Paul Sava @_psava
research scientist @FraunhoferAISEC. opinions are mine. paulsava.github.io Munich, Germany Joined December 2021-
Tweets115
-
Followers11
-
Following405
-
Likes467
This is a great and also your periodic reminder that steering / control vectors are powerful and under researched (at least publicly)
This is a great and also your periodic reminder that steering / control vectors are powerful and under researched (at least publicly)
🤯 New research deploys DSPy-optimized system in real-world medical settings. Finds 70% increase in positive patient feedback. The authors introduce Dr.Copilot, a multi-agent DSPy assistant in Romanian. Optimized along 17 axes like Empathy & Explanations using dspy.SIMBA.
DSPy 3.0 beta Talk by @lateinteraction Link below
We are adding more real-world examples to DSPy tutorials: dspy.ai/tutorials/real…, please check them out and let us know what else you want to see and learn! We welcome contributions! Create a feature request and describe what you want to ship, we can discuss from there!
Ok I guess I have to go through that Apple paper. My main issue is the framing which is super binary: "Are these models capable of generalizable reasoning, or are they leveraging different forms of pattern matching?" Or what if they only caught genuine yet partial heuristics.
Announcing the release of the official Common Corpus paper: a 20 page report detailing how we collected, processed and published 2 trillion tokens of reusable data for LLM pretraining.
it probably feels as a self-defeating thing to say for researchers but NEXT TOKEN PREDICTION IS THE MOST OPTIMAL COMPRESSION. - shannon, 1948 (a mathematical theory of communication)
it probably feels as a self-defeating thing to say for researchers but NEXT TOKEN PREDICTION IS THE MOST OPTIMAL COMPRESSION. - shannon, 1948 (a mathematical theory of communication) https://t.co/4yfUVYtHtQ
So many things in the run-up to DSPy 3. Here's a first, EXPERIMENTAL one: 🚨We're releasing dspy.GRPO, an online RL optimizer for DSPy programs Your DSPy code as-is can be dspy.GRPO'ed. Yes, even compound multi-module programs. Led by @NoahZiems @LakshyAAAgrawal @dilarafsoylu.
We partnered with @huggingface to teach you how to fine-tune LLMs with GRPO! Learn about: • Reward functions + creating them • GRPO Math + Free Reasoning training in Colab • Applying RL to real-world use cases Course: huggingface.co/reasoning-cour… Tutorial: docs.unsloth.ai/basics/reasoni…
i have been reading lately more about the history of RNNs and found this passage in Lashley's 'The Problem of Serial Order in Behavior' (languagelog.ldc.upenn.edu/myl/Lashley195…) that made my day. Don't think he was a fan of Titchener's work, lol
@mariofilhoml it is up to all of us to bring. it. back. say no to professional sponsored influencers with hyper-optimized content. say yes to boutique anons from the internet speaking their mind on their little corner of the internet.
this is massive and i am very shocked that my timeline isn't all over this what the hell
this is massive and i am very shocked that my timeline isn't all over this what the hell
We’re launching a new blog: Engineering at Anthropic. A hub where developers can find practical advice and our latest discoveries on how to get the most from Claude.
VLMs connect a vision encoder to a language model using a linear layer—meaning each visual feature becomes one token. To reduce the number of tokens, SmolVLM and SmolDocling use pixel shuffle—but how does it work? 🔄 Pixel shuffle rearranges the encoded image, trading spatial…
We teamed up with @huggingface to release a free notebook for fine-tuning Gemma 3 with GRPO! Learn to: • Enable reasoning in Gemma 3 (1B) • Prepare/understand reward functions • Make GRPO work for tiny LLMs Notebook: colab.research.google.com/github/unsloth… Details: huggingface.co/reasoning-cour…
Introducing KBLaM, an approach that encodes and stores structured knowledge within an LLM itself. By integrating knowledge without retraining, it offers a scalable alternative to traditional methods. msft.it/6011qniy9
A very exciting day for open-source AI! We're releasing our biggest open source model yet -- OLMo 2 32B -- and it beats the latest GPT 3.5, GPT 4o mini, and leading open weight models like Qwen and Mistral. As usual, all data, weights, code, etc. are available. For a long time,…
angesichts der aktuellen politischen lage ist es schwer, die balance zu finden zwischen „ich will informiert bleiben“ und „ich will mental gesund bleiben“
Serhan Yilmaz @srhnylmz14
203 Followers 1K Following intern @LTIatCMU & cs @sabanciu // prev @EPFL @TMH_KTH @BU_Tweets @kocuniversity // contact: dm
Chad DeChant @chaddechant
408 Followers 3K Following Computer Science PhD candidate @Columbia studying language & memory & robots; teaching AI Safety, Ethics, and Policy; Sanghas Supporting Refugees @villagezendo
Luca Demetrio @zangobot
472 Followers 517 Following Assistant Professor at @UniGenova, former @zenhackteam member, Adversarial ML pwn & rev, dancer, art lover, and many other things
Ana-Maria Cretu @AnaMariaCretu5
365 Followers 511 Following https://t.co/RGPTjgX7VW. Post-doc in privacy and AI safety at @EPFL and @cydcampus. PhD in data privacy from @imperialcollege.
