Elamur Research⌕
☰
home / papers / 2023 / DERA: Enhancing Large Language Model Completions with Dialog-Enabled Resolving Agents

DERA: Enhancing Large Language Model Completions with Dialog-Enabled Resolving Agents

Paper · 2023 · Varun Nair, Elliot Schumacher, Geoffrey Tso et al.

Summary

In clinical settings, LLM outputs must be both correct and complete. Yet single prompts, and even chained prompts, often invent facts or leave details out. The authors (Curai Health) propose DERA, a framework where two GPT-4 instances with separate roles discuss a task to refine the result. A Decider writes the first draft and owns the final answer. A Researcher reviews the source material or the question and points out errors or gaps, which the Decider may accept or reject.

DERA is tested on three clinical tasks: structured summaries of doctor–patient chats, care plan generation, and MedQA questions answered without multiple-choice options. For summaries and care plans, physicians preferred DERA outputs over plain GPT-4, and automatic metrics improved, with fewer hallucinations and omissions. On open-ended QA, DERA added little or nothing, although GPT-4 alone already scored around 70%.

The work matters because a dialog between role-scoped agents offers an iterative, auditable way to fix long-form LLM text in high-stakes domains.

Key points

  • Two agents, one model: both the Researcher and the Decider are played by GPT-4. In summarization, accepted fixes go to a shared scratchpad, and the dialog is capped at 15 turns.
  • Summarization data: 500 de-identified telehealth chats, averaging 27 turns and 646 tokens each, summarized into six sections (e.g., Pertinent Positives/Negatives/Unknowns). Four licensed physicians rated a subset of 50.
  • Open-ended MedQA (USMLE): GPT-4 reaches about 70% against a roughly 60% pass mark, with DERA performing similarly. The authors release this open-ended version of MedQA on GitHub (curai/curai-research).
  • Where it helps: gains appear in long-form generation. Short single-answer QA suffers because DERA tends to add detail, and answers often miss the right level of specificity.
  • Limitations: unknown GPT-4 training data, model versions that are not persisted, data from a single telehealth service, and weak automatic evaluation metrics.

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, topics and concepts

📚 Shelves 6The 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 worksMedicine · 241 worksHealth Informatics · 184 works

🧭 Topics 1Research subjects and languages the work is about.Dialogue systems · 624 works

Cite

Cite as Varun Nair, Elliot Schumacher, Geoffrey Tso and Anitha Kannan. 2023. DERA: Enhancing Large Language Model Completions with Dialog-Enabled Resolving Agents. arXiv:2303.17071.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2303.17071 · arXiv 2303.17071

@misc{nair2023dera,
  title = {{DERA: Enhancing Large Language Model Completions with Dialog-Enabled Resolving Agents}},
  author = {Varun Nair, Elliot Schumacher, Geoffrey Tso and Anitha Kannan},
  year = {2023},
  doi = {10.48550/arxiv.2303.17071},
  eprint = {2303.17071},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2303.17071}
}

