bhavana dalvi (@bhavana_dalvi) 's Twitter Profile
bhavana dalvi

@bhavana_dalvi

Lead Research Scientist at Allen Institute for Artificial Intelligence @allenai_org , NLP/AI Research

ID: 64248393

calendar_today09-08-2009 20:37:32

46 Tweet

165 Followers

47 Following

Bodhisattwa Majumder (@mbodhisattwa) 's Twitter Profile Photo

We are excited to announce CLIN 🤖: The first continually learning language agent that excels in both task adaptation and generalization to unseen tasks and environments in a pure zero-shot setup. Aristo Team at AI2 Ai2 Website: allenai.github.io/clin/ Let's dive in 🧵 (1/n)

Bodhisattwa Majumder (@mbodhisattwa) 's Twitter Profile Photo

(n/n) This was only possible with the huge team effort: bhavana dalvi, Peter Jansen ( @peterjansen-ai.bsky.social ), Oyvind, Niket, Li "Harry" Zhang, Chris Callison-Burch, and Peter Clark! CLIN is open source: github.com/allenai/clin Paper: arxiv.org/pdf/2310.10134…

(n/n) This was only possible with the huge team effort:
<a href="/bhavana_dalvi/">bhavana dalvi</a>, <a href="/peterjansen_ai/">Peter Jansen ( @peterjansen-ai.bsky.social )</a>, Oyvind, Niket, <a href="/liharryzhang/">Li "Harry" Zhang</a>, Chris Callison-Burch, and Peter Clark! 

CLIN is open source: github.com/allenai/clin
Paper: arxiv.org/pdf/2310.10134…
Ai2 (@allen_ai) 's Twitter Profile Photo

Using a dynamic, persistent memory, CLIN is the first continually learning model that excels in both task adaptation & generalization to unseen tasks in a zero-shot setup. Explore this exciting new work by AI2's Bodhisattwa Majumder and bhavana dalvi, and the Aristo Team at AI2 team:

bhavana dalvi (@bhavana_dalvi) 's Twitter Profile Photo

I am thrilled to share our work: Using a dynamic, persistent memory🧠 CLIN is the first continually learning model 🤖 that excels in both task adaptation & generalization to unseen tasks in a zero-shot setup 🧙 Bodhisattwa Majumder Ai2 Aristo Team at AI2 webpage: allenai.github.io/clin/

Kolby Nottingham (@kolbytn) 's Twitter Profile Photo

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 🧵

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
🧵
bhavana dalvi (@bhavana_dalvi) 's Twitter Profile Photo

Excited to share our work on Continual Learning via Skill Set Optimization🥳 Can a language agent explore new interactive tasks, automatically identify useful skills, generalize them to new tasks? allenai.github.io/sso with Kolby Nottingham Bodhisattwa Majumder Sameer Singh Peter Clark Roy Fox (@[email protected])

Bodhisattwa Majumder (@mbodhisattwa) 's Twitter Profile Photo

Is it possible to build end-to-end autonomous discovery systems using Large Generative Models (LGMs)? 🧬 In this position paper, we argue: arxiv.org/pdf/2402.13610… 🧵 (1/n) Ai2 Aristo Team at AI2 Harshit Surana UMass Amherst University of Utah

bhavana dalvi (@bhavana_dalvi) 's Twitter Profile Photo

Excited to share our recent paper led by Nathaniel Weir 🥳 Enhancing Systematic Decompositional Natural Language Inference Using Informal Logic tinyurl.com/bdehyv5k Thanks to all the collaborators!

bhavana dalvi (@bhavana_dalvi) 's Twitter Profile Photo

NLRSE workshop @ ACL 2024: Deadline extended to May 21! Also note that non-archival cross-submissions (papers accepted to other venues, such as ACL Findings) can be submitted on the Google Form here: docs.google.com/forms/d/1OAzZE…

Bodhisattwa Majumder (@mbodhisattwa) 's Twitter Profile Photo

.Christopher Manning in the first keynote Conference on Language Modeling grazed over the meaning of true intelligence saying it involves real-time “adaptation” — but how do you do it? Learn our way of purely in-context continual learning for LLMs! Poster on Wed (4:30 EST) 🦙

Peter Jansen ( @peterjansen-ai.bsky.social ) (@peterjansen_ai) 's Twitter Profile Photo

Can language models perform end-to-end scientific discovery? In our NeurIPS Spotlight paper, we show: very rarely. Our best model found <20% of discoveries, our best PhDs found nearly all. Paper: arxiv.org/pdf/2406.06769 Code/Web: allenai.github.io/discoveryworld Ai2 Microsoft Research

Can language models perform end-to-end scientific discovery? In our NeurIPS Spotlight paper, we show: very rarely.

Our best model found &lt;20% of discoveries, our best PhDs found nearly all.

Paper: arxiv.org/pdf/2406.06769
Code/Web: allenai.github.io/discoveryworld
<a href="/allen_ai/">Ai2</a> <a href="/MSFTResearch/">Microsoft Research</a>
bhavana dalvi (@bhavana_dalvi) 's Twitter Profile Photo

Call for papers: 🤖AI & Scientific Discovery 👩‍🔬 Workshop at NAACL 2025. Submissions due on 30th Jan 2025. 👩‍💻🧑‍💻

Peter Jansen ( @peterjansen-ai.bsky.social ) (@peterjansen_ai) 's Twitter Profile Photo

Only a few more days to submit your AI + Scientific Discovery papers to the AISD workshop NAACL HLT 2025! The cross-submission form is now active -- have a submitted or already-published paper and want to present it to scientific discovery colleagues? Submit today!