Ananya Kumar @ananyaku
Researcher at @openai Previously PhD at Stanford University (@StanfordAILab) advised by Percy Liang and Tengyu Ma ananyakumar.wordpress.com Stanford, CA Joined June 2018-
Tweets313
-
Followers4K
-
Following469
-
Likes3K
Happy to share our work on preference learning methods for LLMs. Key insights: 1. Use more on-policy samples > off-policy samples 2. Contrastive DPO > Pref-FT. Also we provide insights on DPO's training mechanism. 3. Theoretical unification under mode-covering/seeking KL
Happy to share our work on preference learning methods for LLMs. Key insights: 1. Use more on-policy samples > off-policy samples 2. Contrastive DPO > Pref-FT. Also we provide insights on DPO's training mechanism. 3. Theoretical unification under mode-covering/seeking KL
Glad to have made a small contribution to it!
@dlwh has been leading the effort at @StanfordCRFM on developing levanter, a production-grade framework for training foundation models that is legible, scalable, and reproducible. github.com/stanford-crfm/… Here’s why you should try it out for training your next model:
Interestingly, pretraining on unlabeled source/target+finetuning doesn’t improve much over just supervised learning on source in iWildcam-WILDS. Correspondingly, the connectivity conditions on the success of contrastive pretraining for UDA (arxiv.org/abs/2204.00570) also fail!
Interestingly, pretraining on unlabeled source/target+finetuning doesn’t improve much over just supervised learning on source in iWildcam-WILDS. Correspondingly, the connectivity conditions on the success of contrastive pretraining for UDA (arxiv.org/abs/2204.00570) also fail!
What's the best way to use unlabeled target data for unsupervised domain adaptation (UDA)? Introducing Connect Later: pretrain on unlabeled data + apply *targeted augmentations* designed for the dist shift during fine-tuning ➡️ SoTA UDA results! arxiv.org/abs/2402.03325 🧵👇
Introducing Sora, our text-to-video model. Sora can create videos of up to 60 seconds featuring highly detailed scenes, complex camera motion, and multiple characters with vibrant emotions. openai.com/sora Prompt: “Beautiful, snowy…
# on shortification of "learning" There are a lot of videos on YouTube/TikTok etc. that give the appearance of education, but if you look closely they are really just entertainment. This is very convenient for everyone involved : the people watching enjoy thinking they are…
Since the launch of Medusa, we’ve been thrilled to see its adoption in TensorRT, TGI, and numerous open-source projects and companies. Today, we’re unveiling a technical report with fresh features! This includes the Medusa-2 recipe for full-model tuning, self-distillation for…
My group @PrincetonCS is looking for talented PhD students in machine learning systems (deadline Dec 15). If you're excited about fun math, new algorithms / model architectures for new capabilities, or efficient training / inference for LLMs and beyond, pls consider applying!
Quadratic attention has been indispensable for information-dense modalities such as language... until now. Announcing Mamba: a new SSM arch. that has linear-time scaling, ultra long context, and most importantly--outperforms Transformers everywhere we've tried. With @tri_dao 1/
We have reached an agreement in principle for Sam Altman to return to OpenAI as CEO with a new initial board of Bret Taylor (Chair), Larry Summers, and Adam D'Angelo. We are collaborating to figure out the details. Thank you so much for your patience through this.
❤️
i love the openai team so much
I'm on the faculty market! My goal is to build language systems that we understand deeply through discovery and by design, so we can precisely control them and treat their failures. Let's tackle this grand challenge of science and engineering together. nlp.stanford.edu/~johnhew/
Hong (@HongLiu9903) will give an oral presentation at #ICML2023 on this paper (Ballroom A, Jul 27, 16:04 HST). The poster presentation will be at at Poster Session 5 (Exhibit Hall 1), Jul 27, 10:30 HST. Please check them out!
Hong (@HongLiu9903) will give an oral presentation at #ICML2023 on this paper (Ballroom A, Jul 27, 16:04 HST). The poster presentation will be at at Poster Session 5 (Exhibit Hall 1), Jul 27, 10:30 HST. Please check them out!
