Sign in

Jacob Eisenstein

@jacobeisenstein.bsky.social
5.7K followers 2.5K following 230 posts

natural language processing and computational linguistics at google deepmind.

PostsRepliesMedia
Reposted by Jacob Eisenstein
Bryan Cantrill @bcantrill.bsky.social · 05/09/2026
The revolt of the reader bcantrill.dtrace.org/2026/09/05/t...
bcantrill.dtrace.org
The revolt of the reader | The Observation Deck
1112431
Jacob Eisenstein @jacobeisenstein.bsky.social · 22/08/2026
!!!
1160
Reposted by Jacob Eisenstein
Victor Geislinger @victorsvector.com · 13/08/2026
Been testing this 'sign-language-to-text' (SL2T) for almost a year & it's so awesome to see this out! Looked at this almost a decade ago & knew we'd get the tech & ML models there eventually (but super challenging). Talking w/ some Deaf folks, an equivalent of 'voice-to-text' was huge for them
deepmind.google
Putting sign language AI into users’ hands
Introducing sign-language-to-text (SL2T), our breakthrough model powering new sign language features for Deaf and hard of hearing users.
210533
Reposted by Jacob Eisenstein
Eytan Adar @eytan.adar.prof · 11/08/2026
I wrote a post about the emergence of AI-native students who do their research using GenAI tools. I think that there are some real risks to job prospects and the likely changes in their fields.
1378
Jacob Eisenstein @jacobeisenstein.bsky.social · 08/07/2026
a spontaneously-formed triple snaking queue to enter the #icml2026 poster session. the conference sold out before the early bird registration period ended. no shade whatsoever to the organizers, who are making the best of this, but we have strayed from light. conferences cannot be this big.
2322
Jacob Eisenstein @jacobeisenstein.bsky.social · 28/06/2026
It’s a small, biased sample, but my AC batch for @colmweb.org was miles better than the papers i reviewed for NeurIPS: more creative, more rigorous, better written. Maybe AI research isn’t doomed to 2-3 omni-conferences after all?
colmweb.org
COLM 2026
1212
Reposted by Jacob Eisenstein
Nathan Lambert @natolambert.bsky.social · 15/04/2026
I spent some time trying to distill all the complex factors impacting open models -- economics, capabilities, distribution, policy, etc. -- into a clear list of beliefs. Here they are in full. www.interconnects.ai/p/my-bets-on...
interconnects.ai
My bets on open models, mid-2026
What I expect to come next and why, focused on the open-closed gap.
1274
Reposted by Jacob Eisenstein
Tom Schaul @schaul.bsky.social · 16/03/2026
DeepMind's RL team is hiring a research scientist: if you're passionate about RL, come work with us! And if you know people who might be interested, please share: job-boards.greenhouse.io/deepmind/job...
job-boards.greenhouse.io
Research Scientist, Reinforcement Learning
London, UK
12914
Reposted by Jacob Eisenstein
Jonathan Berant @jonathanberant.bsky.social · 06/03/2026
Newish work (arXived in December): Prompts can be ambig., but handling ambiguity is context/user dependent. Sometimes the right thing is to ask a clarifying question, sometimes to give multi. answers, and sometimes to just guess. Can we train steerable models that change their strategy per context?
121
Jacob Eisenstein @jacobeisenstein.bsky.social · 04/03/2026
Are AI models effective collaborators, or mere assistants awaiting your next command? (Preprint: 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.
35621
Jacob Eisenstein @jacobeisenstein.bsky.social · 01/03/2026
This looks like it’ll be a fantastic intro to transformers ⚡️
1224
Reposted by Jacob Eisenstein
Caroline Wang @caroline-wang.bsky.social · 16/02/2026
[1/n] Just wrapped up 7 months interning with @pcastr.bsky.social at Google DeepMind and I'm so excited to share our work: arxiv.org/abs/2602.10324. TLDR: We used LLM-powered program synthesis to automatically model and discover differences between human and LLM strategic behavior
28114
Reposted by Jacob Eisenstein
Quentin Berthet @ AISTATS @qberthet.bsky.social · 16/02/2026
🚨 🔬 PhD positions at Google DeepMind in France 🇫🇷 We are advertising Master Level Intern positions at Google DeepMind within our Frontier AI Unit. These could lead to co-advised PhD positions with Google DeepMind and French academic institutions. job-boards.greenhouse.io/deepmind/job...
job-boards.greenhouse.io
Stagiaire de niveau Master, France / Master Level Intern, France
Grenoble, France ; Paris, France
12917
Jacob Eisenstein @jacobeisenstein.bsky.social · 20/12/2025
meanwhile, in the pnw www.seattletimes.com/seattle-news...
seattletimes.com
Be aware of toilet rats, King County says
Rats could climb into your toilet as a result of recent flooding from atmospheric rivers, King County officials say. Enter "wet rat winter."
