DeeBee @self_supervised
Building dumb machines with supervised learning; AI researcher in practice | past 》MERL, SamsungResearch | Current 》 Credit and Fraud modeling using AI Joined November 2012-
Tweets2K
-
Followers95
-
Following585
-
Likes5K
LoRA in reinforcement learning (RL) can match full-finetuning performance when done right! 💡 A new @thinkymachines post shows how using 10x larger learning rates, applying LoRA on all layers & more, LoRA at rank=1 even works. We're excited to have collaborated on this blog!
LoRA in reinforcement learning (RL) can match full-finetuning performance when done right! 💡 A new @thinkymachines post shows how using 10x larger learning rates, applying LoRA on all layers & more, LoRA at rank=1 even works. We're excited to have collaborated on this blog! https://t.co/LFVVpv2bUK
I love reasoning. Emergence of reasoning is a necessary and sufficient condition for AGI and ASI. Once pre-training scales to reasoning, we (RL) will take it from there.
What if we could predict customer attrition with high accuracy w/o training on sensitive data? MIT’s "FairRAG" is fast & privacy-preserving. It uses differentially private synthetic data & retrieval-augmented generation — & outperforms traditional ML & LLMs:…
🎉Excited to share that LASeR has been accepted to #NeurIPS2025!☀️ RLHF with a single reward model can be prone to reward-hacking while ensembling multiple RMs is costly and prone to conflicting rewards. ✨LASeR addresses this by using multi-armed bandits to select the most…
🎉Excited to share that LASeR has been accepted to #NeurIPS2025!☀️ RLHF with a single reward model can be prone to reward-hacking while ensembling multiple RMs is costly and prone to conflicting rewards. ✨LASeR addresses this by using multi-armed bandits to select the most…
MIT just built PDDL-Instruct — designed to enhance LLMs’ symbolic planning capabilities. They teach LLMs logical chain-of-thought reasoning for PDDL, covering: ✅ Action applicability 🔄 State transitions 🔍 Plan validity Results: Up to 94% accuracy on benchmarks — a +66%…
Took me 4 days to read this blog, but totally worth it. A great detailed guide to post training. Why aren't many people talking about it. Kudos to @Han_Fang_ @karthikabinav for sharing their thoughts. tokens-for-thoughts.notion.site/post-training-…
Interesting! Hai Huang, @ylecun , and @randall_balestr have just proposed LLM-JEPA. It's the first JEPA-style training framework for language models, bridging the gap between vision’s embedding-space objectives and NLP’s generative ones. Results: - Outperforms standard LLM…
Also strongly recommend this paper on diversity reward in RL! The insights line up closely -- well worth reading together. https:// arxiv.org/abs/2509.15194 (Tencent) https:// arxiv.org/abs/2509.02534 (Meta) Not sure which diversity reward wins out 😀 (embedding vs…
Also strongly recommend this paper on diversity reward in RL! The insights line up closely -- well worth reading together. https:// arxiv.org/abs/2509.15194 (Tencent) https:// arxiv.org/abs/2509.02534 (Meta) Not sure which diversity reward wins out 😀 (embedding vs… https://t.co/HTu1NzJMdz
Excited to share that One-shot RLVR is accepted by @NeurIPSConf 2025, see you all at San Diego!
Excited to share that One-shot RLVR is accepted by @NeurIPSConf 2025, see you all at San Diego!
RL often cause 𝐞𝐧𝐭𝐫𝐨𝐩𝐲 𝐜𝐨𝐥𝐥𝐚𝐩𝐬𝐞: generations become shorter, less diverse, and brittle. A simple fix is 𝐝𝐢𝐯𝐞𝐫𝐬𝐢𝐭𝐲 reward to boost exploration. I use it in many of my projects — surprisingly effective! Details in our NEW paper: arxiv.org/abs/2509.15194
So true, this reminds me of time when we had to ensure multiple runs yield the same output.
