Third paper in #ICML2024 brilliantly led by Tarun Kalluri (On the job market!)! A new direction that can propel multimodal pretraining! Allen Institute for AI

Also, Tarun is in job market. If you are looking for a seasoned CV researcher, you know where to look for! 🚀

There's a mountain of great research and datasets that exists today — can AI help sort through the pile? We're excited to present DataVoyager, work by Bodhisattwa Majumder and co-authors that aims to assist scientists in discovering new insights and verify known hypotheses

You might have seen this already but if not, we are super excited with this new direction on hypothesis verification and discovery, grounded in data and code for reproducibility! Catch us in upcoming conferences (NAACL, ICML, ACL) to know more!! 🙌🏼

AI-powered scientific discovery?
Yes we can! blog.allenai.org/data-driven-di… Allen Institute for AI

Let’s talk about Language Agents! 🤖

Here's our incredible speaker lineup covering a wide breadth of these topics!! Very excited for Wordplay at ACL this year
Yoav Artzi Bodhisattwa Majumder Emily Dinan, Ida Momennejad, and Joyce Chai Michigan SLED Lab

🎉We're celebrating our OLMo team's big win last night for Innovation of the Year at the GeekWire Awards! Thank you to the panel for recognizing our efforts towards open science, and to the other nominees for pushing us all towards better AI technology! geekwire.com/2024/geekwire-…

GeekWire Awards 2024 revealed: Winners bask in momentum of AI and potential of region geekwire.com/2024/geekwire-…

martin_casado Technical academia (at least my little corner of it) thinks industry is wasting billions on a technology that is not fit for purpose. We are all using the open weights models and cheering for Allen Institute for AI's fully open LLM effort.

Did you know there’s a task people easily solve but GPT-4 fails? From a few input-output grids, ARC asks you to infer and apply a rule

With Hypothesis Search, we double GPT-4’s score
arxiv.org/abs/2309.05660

w/Ruocheng Wang Gabriel Poesia evanthebouncy Nick Haber noahdgoodman
🧵

🇦🇹 While y'all swarming at ICLR and talking about (language) agents, keep our work from #ICML2024 in heart: how learning 'skills' is effective for in-context policy improvement. Skills can be learned ubiquitously to support continual learning in otherwise frozen agents!

Research from different communities, including Social and Interactive NLP, HRI, Human-AI Interaction, and Multi-Agent RL, are solving the 'social intelligence' problem from different angles.

Therefore, we compose a collection of awesome social agents! bit.ly/SocialAgents

First paper on NBER---hopefully many more to come!

So thankful for brilliant collaborators like John Horton and Kehang Zhu, I don't know how they function in society being so smart ;)

Can generative models assist researchers explore new dimensions within data with novel hypothesis generation and statistical model estimations? Check out our position paper, now accepted ICML Conference. Fantastic collaboration indeed.
More coming soon; stay tuned! Allen Institute for AI EconUofU

It has been a fascinating collab - and much more to come. In the long run, this can help tackle some of the most pressing scientific questions facing humanity.

Incredibly proud of our teamwork, now in ICML Conference! This position starts a series of work on data-driven scientific discovery w generative models.
Follow-ups coming soon on benchmarks, systems, & accessibility in science!
arxiv.org/abs/2402.13610
#ICML2024 Allen Institute for AI Aristo Team at AI2

Skill Set Optimization was accepted to ICML Conference 2024! I'm proud of this work and everything we learned about in-context policy improvement. Big thanks to my collaborators at Allen Institute for AI. Way to go team!

Excited to share our work, 'Skill Set Optimization', a continual learning method for LLM actors that:
- Automatically extracts modular subgoals to use as skills
- Reinforces skills using environment reward
- Facilitates skill retrieval based on state
allenai.github.io/sso
🧵

I am really excited to be part of this project led by Sayash Kapoor and Arvind Narayanan to help improve practices in machine-learning based science. science.org/doi/10.1126/sc…