When and how does known class help discover unknown ones? We provide the first theoretical analysis for Novel Class Discovery through spectral graph theory. (ICML23) Paper: openreview.net/pdf?id=JHodnaW… Video: youtube.com/watch?v=21_P-Q… (w/@zhmeishi, Yingyu Liang, @SharonYixuanLi) (1/n)
Percy Liang @percyliang
49K Followers 408 Following Associate Professor in computer science @Stanford @StanfordHAI @StanfordCRFM @StanfordAILab @stanfordnlp | cofounder @togethercompute | PianistJim Fan @DrJimFan
229K Followers 3K Following @NVIDIA Sr. Research Manager & Lead of Embodied AI (GEAR Lab). Creating foundation models for Humanoid Robots & Gaming. @Stanford Ph.D. @OpenAI's first intern.Behnam Neyshabur @bneyshabur
18K Followers 689 Following Senior Staff Research Scientist @GoogleDeepMind, Interested in reasoning w. LLMs, traveling & backpackingZachary Lipton @zacharylipton
59K Followers 2K Following Professor: CMU/@acmi_lab, CTO / CSO: @AbridgeHQ, Creator: @d2l_ai & https://t.co/QQt98VNLUp, Relapsing 🎷Jonathan Frankle @jefrankle
16K Followers 685 Following Chief Scientist, Neural Networks @Databricks via MosaicML. PhD @MIT_CSAIL. BS/MS @PrincetonCS. DC area native. Making AI efficient for everyone at @DbrxMosaicAIDan Roy @roydanroy
45K Followers 2K Following ML / AI researcher, emphasis on theory. Research Director and Canada CIFAR AI Chair, @VectorInst Professor, @UofT (Statistics/CS)rishi @RishiBommasani
4K Followers 2K Following Stanford CS PhD @StanfordCRFM @StanfordNLP @StanfordAILab @StanfordHAI Advisers: @percyliang @jurafsky Previous: @CornellCIS @clairecardie #FoundationModelsNoam Brown @polynoamial
34K Followers 611 Following Researching reasoning @OpenAI | Co-created Libratus/Pluribus, the first superhuman no-limit poker AIs | Co-created CICERO | PhD from @SCSatCMUGabriel Ilharco @gabriel_ilharco
4K Followers 1K Following Building cool things @xAI. Prev. PhD at UW, Google AIYann Dubois @yanndubs
4K Followers 1K Following PhD student @stanfordAILab | Prev: AI resident @metaai, @vectorinst, @CambridgeMLGAditya Grover @adityagrover_
8K Followers 411 Following CS Prof @UCLA. AI, ML, Climate. Prev: Postdoc @berkeley_ai, PhD @StanfordAILab, bachelors @IITDelhi.yobibyte @y0b1byte
15K Followers 2K Following Kurin ViTaly, senior research scientist @IsomorphicLabs, ML PhD from @UniofOxford on RL, Multitask learning & GraphsDimitris Papailiopoul.. @DimitrisPapail
11K Followers 965 Following prof @ wisconsin; thinking about transformers; learning in context; babas of Inez LilyNaomi Saphra @nsaphra
7K Followers 1K Following Waiting on a robot body. ML/NLP. All opinions are universal and held by both employers and family. Same username on every lifeboat off this sinking ship.Sara Hooker @sarahookr
39K Followers 7K Following I lead @CohereForAI. Formerly Research @Google Brain @GoogleDeepmind. ML Efficiency at scale, LLMs, @trustworthy_ml. Changing spaces where breakthroughs happen.Tengyu Ma @tengyuma
25K Followers 512 Following Assistant professor at Stanford; Co-founder of Voyage AI (https://t.co/wpIITHLgF0) ; Working on ML, DL, RL, LLMs, and their theory.Kayo Yin @kayo_yin
8K Followers 556 Following PhD student @berkeley_ai @berkeleynlp working on interpretability and signed languages. Former @msftresearch @deepmind @carnegiemellon @polytechnique. 🇫🇷🇯🇵Jeremy Cohen @deepcohen
4K Followers 867 Following PhD student in machine learning at Carnegie Mellon. The goal of my research is to turn deep learning into a real engineering discipline.Vansh @vanshg1729
17 Followers 60 FollowingAnurag Mishra @anuragm75160136