Why can’t LLMs easily learn from hundreds of in-context examples like humans? 📊🤖 Meet MachineLearningLM—a continued-pretraining framework that gives LLMs strong in-context ML ability without losing general reasoning. Key ideas: - Pretrains on millions of structural causal…
That’s a solid outcome considering there was recent debate in SFT vs RL. It turns out that one compliment the other
That’s a solid outcome considering there was recent debate in SFT vs RL. It turns out that one compliment the other
RL may supersede SFT when its a non-deterministic 👌
RL may supersede SFT when its a non-deterministic 👌
i suspect society was better off with phone call culture than meeting culture
🚨 New paper alert 🚨 Using LLMs as data annotators, you can produce any scientific result you want. We call this **LLM Hacking**. Paper: arxiv.org/pdf/2509.08825
🚀 Excited to share our new survey paper on RL for Large Reasoning Models (LRMs)! Since early this year, our team has released several RL+LLMs works (PRIME, TTRL, SimpleVLA, MARTI, SSRL, HPT), covering dense rewards, self-evolution, embodied AI, multi-agent, tool learning, and…
HOW INFORMATION FLOWS THROUGH TRANSFORMERS Because I've looked at those "transformers explained" pages and they really suck at explaining. There are two distinct information highways in the transformer architecture: - The residual stream (black arrows): Flows vertically through…
HOW INFORMATION FLOWS THROUGH TRANSFORMERS Because I've looked at those "transformers explained" pages and they really suck at explaining. There are two distinct information highways in the transformer architecture: - The residual stream (black arrows): Flows vertically through… https://t.co/IVsNUq4hUT
Super excited to share our work on variable-length discrete diffusion models! We design a diffusion process that iteratively unmasks tokens and simultaneously adds new tokens. This enables variable-length discrete planning as well as flexible text generation and reasoning!
Super excited to share our work on variable-length discrete diffusion models! We design a diffusion process that iteratively unmasks tokens and simultaneously adds new tokens. This enables variable-length discrete planning as well as flexible text generation and reasoning!
Ever wonder what's really happening when we use RL to teach LLMs to reason? 🤔 The process is full of mysteries. 🤯 What causes those sudden "aha moments" in training? 📏 Why does better reasoning often lead to longer answers ("length-scaling")? 📉 Why does token entropy often…
Ever wonder what's really happening when we use RL to teach LLMs to reason? 🤔 The process is full of mysteries. 🤯 What causes those sudden "aha moments" in training? 📏 Why does better reasoning often lead to longer answers ("length-scaling")? 📉 Why does token entropy often…

Robert Scoble @Scobleizer
543K Followers 24K Following The best from ML/AI community | Ex-Microsoft, Rackspace, Fast Company | Wrote eight books about the future | Silicon Valley robots, holodecks, BCIs, & startups.
EV_BatteryBets🇺�... @Gaju413719
32 Followers 2K Following 15-30% Monthly | 2 High-Conviction Stocks.Short-Term Gains: 15-20% in Days/Weeks.DM "JOIN" for WhatsApp Alerts. Live Trade Signals • Market Analysis
ROIC_KING🇺🇸 @Dralalkeas080
60 Followers 2K Following 15-30% Monthly | 2 High-Conviction Stocks.Short-Term Gains: 15-20% in Days/Weeks.DM "JOIN" for WhatsApp Alerts. Live Trade Signals • Market Analysis
Bedarn @Bedarn595
4 Followers 281 Following
Yuetai Li @yuetai12575
227 Followers 571 Following Second year PhD @UW | Post-Training, LLM reasoning and synthetic dataset. https://t.co/cYAkbnCsCp Open to chat and collaborate!
wen👩🏻💻 @ds_wen_
25K Followers 12K Following 👋 senior data scientist. i do = data science + ai + lifelong learning with a growth mindset