040
Reposted by Jacob Eisenstein
Marc Lanctot @sharky6000.bsky.social · 19/12/2025
Hello! 👋 Are you interested in AI for board games using language models? Want to do some hobby tinkering with fine-tuning or RL? We've released an easy-to-follow example colab that fine-tunes Gemma models via Kauldron to mimic an MCTS player. Details here: github.com/google-deepm... ♟️🎲♦️♠️♥️♣️✨🎉
github.com
2025 Wrap-up: Fine-tuning Gemma with Kauldron Example ✦︎ · Issue #1414 · google-deepmind/open_spiel
Hello everyone! We've been hard at work this year working on OpenSpiel 2.0, which will be better than ever. Major developments have been underway to make working with language models easier. I'm lo...
2407
Reposted by Jacob Eisenstein
Gretchen McCulloch @gretchenmcculloch.com · 01/12/2025
Day 1 of #BooksAreMyJam! Blueberry Maple jam, with Linguaphile: A life of language love by Julie Sedivy. A classic Canadian flavour duo + this book about @juliesedivy.bsky.social's relationship with language through her childhood in Montreal, later research as a linguist, and more
Linguaphile: A life of language love book next to a tiny pot of jam saying "Bonne Maman: myrtille et érable (blueberry and maple) on top of the Bonne Maman jamvent calendar
510811
Reposted by Jacob Eisenstein
mr. TIM @timkellogg.me · 25/11/2025
this is the theme — you can’t have AGI without existing in and learning from the real world
Dwarkesh Patel @dwarkesh_sp
X.com
"The thing that happened with AGI and pretraining is that in some sense they overshot the target.
You will realize that a human being is not an AGI.
Because a human being lacks a huge amount of knowledge. Instead, we rely on continual learning.
If I produce a super intelligent 15-year-old, they don't know very much at all. A great student, very eager. [You can say,] 'You go and be a programmer. You go and be a doctor. Go and learn.'
So you could imagine that the deployment itself will involve some kind of a learning trial and error period. It's a process as opposed to, you drop the finished thing."
@ilyasut
1212
Reposted by Jacob Eisenstein
Kaitlyn Zhou @kaitlynzhou.bsky.social · 06/11/2025
No better time to start learning about that #AI thing everyone's talking about... 📢 I'm recruiting PhD students in Computer Science or Information Science @cornellbowers.bsky.social! If you're interested, apply to either department (yes, either program!) and list me as a potential advisor!
Photo of Cornelll University building surrounded by colorful trees
2259
Jacob Eisenstein @jacobeisenstein.bsky.social · 17/10/2025
knowing how to tie your shoes or order a drink in a crowded bar: not agi naming the big five personality traits: definitely agi
170
Jacob Eisenstein @jacobeisenstein.bsky.social · 17/10/2025
nice summary of everybody’s new fave
041
Jacob Eisenstein @jacobeisenstein.bsky.social · 09/10/2025
Nicholas Carlini asking the right questions at #COLM2025
050
Reposted by Jacob Eisenstein
Maria Antoniak @mariaa.bsky.social · 06/10/2025
Here’s a #COLM2025 feed! Pin it 📌 to follow along with the conference this week!
22617
Reposted by Jacob Eisenstein
Myra Cheng @myra.bsky.social · 03/10/2025
AI always calling your ideas “fantastic” can feel inauthentic, but what are sycophancy’s deeper harms? We find that in the common use case of seeking AI advice on interpersonal situations—specifically conflicts—sycophancy makes people feel more right & less willing to apologize.
Screenshot of paper title: Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence
5544270
Reposted by Jacob Eisenstein
Pete Shaw @ptshaw.bsky.social · 01/10/2025
Excited to share a new paper that aims to narrow the conceptual gap between the idealized notion of Kolmogorov complexity and practical complexity measures for neural networks.