111 Followers 801 Following Building Scalable AI Applications | Senior Data Scientist @ EY | CSE Btech @ NIT MN | Linkedin: https://t.co/pCmSV6FmOeAgamdeep Singh @agammessi10
46 Followers 723 Following Trying to make a business out of RAG and training a foundational pose comparison model @ MOON lab, IISERB.Harsh Desai @dreamerharsh
1 Followers 3K FollowingAmit Raja Naik @AmitRajaNaik
428 Followers 2K Following AI Human @Analyticsindiam, The Belamy, Sector 6 (formerly AIM Daily XO)Chen Wu @ChenHenryWu
331 Followers 573 Following Ph.D. student @CMU_Robotics | Prev. undergrad @Tsinghua_Uni年孟 @c1gFE6DKJ9gtOeX
46 Followers 413 FollowingFiana | A Learner @fianfitr
171 Followers 5K Following @F_Nurfitriana | Psy. | Always go with the choice that scares you the most, because that's the one that is going to help you grow. | Bismillah CHRO.Fushoato @fushoato15711
0 Followers 164 FollowingPankaj Gupta @pankaj_ipynb
25 Followers 920 Following The English language can not fully capture the depth and complexity of my thoughts. So I'm incorporating Emoji into my speech to better express myself 😉.Ryan Chi @ryanandrewchi
17 Followers 42 Following Student Researcher, LLM Reasoning Team @GoogleDeepMind. Led @stanfordnlp's Alexa Prize Team to 1st Place (Science) at @AmazonScience's Socialbot Challenge.T J @tdj11100
321 Followers 4K Following TJ completed a Ph.D. in Physics and then moved into the tech world.Jacob Valdez @jvboid
2K Followers 8K Following surfing a 2nd order phase transition | building @HumanRobotsAI | [email protected] | +1.469.968.9490 | https://t.co/V5Odmcls1FEva Louise Marie Gabr.. @e681554349
8 Followers 3K FollowingDalton_lovegood @dalton_lovegood
45 Followers 263 Following Venture Capitalist @YunqiPartners; Research Assistant @Cambridge_Uni; Alumni @CUHKofficial; Focus on #AI #MachineLearning #SaaSVikram Dutt @vd_
808 Followers 7K FollowingAmir Samani @AmirSamani19
7 Followers 140 Followingzhitong gao @zhitong_gao
37 Followers 142 FollowingMa Sheen Uprising @MaSheenUprising
7 Followers 969 Following “The programme will take me a little while to run.” Fook glanced impatiently at his watch.LelCh @Yongsheng_Si
3 Followers 534 Following Aiming to be a researcher, writer, and startup founder.Makya @Makya12345678
6 Followers 962 FollowingLeonce Nshuti @LeonceNshuti
283 Followers 2K Following Data Engineer @Sony. Ex-UBS, Vanderbilt, Harvard. https://t.co/kOPPM3IA54. Google Scholar: https://t.co/UWXNmktdq0. Opinions my own.Nithish Kannen @NithishKannen
447 Followers 2K Following Languages @GoogleAI | Ex- @AmazonScience London, @IBMResearch | @CNERG @IITKgp | #NLPProcMikaStars★ @MikaStars39_
173 Followers 614 Following Second year B.A. / B.S. in @ZJU_China Prev: Bsc in @Polytechnique Devoted in LLM Architecture & InterpretabilityDhruv Patel @dhruvpatel2012
91 Followers 961 Following MS Robotics @GeorgiaTech @ICatGT | Prev - Google Summer of Code'23 @letsunifyai & Robotics @iiit_hyderabad Passionate about Robotics,AI & Neuroscience.Goutham kumar @goutham0205
24 Followers 427 Following ML engineer at Amazon | Make GPUs go Brrrrr...__vaibhav__ @Sillychap101
105 Followers 3K Following Computer Science and Mathematics undergrad | IIITDPeace=Progress @peaceisprogress
92 Followers 2K Following World peace, one world is true progress of human civilization.Avyay M C @Avyay_M_C
195 Followers 1K FollowingMuhammad Suleman Asif @msulemanas57411
243 Followers 6K Following Current :-Senior Analytic Consultant @wellsfargo. Previously :-Founder of WIFC (Without Internet free Call). I go by Muhammad.Hasnat Sajid Naseer @hasnat_sn
0 Followers 107 FollowingOnur Ünlü @onur_unlu_01
22 Followers 358 Following PhD Student @Cornell | Multi-agent RL Theory, Game theory | Previously @BilkentUniv.mus a h @the_shrouded
