Research Scientist @GoogleAI | previously ML & NLP PhD student @biunlp, intern at @allen_ai, @Microsoft, @AIatMeta.
Joined July 2009
- Tweets309
- Following487
- Followers639
- Likes1.2K
Pinned Tweet
🚨 New Paper 🚨
Are current LLMs up to the task of solving *complex* instructions based on content-rich text?
Our new dataset, TACT, sheds some light on this challenge.
How does it work?
arxiv.org/abs/2406.03618
Work by @GoogleAI & @GoogleDeepMind
👇🧵
Avi Caciularu retweeted
Our team is hiring for core roles on the Gemini Post-Training team in Mountain View!
google.com/about/careers/app…
Targeting candidates with industry experience, but strong PhD grads will also be considered.
Feel free to DM (apologies in advance if I can't reply to everyone)
Avi Caciularu retweeted
In deep research reports, many sentences aren't supported by their citations: recall is 58.7% in NVIDIA AI-Q and 7.1% in TrajectoryKit.
Our #EMNLP2026 paper (main conf) proposes an algorithm that traces each error to the agent that introduced it, and identifies the error type.
Avi Caciularu retweeted
for quite a while I was trying to find a way to measure x-ling consistency without resorting to knowledge, @bobbycxy and @LChoshen came with this games approch. It is honestly brilliant!
Avi Caciularu retweeted
🔥 New paper alert!! Reinforcement Learning with Metacognitive Feedback Elicits Faithful Uncertainty Expression in LLMs 🔥
Can a model that learns to 𝘫𝘶𝘥𝘨𝘦 𝘪𝘵𝘴𝘦𝘭𝘧 get even better at tasks—and learn to be honest about what it doesn't know? Turns out: yes🤯
Details🧵👇
Avi Caciularu retweeted
🚨new ACL paper🚨
🗄️dataset: LocQA🗄️
As we make models multilingual, do we inadvertently make them conflate knowledge across different languages? If we ask a model “who was the 1st president?” will it answer “George Washington” regardless of the language or locale? A 🧵
Avi Caciularu retweeted
Meet Gemini 3.5 Flash — our strongest agentic and coding model yet.
It delivers frontier-level performance at 4x the speed of comparable frontier models — often at less than half the cost.
Generally available, starting today. 🧵
#GoogleIO
ALT Text that says Gemini 3.5 Flash set against a deep blue background.
Avi Caciularu retweeted
GenAI has moved from aspiration to a core requirement. For enterprises, the question is no longer "what can it do?" but "how reliably can it do it?"
At @Google, we view factuality as the essential foundation for this next phase of AI utility. (1/4)
Avi Caciularu retweeted
Introducing ConvApparel, a new human-AI conversation dataset, as well as a comprehensive evaluation framework designed to quantify the "realism gap" in LLM-based user simulators and improve the training of robust conversational agents.
Read all about it →goo.gle/41k5eff
Avi Caciularu retweeted
Are AI models effective collaborators, or mere assistants awaiting your next command? (arxiv.org/abs/2602.24188)
To find out, we make AI collaborate with itself, in private information games: tasks that require sharing private information, like this chess board ordering task.
Avi Caciularu retweeted
Very cool internship project by @AmosaurusRex from our group at @GoogleResearch! Suggests models can start “thinking early” instead of waiting for all the input tokens to be processed.
Can LLMs reason internally while processing their inputs, similar to how humans can think ahead as we process information? Our latest work introduces Thinking States, a novel architectural adaptation that transforms reasoning into a internal recurrent process.
By training models to maintain a dynamic thinking state, we achieve significant inference speedups over Chain-of-Thought while substantially outperforming existing latent reasoning methods.
Paper: arxiv.org/abs/2602.08332
CoT is powerful but slow. @AmosaurusRex, my intern, introduces Thinking States—a way for LLMs to reason while processing input, not after.
🚀 Better latency
📈 (Almost) Matches CoT on complex QA
🧠 Learns from NL supervision
Check out his thread 🧵
Can LLMs reason internally while processing their inputs, similar to how humans can think ahead as we process information? Our latest work introduces Thinking States, a novel architectural adaptation that transforms reasoning into a internal recurrent process.
By training models to maintain a dynamic thinking state, we achieve significant inference speedups over Chain-of-Thought while substantially outperforming existing latent reasoning methods.
Paper: arxiv.org/abs/2602.08332
Avi Caciularu retweeted
Gemini 3.1 Pro is here. Hitting 77.1% on ARC-AGI-2, it’s a step forward in core reasoning (more than 2x 3 Pro).
With a more capable baseline, it’s great for super complex tasks like visualizing difficult concepts, synthesizing data into a single view, or bringing creative projects to life.
We’re shipping 3.1 Pro across our consumer and developer products to bring this underlying leap in intelligence to your everyday applications right away. Rolling out now to:
- Developers in preview via the Gemini API in @GoogleAIStudio
- Enterprises in Vertex AI and Gemini Enterprise
- Everyone through the @Geminiapp and @NotebookLM
Avi Caciularu retweeted
Don't forget to register for ISCOL 2025 😊
Registration closes in less than a week!
Registration for ISCOL is now open! Please register as soon as possible, as spots are limited and we need to get an accurate headcount.
iscol-meeting.github.io/isco…
Avi Caciularu retweeted
My team @GoogleAI is looking for a 2026 research intern in Mountain View! I will be hiring for a project aimed at improving tool-using and search agents via RL training and data generation.
To apply: google.com/about/careers/app…
+ feel free to ping me!
Avi Caciularu retweeted
Introducing Gemini 3 Pro, the world's most intelligent model that can help you being anything to life.
It is state of the art across most benchmarks, but really comes to life across our products (AI Studio, the Gemini API, Gemini App, etc) 🤯
Avi Caciularu retweeted
Registration for ISCOL is now open! Please register as soon as possible, as spots are limited and we need to get an accurate headcount.
iscol-meeting.github.io/isco…
Avi Caciularu retweeted
Save the date for ISCOL'25! The conference will be held on December 18th at Bar-Ilan University. The call for papers is now live on our website: iscol-meeting.github.io/isco…