Bridging Kolmogorov Complexity and Deep Learning: Asymptotically Optimal Description Length Objectives for Transformers
195
Reposted by Jacob Eisenstein
Joshua Raclaw @joshuaraclaw.com · 26/08/2025
Cannot stress enough how good it is that you can come across a post about gorgeous little Yiddish book sitting in someone’s family collection, and within a few seconds you can find the full scanned version of the book available for free through the Yiddish Book Center’s website
yiddishbookcenter.org
Asṭronomye | Yiddish Book Center
4437
Jacob Eisenstein @jacobeisenstein.bsky.social · 25/08/2025
found some books at my parents’ house
yiddish book coverautomatic translation: autonomy by dr. b hoffman
2201
Jacob Eisenstein @jacobeisenstein.bsky.social · 19/08/2025
061
Jacob Eisenstein @jacobeisenstein.bsky.social · 11/08/2025
Baristas still safe from robotic automation, and not just because robots don’t know what coffee tastes like. prompt: “I’m trying to dial in this v60 of huatusco with my vario. temp / grind recommendations?”
130
Jacob Eisenstein @jacobeisenstein.bsky.social · 08/08/2025
I’d guess that the majority position of syntacticians about LLMs (and other NLP beforehand) is roughly what Chomsky says: language tech can’t possibly teach us anything about the human language capability, so whether the LLM writes well doesn’t matter at all.
1100
Jacob Eisenstein @jacobeisenstein.bsky.social · 07/08/2025
boston champaign pittsburgh atlanta, and, uh, let’s count seattle glad i did it, hope i don’t have to do it again
021
Reposted by Jacob Eisenstein
Margaret Mitchell @mmitchell.bsky.social · 06/08/2025
🤖 ICYMI: Yesterday, @hf.co and OpenAI partnered to bring open source GPT to the public. This is a Big Deal in "AI world". Allow me to explain why. 🧵 huggingface.co/openai/gpt-o...
Yellow background with orange border. OpenAI logo top center, "OpenAI's GPT OSS" bottom center. Left side in smaller writing says "From the makers of ChatGPT...A new model is released". On the right and left side are spikey orange announcement banners, reading "Now Open Source!" "Available on Hugging Face! <logo>"
url given in bottom right corner is hf.co/openai/gpt-oss-120b
25217
Reposted by Jacob Eisenstein
Ibn Bassal @ibnbassal.bsky.social · 29/07/2025
roman burrito thread
315138
Reposted by Jacob Eisenstein
Jacob Eisenstein @jacobeisenstein.bsky.social · 29/07/2025
this is very cool and i’m looking forward to reading the paper, but a basic question about this data: isn’t it likely that a congressional rep’s speeches are written by a shifting cast of speechwriters over the course of their career? wouldn’t that explain adoption of new usages?
131
Reposted by Jacob Eisenstein
Gaurav Kamath @grvkamath.bsky.social · 29/07/2025
Our new paper in #PNAS (bit.ly/4fcWfma) presents a surprising finding—when words change meaning, older speakers rapidly adopt the new usage; inter-generational differences are often minor. w/ Michelle Yang, ‪@sivareddyg.bsky.social‬ , @msonderegger.bsky.social‬ and @dallascard.bsky.social‬👇(1/12)
33317
Jacob Eisenstein @jacobeisenstein.bsky.social · 25/07/2025
There's a lot to like in this position paper - and not just the "whiff of Frankenstein" quote. www.arxiv.org/abs/2507.06268
1151
Reposted by Jacob Eisenstein
Maria Antoniak @mariaa.bsky.social · 23/07/2025
What are your favorite recent papers on using LMs for annotation (especially in a loop with human annotators), synthetic data for task-specific prediction, active learning, and similar? Looking for practical methods for settings where human annotations are costly. A few examples in thread ↴
137923
Reposted by Jacob Eisenstein
Ahmad Beirami @abeirami.bsky.social · 09/07/2025
[Thu Jul 17] w/ Ananth Balashankar & @jacobeisenstein.bsky.social, we present a reinforcement learning framework in view of test-time scaling. We show how to optimally calibrate & transform rewards to obtain optimal performance with a given test-time algorithm.
111
Reposted by Jacob Eisenstein
Ahmad Beirami @abeirami.bsky.social · 09/07/2025
[Wed Jul 16] w/ @jacobeisenstein.bsky.social & Alekh Agarwal, we present a theoretical characterization of best-of-N (a simple yet effective method for test-time scaling & alignment). Our results justify the widespread use of BoN as a strong baseline in this space.
