Kyle Lo. LongEval: Guidelines for human evaluation of faithfulness in long-form summarization. In Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, pa… find
  19. Margaret Li, Jason Weston, and Stephen Roller. Acute-eval: Improved dialogue evaluation with optimized questions and multi-turn comparisons. arXiv preprint arXiv:1909.03087, 2019. — [arXiv:1909.03087](https://arxiv.org/abs/1909.03087) find
  20. Chin-Yew Lin. ROUGE: A package for automatic evaluation of summaries. In Text Summarization Branches Out, pages 74–81, Barcelona, Spain, July 2004. Association for Computational Linguistics. find
  21. Alisa Liu, Maarten Sap, Ximing Lu, Swabha Swayamdipta, Chandra Bhagavatula, Noah A. Smith, and Yejin Choi. DExperts: Decoding-time controlled text generation with experts and anti-experts. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th Interna… find our card
  22. Ximing Lu, Sean Welleck, Liwei Jiang, Jack Hessel, Lianhui Qin, Peter West, Prithviraj Ammanabrolu, and Yejin Choi. Quark: Controllable text generation with reinforced unlearning. In Thirty-sixth Conference on Neural Information Processing Systems (NeurIPS), 2022. find our card
  23. Aman Madaan, Niket Tandon, Peter Clark, and Yiming Yang. Memory-assisted prompt editing to improve GPT-3 after deployment. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 2833–2861, Abu Dhabi, United Arab Emirates, December 2022. Association for Compu… find
  24. Jacob Menick, Maja Trebacz, Vladimir Mikulik, John Aslanides, Francis Song, Martin Chadwick, Mia Glaese, Susannah Young, Lucy Campbell-Gillingham, Geoffrey Irving, et al. Teaching language models to support answers with verified quotes. arXiv preprint arXiv:2203.11147, 2022. find our card
  25. Sewon Min, Kalpesh Krishna, Xinxi Lyu, Mike Lewis, Wen-tau Yih, Pang Wei Koh, Mohit Iyyer, Luke Zettlemoyer, and Hannaneh Hajishirzi. Factscore: Fine-grained atomic evaluation of factual precision in long form text generation. arXiv preprint arXiv:2305.14251, 2023. find our card
  26. Sewon Min, Julian Michael, Hannaneh Hajishirzi, and Luke Zettlemoyer. AmbigQA: Answering ambiguous open-domain questions. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 5783–5797, Online, November 2020. Association for Computational Linguisti… find
  27. Reiichiro Nakano, Jacob Hilton, S. Arun Balaji, Jeff Wu, Long Ouyang, Christina Kim, Christopher Hesse, Shantanu Jain, Vineet Kosaraju, William Saunders, Xu Jiang, Karl Cobbe, Tyna Eloundou, Gretchen Krueger, Kevin Button, Matthew Knight, Benjamin Chess, and John Schulman. Webgpt: Browser-assisted q… find our card
  28. OpenAI. Gpt-4 technical report, 2023. find our card
  29. Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul Christiano, Jan Leike, and Ryan Lowe. Training la… find
  30. Debjit Paul, Mete Ismayilzada, Maxime Peyrard, Beatriz Borges, Antoine Bosselut, Robert West, and Boi Faltings. Refiner: Reasoning feedback on intermediate representations, 2023. find our card
  31. Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. Language models are unsupervised multitask learners. 2019. find
  32. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. Exploring the limits of transfer learning with a unified text-to-text transformer. The Journal of Machine Learning Research, 21(1):5485–5551, 2020. find
  33. Rajkumar Ramamurthy*, Prithviraj Ammanabrolu*, Kianté Brantley, Jack Hessel, Rafet Sifa, Christian Bauckhage, Hannaneh Hajishirzi, and Yejin Choi. Is reinforcement learning (not) for natural language processing: Benchmarks, baselines, and building blocks for natural language policy optimization. In… find our card
  34. Jérémy Scheurer, Jon Ander Campos, Jun Shern Chan, Angelica Chen, Kyunghyun Cho, and Ethan Perez. Training language models with natural language feedback. arXiv preprint arXiv:2204.14146, 2022. — [arXiv:2204.14146](https://arxiv.org/abs/2204.14146) find
  35. Jérémy Scheurer, Jon Ander Campos, Tomasz Korbak, Jun Shern Chan, Angelica Chen, Kyunghyun Cho, and Ethan Perez. Training language models with language feedback at scale, 2023. find
  36. John Schulman, Philipp Moritz, Sergey Levine, Michael Jordan, and Pieter Abbeel. High-dimensional continuous control using generalized advantage estimation. In Proceedings of the International Conference on Learning Representations (ICLR), 2016. find
  37. John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347, 2017. find
  38. Weiyan Shi, Emily Dinan, Kurt Shuster, Jason Weston, and Jing Xu. When life gives you lemons, make cherryade: Converting feedback from bad responses into good labels. arXiv preprint arXiv:2210.15893, 2022. find
  39. Ivan Stelmakh, Yi Luan, Bhuwan Dhingra, and Ming-Wei Chang. ASQA: Factoid questions meet long-form answers. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 8273–8288, Abu Dhabi, United Arab Emirates, December 2022. Association for Computational Lingui… find our card
  40. Nisan Stiennon, Long Ouyang, Jeff Wu, Daniel M. Ziegler, Ryan J. Lowe, Chelsea Voss, Alec Radford, Dario Amodei, and Paul Christiano. Learning to summarize from human feedback. ArXiv, abs/2009.01325, 2020. find
  41. Jeff Wu, Long Ouyang, Daniel M Ziegler, Nisan Stiennon, Ryan Lowe, Jan Leike, and Paul Christiano. Recursively summarizing books with human feedback. arXiv preprint arXiv:2109.10862, 2021. find our card
  42. Jing Xu, Megan Ung, Mojtaba Komeili, Kushal Arora, Y-Lan Boureau, and Jason Weston. Learning new skills after deployment: Improving open-domain internet-driven dialogue with human feedback. arXiv preprint arXiv:2208.03270, 2022. find
  43. Wenda Xu, Danqing Wang, Liangming Pan, Zhenqiao Song, Markus Freitag, William Yang Wang, and Lei Li. Instructscore: Towards explainable text generation evaluation with automatic feedback. arXiv preprint arXiv:2305.14282, 2023. find
  44. Zheng Yuan, Hongyi Yuan, Chuanqi Tan, Wei Wang, Songfang Huang, and Fei Huang. Rrhf: Rank responses to align language models with human feedback without tears, 2023. find our card
  45. Xiang Yue, Boshi Wang, Kai Zhang, Ziru Chen, Yu Su, and Huan Sun. Automatic evaluation of attribution by large language models. arXiv preprint arXiv:2305.06311, 2023. find
  46. Daniel M. Ziegler, Nisan Stiennon, Jeff Wu, Tom B. Brown, Alec Radford, Dario Amodei, Paul Christiano, and Geoffrey Irving. Fine-tuning language models from human preferences. ArXiv, abs/1909.08593, 2019. find