References

28 references, 15 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Monica Agrawal, Stefan Hegselmann, Hunter Lang, Yoon Kim, and David Sontag. 2022. Large language models are zero-shot clinical information extractors. arXiv preprint arXiv:2205.12689.
  2. Alan R Aronson. 2001. Effective mapping of biomedical text to the umls metathesaurus: the metamap program. In Proceedings of the AMIA Symposium, page 17. American Medical Informatics Association. find
  3. Republic of Korea. Association for Computational Linguistics. find
  4. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901.
  5. Bharath Chintagunta, Namit Katariya, Xavier Amatriain, and Anitha Kannan. 2021. Medically aware GPT-3 as a data generator for medical dialogue summarization. In Proceedings of the Second Workshop on Natural Language Processing for Medical Conversations, pages 66–76, Online. Association for Computati… find
  6. Aakanksha Chowdhery, Sharan Narang, Jacob Devlin, Maarten Bosma, Gaurav Mishra, Adam Roberts, Paul Barham, Hyung Won Chung, Charles Sutton, Sebastian Gehrmann, et al. 2022. Palm: Scaling language modeling with pathways. arXiv preprint arXiv:2204.02311.
  7. Nouha Dziri, Sivan Milton, Mo Yu, Osmar Zaiane, and Siva Reddy. 2022. On the origin of hallucinations in conversational models: Is it the datasets or the models? In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Techn…
  8. Seppo Enarvi, Marilisa Amoia, Miguel Del-Agua Teba, Brian Delaney, Frank Diehl, Stefan Hahn, Kristina Harris, Liam McGrath, Yue Pan, Joel Pinto, Luca Rubini, Miguel Ruiz, Gagandeep Singh, Fabian Stemmer, Weiyi Sun, Paul Vozila, Thomas Lin, and Ranjani Ramamurthy. 2020. Generating medical reports fro… find
  9. Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Veselin Stoyanov, and Luke Zettlemoyer. 2020. BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. In Proceedings of the 58th Annual Meeting of the…
  10. Valentin Liévin, Christoffer Egeberg Hother, and Ole Winther. 2022. Can large language models reason about medical questions? arXiv preprint arXiv:2207.08143. find
  11. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692.
  12. Joshua Maynez, Shashi Narayan, Bernd Bohnet, and Ryan McDonald. 2020. On faithfulness and factuality in abstractive summarization. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 1906–1919, Online. Association for Computational Linguistics.
  13. Bridget T McInnes, Ted Pedersen, and Serguei VS Pakhomov. 2009. Umls-interface and umls-similarity: open source software for measuring paths and semantic similarity. In AMIA annual symposium proceedings, volume 2009, page 431. American Medical Informatics Association. find
  14. Sewon Min, Xinxi Lyu, Ari Holtzman, Mikel Artetxe, Mike Lewis, Hannaneh Hajishirzi, and Luke Zettlemoyer. 2022. Rethinking the role of demonstrations: What makes in-context learning work?
  15. Harsha Nori, Nicholas King, Scott Mayer McKinney, Dean Carignan, and Eric Horvitz. 2023. Capabilities of GPT-4 on Medical Challenge Problems.
  16. OpenAI. 2023. Gpt-4 technical report.
  17. Michael Hanna and Ondřej Bojar. 2021. A fine-grained analysis of BERTScore. In Proceedings of the Sixth Conference on Machine Translation, pages 507–517, Online. Association for Computational Linguistics.
  18. Laria Reynolds and Kyle McDonell. 2021. Prompt programming for large language models: Beyond the few-shot paradigm. In Extended Abstracts of the 2021 CHI Conference on Human Factors in Computing Systems, pages 1–7.
  19. 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. Applied Sciences, 11(14):6421.
  20. Karan Singhal, Shekoofeh Azizi, Tao Tu, S Sara Mahdavi, Jason Wei, Hyung Won Chung, Nathan Scales, Ajay Tanwani, Heather Cole-Lewis, Stephen Pfohl, et al. 2022. Large language models encode clinical knowledge. arXiv preprint arXiv:2212.13138.
  21. Anirudh Joshi, Namit Katariya, Xavier Amatriain, and Anitha Kannan. 2020. Dr. summarize: Global summarization of medical dialogue by exploiting local structures. In Findings of the Association for Computational Linguistics: EMNLP 2020, pages 3755– 3763, Online. Association for Computational Linguist… find
  22. Tianxiang Sun, Junliang He, Xipeng Qiu, and Xuanjing Huang. 2022. BERTScore is unfair: On social bias in language model-based metrics for text generation. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 3726–3739, Abu Dhabi, United Arab Emirates. Asso… find
  23. Oyvind Tafjord, Bhavana Dalvi Mishra, and Peter Clark. 2022. Entailer: Answering questions with faithful and truthful chains of reasoning. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 2078–2093, Abu Dhabi, United Arab Emirates. Association for Comp… find
  24. Boshi Wang, Xiang Deng, and Huan Sun. 2022. Iteratively prompt pre-trained language models for chain of thought. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 2714–2730, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics. find
  25. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed Chi, Quoc Le, and Denny Zhou. 2022. Chain of thought prompting elicits reasoning in large language models. arXiv preprint arXiv:2201.11903.
  26. Tongshuang Wu, Michael Terry, and Carrie Jun Cai. 2022. Ai chains: Transparent and controllable human-ai interaction by chaining large language model prompts. In Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems, CHI ’22, New York, NY, USA. Association for Computing Machin… find
  27. Longxiang Zhang, Renato Negrinho, Arindam Ghosh, Vasudevan Jagannathan, Hamid Reza Hassanzadeh, Thomas Schaaf, and Matthew R Gormley. 2021. Leveraging pretrained models for automatic summarization of doctor-patient conversations. arXiv preprint arXiv:2109.12174. find
  28. Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q Weinberger, and Yoav Artzi. 2019. Bertscore: Evaluating text generation with bert. arXiv preprint arXiv:1904.09675. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/nair-2023-dera-enhancing-large-language-model/