Teardeaut @TeardeautAVFHu
56 Followers 2K Following Life has no limitations, except the ones you make.
Carlo Fahey @CarloFahey95200
114 Followers 5K Following
SrinivasanSS @SrinivasanSS52
7K Followers 5K Following Software Engineer from Bangalore Native of Hyderabad Ambivert Can converse in English, Telugu, Hindi
Theesleet @TheesleetiNH6S
61 Followers 2K Following
Neelansh Joshi @NeelanshJoshi
1 Followers 11 Following
Swecko @SweckogsTz
72 Followers 1K Following
PhyllisDillon @y4CcLlXVkOEF692
28 Followers 3K Following
Nicoleta Luca @NicoletaLuca15
30 Followers 445 Following
Telikesh @Telikesh130887
50 Followers 530 Following
Eldor Fozilov @EldorFozilov
10 Followers 179 Following MS Student in CSE @ UNIST | Researcher @ Machine Learning, Vision & Language (MVL) Lab
SwissCognitive, AI Ve... @SwissCognitive
146K Followers 100K Following We are committed to unleashing the power of AI in the business world. With our AI research, advisory, and ventures, we bring a blend of expertise to the Table.
hasan davulcu #progra... @hdavulcu
470 Followers 2K Following An advocate for #ethicalAI #explainableAI #programmableAI
F*ckedUpCompanies @fedupcompanies
97 Followers 576 Following View and participate in layoff discussions, share interview experiences, tell insider stories and more. Uncensored and 100% anonymous.
AI/CC ::Artificial In... @aicccreators
1K Followers 3K Following Global collective of artists, designers, and innovators reshaping creativity with AI. Empowering voices. Showcasing visions. The future of artistic expression.
Gio at QRC @GioQrc
294 Followers 5K Following Gio is a machine learning enthusiasts teaching at Queenstown Resort College (QRC) as well as a trading manager in the high-frequency trading world.
Acacia Mei Larson @AcaciaMei
382 Followers 4K Following
Reggie @ReggieCYoung
2K Followers 3K Following Chief of Staff @Lithic Previously @BlueVine Host of Fintech Layer Cake https://t.co/VMshJ0MoFR Tweets my own, not legal advice!
Jonathan Lorraine @jonLorraine9
7K Followers 6K Following Research scientist @NVIDIA | PhD in machine learning @UofT. Opinions are my own. 🤖 💻 ☕️
King Solomon @NeelNotFound
2 Followers 41 Following Start changing your mindset to make yourself successful. What you think, you get it.
Lev Telyatnikov @lev_telyatnikov
329 Followers 312 Following Moving forward Topological Deep Learning @SapienzaRoma (PhD)
Sonal Goyal @sonalgoyal
1K Followers 5K Following Open source entity resolution, data matching and deduplication
Nadeem Kafi @nakafi
195 Followers 3K Following Human Intensive Computing to Humanizing AI. Love to teach and travel.
Bhapi @bhapi_io
397 Followers 4K Following We’re bhapi, a safe #socialmedia app that’s free from toxic #content & ads, & never sells your #data. Available to download now.
Sizttt @Sizttt
25 Followers 148 Following ( 'ω')/プロ野球垢 🦅2021年からホークスを応援しています/ファン歴浅い留学生です/無言フォロー失礼します🙇♀️
Reliable Home VoIP Ph... @ReliablePhone
559 Followers 3K Following The reliable service of the home phone, keeping in touch with family and friends has been easier and more affordable.
Praneeth Vepakomma @proneat
1K Followers 1K Following Visiting Assistant Professor @MIT (IDSS), Assistant Professor @MBZUAI, Responsible AI/ML/Stats/Data Science, Collaborative, Private ML, Statistical Inference.
Pasquale Minervini @PMinervini
9K Followers 5K Following Research in ML/NLP at the U of Edinburgh (tenured faculty @InfAtEd @EdinburghNLP), Co-Founder @Miniml_AI, @ELLISforEurope Scholar, https://t.co/5dUI3EFexo
Protagion: Profession... @protagion
732 Followers 4K Following Talented professional? Wondering what's next in your career? Try our Career Support Guide to get you started 👇
Gautham Krishna Gudur @gauthamkrishna_
599 Followers 5K Following Ph.D. student @UTAustin and @wncg_UT. Prev. @BellLabs @ericsson. Mostly into efficient/human-centered AI/ML + multimodal sensing. Always GPU poor.
John C. Havens (he/hi... @johnchavens
16K Followers 12K Following Planet Positive 203O Lead. Founding ED, IEEE Global Initiative on AI Ethics. Book: Heartificial Intelligence. https://t.co/HgZJsmTJfT
Tom Goldstein @tomgoldsteincs
27K Followers 2K Following Professor at UMD. AI security & privacy, algorithmic bias, foundations of ML. Follow me for commentary on state-of-the-art AI.