101
Jacob Eisenstein @jacobeisenstein.bsky.social · 10/06/2025
Cheap but noisy? Or accurate but expensive? How to split a limited annotation budget between different types of judges?👩‍⚖️🤖🦧 www.arxiv.org/abs/2506.07949
arxiv.org
Cost-Optimal Active AI Model Evaluation
The development lifecycle of generative AI systems requires continual evaluation, data acquisition, and annotation, which is costly in both resources and time. In practice, rapid iteration often makes...
193
Reposted by Jacob Eisenstein
Ned Resnikoff @resnikoff.bsky.social · 06/06/2025
Everyone should check out People Time, the recording of his dates with Kenny Barron in Copenhagen three months before his death. Getz knew he was dying, and produced some of the most moving music of his career. www.youtube.com/watch?v=c3jd...
youtube.com
East Of Sun (And West Of The Moon)
YouTube video by Kenny Barron - Topic
1242
Jacob Eisenstein @jacobeisenstein.bsky.social · 28/05/2025
Great topic! looking forward to this
020
Jacob Eisenstein @jacobeisenstein.bsky.social · 25/05/2025
most relatable parenting thread
000
Reposted by Jacob Eisenstein
Ferenc Huszár @inference.vc · 22/05/2025
A new blog post with intuitions behind continuous-time Markov chains, a building block of diffusion language models, like @inceptionlabs.bsky.social's Mercury and Gemini Diffusion. This post touches on different ways of looking at Markov chains, connections to point processes, and more.
inference.vc
Discrete Diffusion: Continuous-Time Markov Chains
A tutorial explaining some key intuitions behind continuous time Markov chains for machine learners interested in discrete diffusion models: alternative representations, connections to point processes...
1215
Reposted by Jacob Eisenstein
Stella Biderman @stellaathena.bsky.social · 23/05/2025
People keep plugging AI "Co-Scientists," so what happens when you ask them to do an important task like finding errors in papers? We built SPOT, a dataset of STEM manuscripts across 10 fields annotated with real errors to find out. (tl;dr not even close to usable) #NLProc arxiv.org/abs/2505.11855
411931
Jacob Eisenstein @jacobeisenstein.bsky.social · 22/05/2025
‘intermediate tokens-often anthropomorphized as "thoughts" or reasoning traces’ 🌶️ but true! really glad to see work approaching inference scaling more skeptically and objectively
0132
Reposted by Jacob Eisenstein
Myra Cheng @myra.bsky.social · 21/05/2025
Dear ChatGPT, Am I the Asshole? While Reddit users might say yes, your favorite LLM probably won’t. We present Social Sycophancy: a new way to understand and measure sycophancy as how LLMs overly preserve users' self-image.
614735
Jacob Eisenstein @jacobeisenstein.bsky.social · 21/05/2025
Google Deepmind is hiring a research scientist in Seattle to work on foundational research in language! job-boards.greenhouse.io/deepmind/job...
job-boards.greenhouse.io
Research Scientist, Foundational Research in Language, USA
Seattle, Washington, US
091
Reposted by Jacob Eisenstein
Ahmad Beirami @abeirami.bsky.social · 09/05/2025
#ICML2025 Is standard RLHF optimal in view of test-time scaling? Unsurprisingly no. We show a simple change to standard RLHF framework that involves 𝐫𝐞𝐰𝐚𝐫𝐝 𝐜𝐚𝐥𝐢𝐛𝐫𝐚𝐭𝐢𝐨𝐧 and 𝐫𝐞𝐰𝐚𝐫𝐝 𝐭𝐫𝐚𝐧𝐬𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐨𝐧 (suited to test-time procedure) is optimal!
1176
Reposted by Jacob Eisenstein
Naomi Saphra @nsaphra.bsky.social · 08/05/2025
uh oh! uh oh!
10354
Reposted by Jacob Eisenstein
Gus @gusthema.bsky.social · 30/04/2025
Gemma 3 explained: Longer context, image support, and a new 1B model. → goo.gle/4lV8iaw Other key enhancements: 🔸 Best model that fits in a single consumer GPU or TPU host 🔸 KV-cache memory reduction with 5-to-1 interleaved attention 🔸 And more! Read the blog for the full details on Gemma 3.
goo.gle
Gemma explained: What’s new in Gemma 3- Google Developers Blog
Google's Gemma 3 model includes vision-language support and architectural changes for resource-friendly multimodal language models.
1228