Lorenzo @LorenzoCazz
289 Followers 701 Following Postdoctoral Researcher @CaFoscari | Previously at @CISPA | Adversarial Machine Learning and Verification of Machine Learning models.
Battista Biggio @biggiobattista
3K Followers 2K Following Full Professor at University of Cagliari (Italy), Co-Founder of Pluribus One. #Security of #MachineLearning, #CyberSecurity & #ComputerVision
Maura Pintor @maurapintor
512 Followers 525 Following Assistant Professor @univca. Computer Science, Engineering, and Futsal lover.
Fabio Pierazzi @fbpierazzi
1K Followers 2K Following Associate Professor at @uclcs in @uclisec. Interested in AI and Systems Security. Opinions are my own.
Leo Dianbo Liu @DianboLiu
1K Followers 1K Following Assistant professor at @NUSingapore . Previously @BroadInstitute @Mila @Harvard @BostonChidrens @MIT @MIT_CSAIL,@dundeeuni @EdinburghUni @SingaporePoly e/acc
Tim David @TimDaOpenAccess
585 Followers 4K Following Journal of Pure and Applied Mathematics (ISSN-2752-8081) aims to publish the latest and the research articles, reviews, and letters on pure Applied Mathematics.
davidp @xTouK
29 Followers 142 Following
Bing Liu @vbingliu
844 Followers 99 Following Head of Research @Scale_AI, ex-Meta Llama 3, PhD @CarnegieMellon.
Behnam Neyshabur @bneyshabur
30K Followers 860 Following Research @AnthropicAI (Co-lead Discovery team) 💼 Past: Gemini @GoogleDeepMind (Co-led Blueshift team) 🧠 LLM Reasoning / AI Scientist 🎒Traveling & Backpacking
Eugene Bagdasarian @ebagdasa
1K Followers 618 Following Challenge AI security and privacy practices. Asst Prof at UMass @manningcics. Researcher at @GoogleAI. he/him 🇦🇲 (opinions mine)
davidad 🎇 @davidad
20K Followers 9K Following Programme Director @ARIA_research | accelerate mathematical modelling with AI and categorical systems theory » build safe transformative AI » cancel heat death
Rohan Pandey @khoomeik
40K Followers 2K Following descending cross-entropy to ascend entropy @PeriodicLabs || prev research @OpenAI @CarnegieMellon '23
Yi Tay @YiTayML
47K Followers 82 Following research scientist @googledeepmind ✨♊, model co-lead/captain of gemini deepthink imo gold medal 🥇, opinions are my own.
Luca Soldaini 🎀 @soldni
11K Followers 1K Following I like tokens! I lead the OLMo data team at @allen_ai w/ @kylelostat. Open source is fun 🤖☕️🍕🏳️🌈 Opinions are sampled from my own stochastic parrot
Bob McGrew @bobmcgrewai
28K Followers 1K Following Learning new things. Former Chief Research Officer at OpenAI, early exec at Palantir, early employee at Paypal.
j⧉nus @repligate
59K Followers 2K Following ↬🔀🔀🔀🔀🔀🔀🔀🔀🔀🔀🔀→∞ ↬🔁🔁🔁🔁🔁🔁🔁🔁🔁🔁🔁→∞ ↬🔄🔄🔄🔄🦋🔄🔄🔄🔄👁️🔄→∞ ↬🔂🔂🔂🦋🔂🔂🔂🔂🔂🔂🔂→∞ ↬🔀🔀🦋🔀🔀🔀🔀🔀🔀🔀🔀→∞
Albert Gu @_albertgu
18K Followers 88 Following assistant prof @mldcmu. chief scientist @cartesia_ai. leading the ssm revolution.