45 Followers 1K FollowingLevi Githaiga @CodeTitanium
12 Followers 376 FollowingNino Scherrer @ninoscherrer
581 Followers 2K Following Interested in rigorous LLM evals, synthetic data, causality, robustness & AI safety/ethics | RS @PatronusAI | Prev: @VectorInst, @Mila_Quebec, @MPI_IS, @ETH_enSubodh Kumar @Ksub_27
1 Followers 351 Following Senior Year Undergrad @iitroorkee. Previously AI Research intern @IBMResearch. Loves Talking about AI, Software Development and ...Bruce Cheng @BruceChengXD
11 Followers 28 FollowingManu Gaur @gaur_manu
71 Followers 661 Following Doing matrix multiplications @IIIT_Hyderabad | Applied Physics DCE’23 | Prev @amazon, UTS,SydneyMuktin Cobus @MuktinCobus
4 Followers 168 Following🌎 @ascetic_one
30 Followers 363 FollowingLawrence Phillips @LawrencePh87045
28 Followers 30 FollowingNjdeh Satourian @satourian
299 Followers 1K FollowingLang Xu @xulang7
18 Followers 230 FollowingPercy Liang @percyliang
49K Followers 408 Following Associate Professor in computer science @Stanford @StanfordHAI @StanfordCRFM @StanfordAILab @stanfordnlp | cofounder @togethercompute | PianistJim Fan @DrJimFan
229K Followers 3K Following @NVIDIA Sr. Research Manager & Lead of Embodied AI (GEAR Lab). Creating foundation models for Humanoid Robots & Gaming. @Stanford Ph.D. @OpenAI's first intern.Gautam Kamath @thegautamkamath
44K Followers 504 Following Assistant Prof of CS @UWaterloo, Faculty @VectorInst, Canada @CIFAR_News AI Chair. Co-EiC @TmlrOrg. I lead @TheSalonML. Privacy, robustness, machine learning.Behnam Neyshabur @bneyshabur
18K Followers 689 Following Senior Staff Research Scientist @GoogleDeepMind, Interested in reasoning w. LLMs, traveling & backpackingZachary Lipton @zacharylipton
59K Followers 2K Following Professor: CMU/@acmi_lab, CTO / CSO: @AbridgeHQ, Creator: @d2l_ai & https://t.co/QQt98VNLUp, Relapsing 🎷Jonathan Frankle @jefrankle
16K Followers 685 Following Chief Scientist, Neural Networks @Databricks via MosaicML. PhD @MIT_CSAIL. BS/MS @PrincetonCS. DC area native. Making AI efficient for everyone at @DbrxMosaicAIDan Roy @roydanroy
45K Followers 2K Following ML / AI researcher, emphasis on theory. Research Director and Canada CIFAR AI Chair, @VectorInst Professor, @UofT (Statistics/CS)Sergey Levine @svlevine
79K Followers 122 Following Associate Professor at UC Berkeley Co-founder, Physical Intelligencerishi @RishiBommasani
4K Followers 2K Following Stanford CS PhD @StanfordCRFM @StanfordNLP @StanfordAILab @StanfordHAI Advisers: @percyliang @jurafsky Previous: @CornellCIS @clairecardie #FoundationModelsNoam Brown @polynoamial
34K Followers 611 Following Researching reasoning @OpenAI | Co-created Libratus/Pluribus, the first superhuman no-limit poker AIs | Co-created CICERO | PhD from @SCSatCMUGabriel Ilharco @gabriel_ilharco
4K Followers 1K Following Building cool things @xAI. Prev. PhD at UW, Google AINeurIPS Conference @NeurIPSConf
111K Followers 35 Following New Orleans, Dec 10-16, 23. https://t.co/ga8aOw615g Tweets to this account are not monitored. Please send feedback to [email protected].Yann Dubois @yanndubs
4K Followers 1K Following PhD student @stanfordAILab | Prev: AI resident @metaai, @vectorinst, @CambridgeMLGAlex Ratner @ajratner
5K Followers 545 Following @SnorkelAI @uwcse / prev @StanfordAILab – Interested in data management systems for machine learning, weak supervision, and impactful applications.Christopher Manning @chrmanning
126K Followers 115 Following Director, @StanfordAILab. Assoc. Director, @StanfordHAI. Founder, @stanfordnlp. Prof. CS & Linguistics, @Stanford. IP @aixventureshq. 🇦🇺 Do #NLProc & #AI. 👋Irina Rish @irinarish