SBX Robotics @sbxrobotics
168 Followers 73 Following Build better vision models with synthetic data. SBX improves computer vision models with training data that combines the best of real and synthetic images.
Jeremy Georges-Filtea... @jgeofil
232 Followers 3K Following Machine Learning Engineer and Computational Biologist.
Eleftherios Cheliouda... @EleChelioudakis
776 Followers 1K Following Co-founder of @Homo_Digitalis_ Tweets express personal views only. Cover pic by Tom Gauld, published at NewScientist.
Maria Luciana Axente @maria_axente
15K Followers 8K Following Founder and CEO #ResponsibleIntelligence
ZenML @zenml_io
2K Followers 2K Following Open-source pipeline framework that integrates all your ML tools. Get all your ML workflows running on any tooling stack with minimum effort.
Matt Janiga 💥♻�... @regulatorynerd
10K Followers 9K Following Legal nerd at Trustly! Former Stripe, SQ, BlueVine, Lithic, Cap One. Views are mine, and not my company's.
Sachin @sachdh
3K Followers 742 Following cooking reasoning models and agents at @AthenaAgentRL - a narrow intelligence lab
Gaurav Aggarwal @fooobar
12K Followers 1K Following VP & Chief AI Scientist@Reliance Jio, Volunteer@iSPIRT. Visiting Faculty @ ISB, 2x founder, Prev: Google Research, Fashiate, Yahoo Labs. Views personal!
Learning on Graphs Co... @LogConference
10K Followers 771 Following LoG is a new annual research conference that covers areas broadly related to machine learning on graphs and geometry, with a special focus on review quality.
Gautam Kedia @thegautam
4K Followers 528 Following Angel Investor @braintrustdata, @trybasis, @modal_labs, @_hex_tech & Ex Applied ML @stripe @lyft, @office, @bing.
Manling Li @ManlingLi_
8K Followers 760 Following Assistant Professor@NU, Amazon Scholar, Postdoc@Stanford, PhD@UIUC #NLP #CV Language+Vision/EmbodiedAI, Reasoning, Planning, Compositionality, Trustworthiness
Guohao Li 🐫 @guohao_li
8K Followers 1K Following Founder of @Eigent_AI / @CamelAIOrg. Finding the Scaling Laws of Agents. Prev UniofOxford, KaustVision, Intel, Kumo.
Yi-01.AI @01AI_Yi
9K Followers 73 Following A global company building generative AI LLM and applications
Fintech Meetup @FintechMeetup
3K Followers 347 Following Fintech Meetup is the future of fintech events. Join 6,000+ fintechers at Mandalay Bay, Las Vegas on March 30-April 1, 2026.
λux @novasarc01
20K Followers 2K Following tensor shepherd in a non-euclidean pasture | grazing on cuda cores
SemiAnalysis @SemiAnalysis_
37K Followers 18 Following
Eric Zhu @ekzhu
1K Followers 606 Following Building AI Agent Systems, Principal Architect at Microsoft
Z Fellows @zfellows
41K Followers 3 Following 1 week, $10,000. Your Fast-Track to Silicon Valley. Apply @ https://t.co/kc29ux5GVI
MetaGPT @MetaGPT_
9K Followers 219 Following The Multi-Agent Framework The World's First AI Dev Team: https://t.co/5ONAO5tqCq Discord: https://t.co/vlkPJDMSQZ @atoms_dev New Soon!
DeepSeek @deepseek_ai
972K Followers 0 Following Unravel the mystery of AGI with curiosity. Answer the essential question with long-termism.
Rajeev Chandrasekhar ... @RajeevRC_X
348K Followers 3K Following State President @BJP4Keralam | Chip Designer, Cellular pioneer,Tech Investor, Aviator, Former MP/Minister Govt of India 🇮🇳
Math Cafe @Riazi_Cafe_en
39K Followers 115 Following The goal of this account is to encourage mathematical thinking. Farsi account: @Riazi_Cafe
Congressman Greg Casa... @RepCasar
74K Followers 741 Following Representing #TX35 from East Austin to West San Antonio. Chair of @USProgressives. Labor organizer. Lover of migas and long runs, just not at the same time.
