--- title: "Reiichiro Nakano — author profile and papers" description: "Reiichiro Nakano: 2 papers on Elamur Research (2021) — affiliations, topics, co-authors and who cites them in our library." canonical: "https://research.elamur.ai/free/en/authors/reiichiro-nakano/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/authors/reiichiro-nakano/) [home](https://md.research.elamur.ai/free/en/) / [authors](https://md.research.elamur.ai/free/en/authors/) / Reiichiro Nakano # Reiichiro Nakano Reiichiro Nakano works on large language models and on ways to make what they produce more reliable. Both papers here date from 2021. They share two concerns. One is multi-step reasoning, where a single slip can derail an answer. The other is open-ended [question answering](https://md.research.elamur.ai/free/en/topics/question-answering/), where answers should rest on cited sources. The methods pair finetuned GPT-3 models with separately trained scoring components, human judgements and reinforcement learning. The more widely cited paper, [Training Verifiers to Solve Math Word Problems](https://md.research.elamur.ai/free/en/papers/cobbe-2021-training-verifiers-solve-math-word/), introduced GSM8K. This is a collection of roughly 8.5K grade-school word problems written by people, each paired with a worked solution in ordinary language. The paper also proposed drawing many candidate solutions from a generator and letting a trained verifier choose the most promising one, a contribution to work on [reasoning](https://md.research.elamur.ai/free/en/topics/reasoning/). The second paper, [WebGPT](https://md.research.elamur.ai/free/en/papers/nakano-2021-webgpt-browser-assisted-question-answering/), had GPT-3 navigate a stripped-down, text-based browser. The model ran searches, followed links and quoted passages that then served as references in paragraph-length answers. It was trained from human demonstrations and pairwise answer preferences, then refined with reinforcement learning and rejection sampling. Both works appeared in 2021 and point the same way: using learned evaluators and human feedback to steer language model output, whether checking chains of arithmetic or judging which web evidence supports an answer. Nakano's papers here are collaborative. [Karl Cobbe](https://md.research.elamur.ai/free/en/authors/karl-cobbe/), [Jacob Hilton](https://md.research.elamur.ai/free/en/authors/jacob-hilton/), [Christopher Hesse](https://md.research.elamur.ai/free/en/authors/christopher-hesse/) and [Vineet Kosaraju](https://md.research.elamur.ai/free/en/authors/vineet-kosaraju/) appear on both, and John Schulman and Long Ouyang are among the other co-authors. Reiichiro Nakano has 2 papers in our library, in 2021, most of them on [Question answering](https://md.research.elamur.ai/free/en/topics/question-answering/), [Reasoning](https://md.research.elamur.ai/free/en/topics/reasoning/); often with [Christopher Hesse](https://md.research.elamur.ai/free/en/authors/christopher-hesse/), [Jacob Hilton](https://md.research.elamur.ai/free/en/authors/jacob-hilton/), [Karl Cobbe](https://md.research.elamur.ai/free/en/authors/karl-cobbe/). 2 papers in our library · 2021 No OpenAlex profile matched yet. [2 works in our library](https://md.research.elamur.ai/free/en/authors/reiichiro-nakano/#papers) · [4 306 citations in our library](https://md.research.elamur.ai/free/en/authors/reiichiro-nakano/cited-by/) · [⭐ Paper Star rank #595](https://md.research.elamur.ai/free/en/paper-stars/page/6/#star-595) The profile is the library's match of the name; two people with one name are not always told apart. Elsewhere: [Google Scholar](https://scholar.google.com/citations?view_op=search_authors&mauthors=Reiichiro+Nakano) · [Semantic Scholar](https://www.semanticscholar.org/search?q=Reiichiro+Nakano) · [dblp](https://dblp.org/search?q=Reiichiro+Nakano) · [ACL Anthology](https://aclanthology.org/search/?q=Reiichiro+Nakano) ## Topics [Question answering 1](https://md.research.elamur.ai/free/en/topics/question-answering/) · [Reasoning 1](https://md.research.elamur.ai/free/en/topics/reasoning/) ## Works by year ### 2020s [**2021** 2 papers](https://md.research.elamur.ai/free/en/authors/reiichiro-nakano/year/2021/) ## Papers The most cited first (OpenAlex and our library). 1. **[Training Verifiers to Solve Math Word Problems](https://md.research.elamur.ai/free/en/papers/cobbe-2021-training-verifiers-solve-math-word/)** *(2021 · 3 675 citations)* Large language models still make serious mistakes when a math problem takes several reasoning steps. An autoregressive generator can't go back and fix an error once its solution goes wrong. The OpenAI authors tackle this in two ways. First, they release GSM8K, a set of 8.5K grade-school math word problems written by… 2. **[WebGPT: Browser-assisted question-answering with human feedback](https://md.research.elamur.ai/free/en/papers/nakano-2021-webgpt-browser-assisted-question-answering/)** *(2021 · 687 citations)* Long-form question answering (LFQA) asks a system to write a paragraph-length reply to an open-ended question, and such systems still fall short of human quality. ## Co-authors here 12 of 14 co-authors with a page, the most papers together first. - [Christopher Hesse](https://md.research.elamur.ai/free/en/authors/christopher-hesse/) ×2 - [Jacob Hilton](https://md.research.elamur.ai/free/en/authors/jacob-hilton/) ×2 - [Karl Cobbe](https://md.research.elamur.ai/free/en/authors/karl-cobbe/) ×2 - [Vineet Kosaraju](https://md.research.elamur.ai/free/en/authors/vineet-kosaraju/) ×2 - [Heewoo Jun](https://md.research.elamur.ai/free/en/authors/heewoo-jun/) - [Jeff Wu](https://md.research.elamur.ai/free/en/authors/jeff-wu/) - [Jerry Tworek](https://md.research.elamur.ai/free/en/authors/jerry-tworek/) - [John Schulman](https://md.research.elamur.ai/free/en/authors/john-schulman/) - [Long Ouyang](https://md.research.elamur.ai/free/en/authors/long-ouyang/) - [Lukasz Kaiser](https://md.research.elamur.ai/free/en/authors/lukasz-kaiser/) - [Mark Chen](https://md.research.elamur.ai/free/en/authors/mark-chen/) - [Mohammad Bavarian](https://md.research.elamur.ai/free/en/authors/mohammad-bavarian/) ## Cited by Papers of our library cite their works 4 306 times (of the 140 884 papers whose bibliographies we read); [334 citing papers with a page here →](https://md.research.elamur.ai/free/en/authors/reiichiro-nakano/cited-by/).