9K Followers 994 Following prof UdeM/Mila; Canada Excellence Research Chair; AAI Lab head https://t.co/UzlrC7ZrGF; INCITE project PI https://t.co/0rV7szd7rH; CSO https://t.co/XDhj6MEtUjAditya Grover @adityagrover_
8K Followers 411 Following CS Prof @UCLA. AI, ML, Climate. Prev: Postdoc @berkeley_ai, PhD @StanfordAILab, bachelors @IITDelhi.Aleksander Madry @aleks_madry
31K Followers 165 Following Head of Preparedness at OpenAI and MIT faculty (on leave). Working on making AI more reliable and safe, as well as on AI having a positive impact on society.Jacob Austin @jacobaustin132
3K Followers 798 Following @Google @DeepMind researcher. AI for math and science. Coding. Gemini. I also play piano. NYC. Opinions my ownBanghua Zhu @BanghuaZ
2K Followers 785 Following PhD @Berkeley_EECS, statistics, info theory, LLM, RL, Human-AI Interactions.Elizabeth Yang @eytyang
768 Followers 677 Following technical staff @openai, previously theory @berkeleyeecs | fan of graphs, crosswords, turtles, bad puns, running, and Survivor, among other thingsHelen Qu @_helenqu
227 Followers 66 Following supernovae / cosmology / machine learning ✨ incoming research fellow @FlatironCCA, prev: PhD @physatpenn ‘24, BSE @CIS_Penn '17Lawrence H. Summers @LHSummers
326K Followers 706 Following Charles W. Eliot University Professor and President Emeritus at Harvard. Secretary of the Treasury for President Clinton and Director of NEC for President ObamaStephen Wright @madsjw
2K Followers 1K FollowingVoyage AI @Voyage_AI_
2K Followers 164 Following Building embedding/vectorization models, customized for your domain and company, for better retrieval quality https://t.co/MEAhTpBQqdSabri Eyuboglu @EyubogluSabri
611 Followers 261 Following Computer Science PhD student @Stanford working with @HazyResearch and @james_y_zou 🪬Lukasz Kaiser @lukaszkaiser
7K Followers 47 FollowingGuillaume Lample @GuillaumeLample
37K Followers 648 Following Cofounder & Chief Scientist https://t.co/hLfvKLkFHd (@MistralAI). Working on LLMs. Ex @MetaAI | PhD @Sorbonne_Univ_ | MSc @CarnegieMellon | X11 @PolytechniqueTogether AI @togethercompute
27K Followers 303 Following The future of AI is open-source. Let's build together.Mistral AI @MistralAI
90K Followers 0 Following Fast, open-source and secure language models. Join us https://t.co/INALdNGvCPEric @ericmitchellai
4K Followers 487 Following I like AI & music. Working on making LLMs easier & safer to use. Final year PhD student at Stanford advised by Chelsea Finn & Chris Manning.Joshua Achiam ⚗️ @jachiam0
14K Followers 944 Following Human. Trying to make safe alchemy machines. Thinking about humanist alchemism (h/alc ⚗️, maybe). Main author of https://t.co/cKuSh210l1Brydon Eastman @brhydon
877 Followers 729 Following Mathematician (Heavy on the ish) Research Scientist @OpenAI, Previously Ph.D. @WaterlooMath. ☕ //🚴//🧗♂️ // 🤔➡️💻Hongyu Ren @ren_hongyu
3K Followers 595 Following Research Scientist @openai. CS PhD @stanford. Previously @apple, @googleai and @nvidiaai. I train language models.Chelsea Sierra Voss @csvoss
10K Followers 1K Following engineeress ✨ Member of Technical Staff @openai serious play // notice your curiositySean Metzger @SeanMetzger5
123 Followers 135 Following @OpenAI Prev @ChangLabUCSF/@UCBerkeley. EE MS, BS @Stanford.Miles Brundage @Miles_Brundage
43K Followers 10K Following Policy research at @openai. I mostly tweet about AI, animals, and sci-fi. He/him. Views my own.Ilya Kostrikov @ikostrikov
8K Followers 614 Following Researcher @OpenAI, previously @Postdoc at UC Berkeley @berkeley_ai, PhD in CS @CILVRatNYUTao Xu @txhf
