Valentina Pyatkin (@valentina_

Valentina Pyatkin

834 posts

Valentina Pyatkin

@valentina__py

Postdoc at the Allen Institute for AI @allen_ai , @uwnlp , and @ETH_AI_Center

🇨🇭 Zürich

valentinapy.github.io

Joined October 2016

Valentina Pyatkin
@valentina__py
Oct 15, 2024
📣Research internship hiring at Ai2 is now open! We're hiring research interns at Ai2 to help at every stage of our Open Language Model (OLMo) workstream and we want people who can contribute meaningfully to our best models! Preferred deadline for consideration: Nov. 3rd, 2024
47K
Valentina Pyatkin
@valentina__py
Jul 3, 2025
💡Beyond math/code, instruction following with verifiable constraints is suitable to be learned with RLVR. But the set of constraints and verifier functions is limited and most models overfit on IFEval. We introduce IFBench to measure model generalization to unseen constraints.
95K
Valentina Pyatkin
@valentina__py
May 15, 2024
My thesis got approved with no corrections and I just received the news that I am now officially a doctor! 😊 A big thank you to my amazing advisors Ido Dagan and @rtsarfaty !
12K
Valentina Pyatkin
@valentina__py
Jun 1, 2023
Life update: my PhD at @biunlp with my advisors @rtsarfaty and Ido Dagan is soon coming to a completion! Next I am excited to be joining @allen_ai @ai2_mosaic @uwnlp for a postdoc with @YejinChoinka I'll get to cross the Fremont Bridge for some more years (usually up when I do):
14K
Valentina Pyatkin
@valentina__py
Nov 21, 2024
Open Post-Training recipes! Some of my personal highlights: 💡 We significantly scaled up our preference data! (using more than 330k preference pairs for our 70b model!) 💡 We used RL with Verifiable Rewards to improve targeted skills like math and precise instruction following
15K
Valentina Pyatkin
@valentina__py
May 23, 2023
Inspired by Yoav’s questions on twitter a while ago, we started to explore “sentence similarity” based on abstract descriptions. Check out the preprint!
37K
Valentina Pyatkin
@valentina__py
Oct 23, 2024
📈New paper on implicit language and context! She bought the largest pumpkin? - Largest pumpkin out of what? All pumpkins in the store? Out of all pumpkins bought by her friends? In the world? Superlatives are (often) ambiguous and their interpretation is extremely context
8.9K
Valentina Pyatkin
@valentina__py
Oct 9, 2024
COLM #COLM2024 was one of the best conferences I ever attended! Perfect size, loved the single-track setup, cool research and Philly is lovely 😍
13K
Valentina Pyatkin
@valentina__py
Dec 21, 2022
⚡️New preprint! 📰 We introduce a question generation approach for disambiguating social and moral situations using RL. With: @liweijianglw, @GXiming, @viveksrikumar, @JenaHwang2, @chandra_bhagav, @YejinChoinka Preprint: arxiv.org/pdf/2212.10409…
20K
Valentina Pyatkin
@valentina__py
May 1, 2025
Accepted to #ICML2025 ! 🥳
Michael Zhang
@mjqzhang
Nov 4, 2024
Why and when do preference annotators disagree? And how do reward models + LLM-as-Judge evaluators handle disagreements? We explore both these questions in a ✨new preprint✨ from my @allen_ai internship! [1/6]
7.1K
Valentina Pyatkin
@valentina__py
Sep 27, 2024
🎉 Very excited that our work on best practices for preference learning has been accepted to #NeurIPS2024! See you in Vancouver!
Hamish Ivison
@hamishivi
Jun 14, 2024
How well do DPO and PPO work on public preference datasets? Excited to share some work exploring the effects of data, reward models, and prompts! We also find that PPO generally beats DPO, despite being more challenging engineering-wise. 📜: arxiv.org/abs/2406.09279 More below 👇
8.3K
Valentina Pyatkin
@valentina__py
Jan 26, 2022
🛫I'm really looking forward to be spending this summer in Seattle, where I'll be interning at the @ai2_mosaic team at @allen_ai ! 🏞️
Valentina Pyatkin
@valentina__py
Mar 10, 2025
I'm honored to be giving an invited talk at Stanford next week at the "Designing Trustworthy AI in a Polarized World" symposium! I'll be staying in the Bay Area and SF for a few days after: happy to meet up and grab a coffee! 🌉
8.5K
Valentina Pyatkin
@valentina__py
Nov 18, 2024
📣 We are looking forward to an interesting line up of invited speakers at the SoLaR workshop at NeurIPS! Come join us on Saturday, Dec. 14th in Vancouver #neurips2024 @solarneurips @_beenkim @PeterHndrsn @zicokolter @qadrida @hannahrosekirk
21K