Mustafa Suleyman @mustafasuleyman
170K Followers 486 Following CEO, Microsoft AI | Author: The Coming Wave | Past: Co-founder, @InflectionAI & @GoogleDeepMind
Christopher Manning @chrmanning
152K Followers 229 Following Founder, @stanfordnlp and cs224n. Assoc. Director, @StanfordHAI. Prof. CS & Linguistics, @Stanford. GP @aixventureshq. Australian🇦🇺. Do #NLProc & #AI. 👋
himanshu @himanshustwts
16K Followers 3K Following ai research x product • @bhindiai • @anannas_ai • pods @groundzero_twt
Sebastian Nehrdich @SebastianNehrd2
8K Followers 2K Following CTO of the MITRA project @BAIR, UC Berkeley. Research in ancient Asian low resource languages, with focus on machine translation and semantic search.
Michael Hartung @DrMHartung
4K Followers 151 Following Radiologist - Creator https://t.co/SOPQmhvkpb - Editor @Radiopaedia - Abdominal CT course https://t.co/GB83ZyE5vc - educational purposes only
Adam Silverman (Hirin... @AtomSilverman
41K Followers 2K Following Agent Consulting: https://t.co/goijjAdzVD Agent Observability: @AgentOpsAI + https://t.co/JCGSBr6B0L Funding Agents: https://t.co/ufCO2UI9yU
Evan @StockMKTNewz
641K Followers 395 Following Free Stock Market News that is FAST, ACCURATE, CONSISTENT, and RELIABLE | Not Just Stock News | My Daily Stock Market Recap is the link in my bio ⬇️
BRICS News @BRICSinfo
1.7M Followers 3 Following We are an independent media company bringing you unparalleled coverage of all-things geo-politics & BRICS News in real-time.
Deepak @deepak4748
28K Followers 146 Following IIT Kanpur, Mechanical Engineer, SEBI unregistered, Stoploss is your friend. Collab @EquityBite
Google Pay Developers @GooglePayDevs
7K Followers 45 Following Official channel for the Google Pay Developers Network. Enabling developers to create better checkout experiences, and beyond.
Adyen Developers @AdyenDevs
2K Followers 80 Following Helping developers build payment processing solutions with @Adyen’s all-in-one financial technology platform. https://t.co/vVe6huXVyh
This Week in Fintech @twifintech
7K Followers 95 Following Welcome to the fintech fiesta. news: https://t.co/UHEN4EW1Yc fund: https://t.co/04zxq5B3Gk community: https://t.co/pMR1wqkSiS events: https://t.co/bXEJytfjhj
The Paperclip @Paperclip_In
83K Followers 115 Following A digital media house. Binding stories from India & beyond. History | Culture | Sports | Politics | Life
Santosh Ojha @DriftyAfter50
5K Followers 374 Following Drifting through Indian history, books, words, music, travel. Information sharing.
DotPe @DotPe_India
925 Followers 25 Following We are committed to empowering entrepreneurs with the power of digital technology and promoting online growth for every entrepreneur out there with a vision.
Nityānanda Miśra (�... @MisraNityanand
65K Followers 47 Following IIM Bangalore (2007). Finance professional. Onomastician. Author of fifteen books on Sanskrit & Hinduism. Editor of another fifteen. Please DM for name advice.
तहक्षी™... @yajnshri
170K Followers 466 Following Dharm ॥ History || Jyotish || Politics ॥ Posting Threads🧵Everyday ॥ नमश्चण्डिकायै
Hiring @XHiring
50K Followers 9 Following Try job search at https://t.co/7dVBTNIpt5 (available on web and iOS).
South Park Commons @southpkcommons
34K Followers 727 Following A community of technologists dedicated to helping each other explore, learn, and go from -1 to 0.
sophia goldberg @sophgoldb
5K Followers 2K Following probably too into payments 🪐 building @cheaperpayments, bestselling author, prev. @Adyen, IPE @ the LSE