Valentina Pyatkin
Researcher at the ETH AI Center working on LLM post-training; core contributor to Tülu 3
- Researcher working on LLM post-training for the Swiss AI Initiative, ETH AI Center1
Valentina Pyatkin is a researcher at the ETH AI Center, working on post-training of large language models for the Swiss AI Initiative and contributing to Apertus. Pyatkin was previously a postdoctoral researcher at the Allen Institute for AI (Ai2) and the University of Washington, where Pyatkin worked on open post-training efforts including Tülu 3 and contributed to OLMo. Pyatkin is the third author of the Tülu 3 report and has built benchmarks such as RewardBench and IFBench.12
Career
Pyatkin completed a PhD in computer science at the NLP lab of Bar-Ilan University, supervised by Ido Dagan and Reut Tsarfaty, and interned twice at Ai2. Pyatkin then held a postdoctoral position at Ai2 and the University of Washington, advised by Hannaneh Hajishirzi and Yejin Choi.1
Pyatkin is now a researcher at the ETH AI Center, working on post-training of language models for the Swiss AI Initiative and contributing to the Apertus models.1
Research
Pyatkin's research centres on post-training and adaptation of language models, including how models follow instructions and constraints and how preference and reward signals shape their behaviour. At Ai2, Pyatkin worked on the Tülu post-training pipelines, which combine supervised finetuning, direct preference optimization and reinforcement learning with verifiable rewards, and contributed to OLMo and OLMo 2.12
Pyatkin has also built evaluation benchmarks, including RewardBench and IFBench, and co-authored RewardBench 2, published at ICLR 2026.1
Work
Llama 3.1 Tülu 3.1 8BOpen weights
Third author of the Tülu 3 post-training report, which the Tülu 3.1 8B model card cites as its paper; Pyatkin describes being a core contributor to the Tülu and Open-Instruct project.231
IF-RLVR: Generalizing Verifiable Instruction Following (external site: valentinapy.github.io)Published research
First author of a NeurIPS 2025 paper on generalizing verifiable instruction following, with co-authors including Pradeep Dasigi and Nathan Lambert.1
Career timeline
Jan 2023
Received the AI2 Outstanding Intern of the Year Award.1
Aug 2024
Received the ACL Best Theme Paper Award for OLMo and an ACL Outstanding Paper Award.1
Nov 2024
Co-authored the Tülu 3 report on open language model post-training as third author.2
Dec 2025
Had the IFBench paper accepted to NeurIPS 2025.1
2026
Co-authored RewardBench 2, published at ICLR 2026.1
Other documented roles
Roles the sources document but that are not confirmed as current.
- Postdoctoral researcher and Young Investigator (past), Allen Institute for AI (Ai2) and University of Washington1
Honors and recognition
- AI2 Outstanding Intern of the Year Award · Allen Institute for AI, 20231
- DAAD AInet Fellow · German Academic Exchange Service (DAAD), 20231
- ACL Best Theme Paper Award (for OLMo) · Association for Computational Linguistics, 20241
- ACL Outstanding Paper Award · Association for Computational Linguistics, 20241
Related people
More sourced profiles at the same organization, in the same field, or from the same school.
- Akshita BhagiaAlso at Ai2
- Adina WilliamsEvaluation and benchmarks
- Ali FarhadiAlso at Ai2
- Christopher ClarkAlso at Allen Institute for AI
- AI ExplainedEvaluation and benchmarks
- Aylin CaliskanAlso from UW
- David WaddenAlso at Ai2
- Collin BurnsLanguage models
- Byron BootsAlso from UW
- Dieter FoxAlso at Ai2
- Diogo AlmeidaLanguage models
- Carlos GuestrinAlso from UW
Is this you?
Profiles cover public professional work only, and every line cites a source. If something is out of date, missing or wrong, send a correction with a link to an official page and we will update it. Suggest a correction
Add this badge to your website or GitHub profile README. It links back to this page. It is not an endorsement or a rating.
Markdown (for a README)
[](https://unitedstatesofamericasuperintelligence.com/people/valentina-pyatkin/)
HTML
<a href="https://unitedstatesofamericasuperintelligence.com/people/valentina-pyatkin/"><img src="https://unitedstatesofamericasuperintelligence.com/badges/profile.svg" alt="Valentina Pyatkin — Profile on USASI" height="20"></a>
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.