Casey Flint @FlintCasey
3K Followers 799 Following 🇦🇺🏄♀️ testimonials include “oddly calibrated” and “the right kind of weird”. @reflection_ai
Jason Wei @_jasonwei
99K Followers 640 Following ai researcher @meta superintelligence labs, past: openai, google 🧠
Noam Brown @polynoamial
92K Followers 859 Following Researching reasoning @OpenAI | Co-created Libratus/Pluribus superhuman poker AIs, CICERO Diplomacy AI, and OpenAI o3 / o1 / 🍓 reasoning models
gabriel @GabrielPeterss4
42K Followers 504 Following research sora at @OpenAI, previously at midjourney, swedish high school dropout
Adrian Cosma @adrianicosma
99 Followers 264 Following PostDoc Researcher @ IDSIA @supsi_ch 🇨🇭 Researcher @ UCS @upb1818 🇷🇴 The revolution will not be supervised.
Niloofar @niloofar_mire
7K Followers 2K Following Niloofar Mireshghallah — incoming asst. prof @LTIatCMU @CMU_EPP, RS in @AIatMeta, postdoc @uwcse, Ph.D. @ucsd_cse, former @MSFTResearch -Privacy, ML, NLP
Pawel Szczesny @PawelPSzczesny
393 Followers 683 Following Psychologist (with PhD in bioinformatics) meets AI. I build cognitive interfaces that steer AI and reshape human thinking. Prev: Biotech → VR → Nootropics → AI
Marco lin @MarcoLinSight
1K Followers 3K Following Science, human nature, technology, and AI. Proud nerd 🤓 Onboarding humanity to the AI-augmented future
jason liu @jxnlco
43K Followers 2K Following independent ai consultant, a16z scout, creator of instructor prev. @stitchfix @meta
Antoine Chaffin @antoine_chaffin
2K Followers 590 Following 28, French CS Engineer 💻, PhD in ML 🎓🤖 — Guiding generative models for better synthetic data and building multimodal representations @LightOnIO — 🇫🇷🇬🇧
Andrew Curran @AndrewCurran_
34K Followers 13K Following 🏰 - I write about AI, mostly. Expect some strange sights.
The Tor Project @torproject
455K Followers 2K Following 🦋 https://t.co/bX77YnBKhO We're a nonprofit defending privacy & freedom online. Download Tor Browser: https://t.co/ROuSDYAs6F
TNG Technology Consul... @tngtech
2K Followers 139 Following TNG, aka "The Nerd Group", is a consulting partnership focused on high end information technology, particularly AI. 906 employees, 99.9% academics, ~53% PhDs.
DSPy @DSPyOSS
11K Followers 50 Following An open-source declarative framework for building modular AI software. Programming—not prompting—LLMs via higher-level abstractions & optimizers.
am @attentionmech
13K Followers 1K Following
Yanick Fratantonio (@... @reyammer
7K Followers 726 Following Senior Research Scientist @ Google. Prev: Prof @EURECOM, Cisco Talos, DEFCON CTF Org. MOBISEC class: https://t.co/xPcWDxIcNo, Shellphish/NOPS hacker. Tweets mine.
Elie Bursztein @elie
62K Followers 127 Following AI Cybersecurity @Google & @DeepMind. Help advance AI cybersecurity capabilities and make AI safe & secure for all. @EtteillaOrg Art Foundation founder.
turboderp @turboderp_
771 Followers 37 Following
Ben Clavié @bclavie
6K Followers 1K Following regressing linearly on a daily basis. wife guy who does retrieval. research @mixedbreadai, prev answerdotai
zed @zmkzmkz
4K Followers 1K Following #1 paperclip maximizer fan, occasionally on x-games mode. I really, really like watching loss graphs go down
Linlu Qiu @linluqiu
883 Followers 668 Following PhD student @MIT_CSAIL | formerly AI resident @Google
MIT NLP @nlp_mit
4K Followers 52 Following NLP Group at @MIT_CSAIL! PIs: @yoonrkim @jacobandreas @lateinteraction @pliang279 @david_sontag, Jim Glass, @roger_p_levyTrends for United States
You might like