6K Followers 888 Following Learning Machine at OpenAI, previously Airbnb, Quora, Facebook and Microsoft.Jerry Tworek @MillionInt
7K Followers 281 Following I teach programs how to program @ OpenAI | putting the ball in the damn hoop - @jacobmenickJong Wook Kim 💟 @_jongwook_kim
4K Followers 467 Following Member of Technical Staff @OpenAI, authored CLIP and Whisper; previously at @nyuMARL, @SpotifyResearch, @pandoramusic, @kakaocorpglobal, and @NCSOFTTony Lee @tonyh_lee
401 Followers 86 Following Incoming PhD Candidate @StanfordAILab @StanfordNLP @Stanford. Author of HELM + extensions (https://t.co/f9UOXPWkpR). Prev: Research Eng at @StanfordCRFM.Song Mei @Song__Mei
1K Followers 548 Following Assistant Professor at UC Berkeley, Department of Statistics and EECS. Researcher working on foundations of generative AI.Jeff Wu @WuTheFWasThat
258 Followers 245 FollowingMark Chen @markchen90
10K Followers 245 Following Head of Frontiers Research at OpenAI. Coach for the USA IOI Team.Ishaan Gulrajani @__ishaan
3K Followers 473 Following Hi! I’m a machine learning researcher @openai. Previously @stanford @facebook @google @mila_quebecKangwook Lee @Kangwook_Lee
2K Followers 666 Following Assistant Professor, ECE, UW-Madison / Leading deep learning research @ KRAFTONChris Albon @chrisalbon
86K Followers 2K Following Director of Machine Learning at the Wikimedia Foundation. We host Wikipedia.Quanquan Gu @QuanquanGu
9K Followers 2K Following Professor @UCLA | Head of AIDD, ByteDance Research | Recent work: Self-play fine-tuning (SPIN) | Opinions are my ownGreg Yang @TheGregYang
53K Followers 660 Following Cofounder https://t.co/SpHbO7FZNV. Morgan Prize Honorable Mention 2018. Developing the theory of #TensorPrograms and the practice of scaling #neuralnetworks.Arvind Narayanan @random_walker
119K Followers 412 Following Princeton CS prof. Director @PrincetonCITP. I write about the societal impact of AI, tech ethics, & social media platforms. BOOK: AI Snake Oil. Views mine.Yiyou Sun @YiyouSun
314 Followers 132 Following Researcher in Open-world Machine Learning. Graduated from the University of Wisconsin-Madison, advised by @SharonYixuanLi.Priya Sundaresan @priyasun_
734 Followers 338 Following CS PhD student @Stanford, prev. BS/MS @Berkeley_EECS | learning from humans and teaching robots 🤖Caroline Choi @carolineschoi
40 Followers 60 Following CS + Math @Stanford. Researcher @StanfordAILab. Prev @Meta, @SnapSeong Joon Oh @coallaoh
1K Followers 870 Following Leading the STAI group at the University of Tübingen https://t.co/qrSPDDcdOy Advising @ParameterLab.Pavel Izmailov @Pavel_Izmailov
6K Followers 1K Following Incoming Assistant Professor @nyuniversity 🏙️ Previously @OpenAI #StopWar 🇺🇦Time to transition my jobs to this gpu
First @nvidia DGX H200 in the world, hand-delivered to OpenAI and dedicated by Jensen "to advance AI, computing, and humanity":
Correction: I did the math wrong (not considering log/log scales). Sophia is ~1.6x times more efficient than Adam (thanks for pointing out @tengyuma).
Putting together all the experiments, scaling looks very healthy. We're slightly more than 1.2x more efficient with Sophia vs. AdamW at scale. Doesn't get close to 2x the original paper stated but also original paper used a lot less compute. Seems like free lunch!
First @nvidia DGX H200 in the world, hand-delivered to OpenAI and dedicated by Jensen "to advance AI, computing, and humanity":
Some personal updates: I joined OpenAI a few months ago, working on all things robustness/safety/privacy. Also, we are working to publish more of our safety work. See my first project here below, where we make initial progress on prompt injections and other attacks!
