Nathan Lambert

Career

Lambert was a senior research scientist and post-training lead at the Allen Institute for AI (Ai2), where Lambert trained the open Olmo language models. In October 2026 Lambert announced Trillium Labs, a nonprofit co-founded to support open science of frontier AI, with Lambert as Executive Director.²¹¹⁰

Research

Lambert's research centres on post-training of open language models. Lambert co-authored the Zephyr alignment paper (2023) and Tulu 2 (2023), was first author of RewardBench (2024), a benchmark for evaluating reward models, and first author of the Tülu 3 post-training report (2024), which released model weights, data, training code and the full recipe.⁸¹¹¹²³

Lambert is also a co-author of the OLMo (2024) and OLMoE (2024) papers on fully open language models, and is listed among the authors of Hugging Face's TRL post-training library and the Diffusers library.⁴⁵⁶⁷

Teaching and writing

Lambert writes the Interconnects newsletter, which grew into a weekly publication after the release of ChatGPT and covers RLHF, reasoning models and open models, and hosts the Interconnects Interviews series. Lambert's book Reinforcement Learning from Human Feedback introduces post-training methods organized around the standard RLHF recipe, and is accompanied by a post-training course.²¹⁹

Work and contributions

• Tülu 3 Open weights

  First author and a marked core contributor of "Tulu 3: Pushing Frontiers in Open Language Model Post-Training" (2024), written with an Allen Institute for AI affiliation. The abstract says the release includes the Tülu 3 model weights along with datasets, training code, and the full recipe.³¹³

• OLMo and OLMoE Published research

  Co-author of "OLMo: Accelerating the Science of Language Models" (2024) and "OLMoE: Open Mixture-of-Experts Language Models" (2024).⁴⁵

• TRL (Transformers Reinforcement Learning) Open-source code

  Listed among the authors in TRL's citation file; TRL is an Apache-2.0 post-training toolkit for transformer models.⁶

• Diffusers Open-source code

  Listed among the authors in the Diffusers README citation (2022).⁷

• Reinforcement Learning from Human Feedback (book) Published research

  Sole author of the book, an introduction to RLHF and post-training methods for language models, available on the web and as a Manning print edition.⁹

• Llama 3.1 Tülu 3.1 8B Open weights

  First author of the Tülu 3 paper, which the Llama 3.1 Tülu 3.1 8B model card cites.³

• Olmo 3 7B Instruct Open weights

  Credited in the Olmo 3 paper's author contributions for technical leadership and for post-training infrastructure, methodology, and data work; the model card cites that paper.¹⁴

Timeline

1. Oct 25, 2023Co-authored "Zephyr - Direct Distillation of LM Alignment".⁸
2. Nov 17, 2023Co-authored the Tulu 2 paper on adapting language models.¹¹
3. Feb 1, 2024Co-authored "OLMo - Accelerating the Science of Language Models".⁴
4. Mar 20, 2024Released RewardBench, a benchmark for reward models, as first author.¹²
5. Sep 3, 2024Co-authored "OLMoE - Open Mixture-of-Experts Language Models".⁵
6. Nov 22, 2024Released the Tülu 3 post-training report as first author.³
7. 2026Completed the book Reinforcement Learning from Human Feedback, published in print by Manning.¹⁹
8. Oct 1, 2026Announced Trillium Labs, a nonprofit for the open science of frontier AI, as co-founder and Executive Director.¹⁰¹

Other documented roles

Roles the sources document that are not confirmed as current.

• Senior research scientist and post-training lead, Allen Institute for AI (Ai2)²¹³

Where they work

• Trillium Labs 1 person in the wiki

Related people

• Lilian WengTeaching and writing · Language models
• Denny BritzLanguage models · Reinforcement learning
• Abigail SeeLanguage models · Teaching and writing
• Zvi MowshowitzTeaching and writing · Language models
• Younes BelkadaLanguage models · Reinforcement learning
• Dwarkesh PatelTeaching and writing · Language models
• Percy LiangLanguage models · Teaching and writing
• Albert JiangLanguage models · Reinforcement learning
• Susan ZhangLanguage models · Reinforcement learning
• Weizhu ChenLanguage models · Reinforcement learning