Introducing the Instruction Hierarchy, our latest safety research to advance robustness for prompt injections and other ways of tricking LLMs into executing unsafe actions. More details: arxiv.org/abs/2404.13208
Happy to share our work on preference learning methods for LLMs. Key insights: 1. Use more on-policy samples > off-policy samples 2. Contrastive DPO > Pref-FT. Also we provide insights on DPO's training mechanism. 3. Theoretical unification under mode-covering/seeking KL
Many LLM fine-tuning methods. Unclear what you should use & why? In our new paper, we did an extensive study of on-policy RL, supervised & offline contrastive methods (DPO, IPO) to answer this... 🧵⬇️ On-policy > offline, mode-seeking > mode-covering understanding-rlhf.github.io
We often hear about the theory-practice gap. At this workshop we will take a thorough look at this. Is there a gap? What is the nature of the gap? Who made it? Is it good to have the gap? If not, how to close it? I think this is super important for the healthiness of the field!
🧵 Thrilled to announce the #ICML RL workshop 'Aligning RL Experimentalists and Theorists'! We will have several talks and a panel delivered by a super lineup of speakers: @white_martha, @ShamKakade6, @yayitsamyzhang, Dylan Foster, Niao He, @svlevine, and @MengdiWang10. 1/3
Introducing Arena-Hard – a pipeline to build our next generation benchmarks with live Arena data. Highlights: - Significantly better separability than MT-bench (22.6% -> 87.4%) - Highest agreement to Chatbot Arena ranking (89.1%) - Fast & cheap to run ($25) - Frequent update…
It's a great week for open source AI! Data is among the highest impact work to push the field forward. Bravo to 🤗
Llama3 was trained on 15 trillion tokens of public data. But where can you find such datasets and recipes?? Here comes the first release of 🍷Fineweb. A high quality large scale filtered web dataset out-performing all current datasets of its scale. We trained 200+ ablation…
I got tenure! It was fitting that I got to celebrate with the lab right after the news. Working together for the last 6.5 years has been a blast.
Congratulations to Brenden! I remember when we offered him the role of Moore-Sloan data science fellow at NYU. A super-postdoc designed to foster independent, multidisciplinary, cutting-edge research. We were so happy that he accepted and that he stayed to join the faculty. 🎉
I got tenure! It was fitting that I got to celebrate with the lab right after the news. Working together for the last 6.5 years has been a blast.
Sebastien Bubek giving a keynote talk in the launch of our Generative AI center in UT Austin. "The small language models revolution"
Thanks Dr. @SebastienBubeck for this exciting talk. The insights are super new interesting.
Sebastien Bubek giving a keynote talk in the launch of our Generative AI center in UT Austin. "The small language models revolution"
Llama3 8B and 70B are out, with pretty exciting results! * The ~400B is still training but results already look promising. * Meta's own Chat interface is also live at meta.ai * TorchTune integration is shortly going live: github.com/pytorch/torcht…
It’s here! Meet Llama 3, our latest generation of models that is setting a new standard for state-of-the art performance and efficiency for openly available LLMs. Key highlights • 8B and 70B parameter openly available pre-trained and fine-tuned models. • Trained on more…
Meta released Llama 3 on my birthday! 🎂 Best present ever, thanks Meta! 😀
Congrats to @AIatMeta on Llama 3 release!! 🎉 ai.meta.com/blog/meta-llam… Notes: Releasing 8B and 70B (both base and finetuned) models, strong-performing in their model class (but we'll see when the rankings come in @ @lmsysorg :)) 400B is still training, but already encroaching…
ICPC Alumni Series talk by Jakub Pachocki, Director of Research at OpenAI (and my former postdoc!). We got to learn about “Scaling up Deep Learning”. #icpcwfluxor
I’m a Dr now!! so grateful to my advisor and all my collaborators, friends, and family for supporting me every step of the way 🥰
Hmmm, I have a feeling this plot might need an overhaul rather soon🤣. I guess phi-2 was the lower left part of the triangle. I wonder what those guys have been up to in the last 6 months? 🤔
Congrats Jiefeng. Jiefeng was co-advised by me and Prof. Yingyu Liang and did impressive work during his PhD at @WisconsinCS Eager to see his research trajectory at @Google
Thrilled to announce that I've joined Google as a Research Scientist! I'm excited to dive into LLM research and contribute to cutting-edge developments in the field. Looking forward to this new chapter! 🚀 #NewJob #ResearchScientist #Google