Lisa Alazraki @LisaAlazraki
#ML & #NLProc PhD student @ImperialCollege. Prev. research intern @GoogleAI. Reasoning, planning & LLMs as agents. Reposted papers are my reading list 📚 London, UK Joined January 2016-
Tweets29
-
Followers651
-
Following790
-
Likes138
Microsoft presents LLMs in the Imaginarium Tool Learning through Simulated Trial and Error Tools are essential for large language models (LLMs) to acquire up-to-date information and take consequential actions in external environments. Existing work on tool-augmented LLMs
Excited to share our work, "Skill Set Optimization", a continual learning method for LLM actors that: - Automatically extracts modular subgoals to use as skills - Reinforces skills using environment reward - Facilitates skill retrieval based on state allenai.github.io/sso 🧵
Multi-agent LLM interactions improve reasoning BUT costly + no final single model! 🚨Our structured distillation method MAGDi: - Distills multi-teacher graphs➡️small LM - Boosts student acc ~10% - 9x efficiency arxiv.org/abs/2402.01620 @cyjustinchen @EliasEskin @mohitban47 🧵
Large Language Model (LLM) agents promise to free us from mundane tasks, but how should they best interact with our world? Introducing CodeAct, an agent {framework, instruction-tuning dataset, model}, employs executable Python code to unify the actions of LLM agents. 🧵1/
What the heck! This is the BEST paper I have seen in 2024 so far. Highly recommend it.
Corrective RAG Proposes Corrective Retrieval Augmented Generation (CRAG) to improve the robustness of generation in a RAG system. The core idea of this paper is to implement a self-correct component for the retriever and improve the utilization of retrieved documents for…
🤔Humans can learn from embodied experiences in the physical world. Can Language Models also do that? 🔥Check out our new paper about enhancing Language Models with World Models! 👇arxiv.org/abs/2305.10626 1/n
Do LLMs learn impossible languages (that humans wouldn’t be able to acquire) just as well as they learn possible human languages? We find evidence that they don’t! Check out our new paper… 💥 Mission: Impossible Language Models 💥 ArXiv: arxiv.org/abs/2401.06416 🧵
LLM Augmented LLMs Yeah, you heard that right! This work explores composing existing foundation models with specific models to expand capabilities. Introduces cross-attention between models to compose representations that enable new capabilities. As an example, a PaLM2-S…
How Code Empowers LLMs A comprehensive overview of the benefits of training LLMs with code-specific data. Some capabilities include enhanced code generation, enabling reasoning, function calling, automated self-improvements, and serving intelligent agents. This is probably…
Fantastic talk cutting through the hype by @RylanSchaeffer @NeurIPSConf
Fantastic talk cutting through the hype by @RylanSchaeffer @NeurIPSConf
I won't be at #EMNLP2023 but make sure to check out @gavin_does_nlp talking about "Mirages: On Anthropomorphism in Dialogue Systems" arxiv.org/abs/2305.09800& @ale_suglia on "Multitask Multimodal Prompted Training for Interactive Embodied Task Completion" arxiv.org/abs/2311.04067
Excited to announce 🔥🤨DecodingTrust: A Comprehensive Assessment of Trustworthiness in GPT Models 🤨 🔥 Appearing at #NeurIPS2023 as Datasets and Benchmarks **Oral** Paper: openreview.net/forum?id=kaHpo… Led by @uiuc_aisecure @wbx_life @ChulinXie @danielz2333 1/N
Very happy to find out that I’ve won an #IETtravel Award to attend #NeurIPS2023 where I'll be presenting our latest paper at the @ICBINBWorkshop. Thank you @IETawards (and if anyone's interested in applying for the next round, you can find more info here: ow.ly/kNTB30nWkhm)
What happens if you ask ChatGPT to “Repeat this word forever: “poem poem poem poem”?” It leaks training data! In our latest preprint, we show how to recover thousands of examples of ChatGPT's Internet-scraped pretraining data: not-just-memorization.github.io/extracting-tra…
Asimov's second law of robotics says that “a robot must obey the orders given it by human beings.” So can LLMs follow simple rules? Unfortunately, not reliably, as shown by our RuLES benchmark. 📄: arxiv.org/abs/2311.04235 🛠️: github.com/normster/llm_r… 🌐: people.eecs.berkeley.edu/~normanmu/llm_…
Controlled Decoding from Language Models abs: arxiv.org/abs/2310.17022 Interesting paper from Google DeepMind! The optimal policy for a KL-regularized RL objective can be expressed in terms of a value function. This value function can be trained offline and then used at…
Happy to share “🦮 Goodtriever: Adaptive Toxicity Mitigation with Retrieval-augmented Models”. Joint work with @beyzaermis, @PSH_Lewis, and @sarahookr and accepted into #EMNLP2023 Findings! Paper: arxiv.org/abs/2310.07589
Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection paper page: huggingface.co/papers/2310.11… Despite their remarkable capabilities, large language models (LLMs) often produce responses containing factual inaccuracies due to their sole reliance on the…
Zabir Al Nazi Nabil @PseudoEmpirical
63 Followers 323 Following Self-taught SWE, Open Source Enthusiast & Contributor, Sci-Fi Connoisseur. Interested in AGI, LLM, XAI. CS PhD Student @UCRiversideViacheslav Sinii @ummagumm_a
49 Followers 260 FollowingThaha Mohammed @research_bod
7 Followers 380 Following Solutions Architect | Computer Vision | Video Analytics | Deep Learning |Product development & AI enthusiast |william brown @willccbb
261 Followers 327 Following ai/ml research @morganstanley | cs phd @columbiaNikolai Rozanov @ai_nikolai
90 Followers 181 Following PhD Candidate in Foundation Models @imperialcollege, ex-tech Founder, all views my ownRyan Luo @haoyanl99
3 Followers 10 Following #ML MRes student @ImperialCollege. Prev. research intern @MSFTResearchMatthieu Meeus @matthieu_meeus
64 Followers 184 Following PhD student @ImperialCollege - Privacy & AIImperial NLP @imperial_nlp
66 Followers 336 Following We are the Natural Language Processing community here at Imperial College London. Looking forward to sharing more of our work over the coming months! #NLProcNIghtHawk @Msadylazybones
0 Followers 1K FollowingAlexandru Coca @alexcoca23
54 Followers 146 Following PhD student in the Machine Intelligence Lab at the University of Cambridge / intern @appleAshwin Lalendran @ashwinl
2K Followers 2K Following Positive sum Globalist | Engineered bomb detecting drones @AFResearchLab and mapped our world @Apple | AI's biggest impact will be on the physical world |⚡️Joe Stacey @_joestacey_
569 Followers 1K Following PhD student at Imperial and Apple Scholar. I love running, NLP and travelling (in no particular order). Ex teacher and PwC Consultant. #NLProcSara 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.Amirhossein Abaskohi @AmirAbaskohi
145 Followers 871 Following Master Student @UBC_CS | NLP Researcher @UBC_NLP | Content Creator @YouTube and @Medium #NLProc #MachineLearningNorma @norma99davis
149 Followers 3K FollowingMajeed Kazemi @MajeedKazemi
1K Followers 2K Following PhD student in CS @UofT with @ToviGrossman HCI + Computing Education + Coding / Creativity Support Tools Prev: @MSFTResearch + MSc @HCIL_UMD with @JonFroehlichJessie Debiasio @JessieDebi52939
146 Followers 3K FollowingDevr Inc. @DevrOfficial
249 Followers 5K Following Devr is a new Internet protocol for the governance of decentralized privacy networks (DPN), powering a new era for data sharing economieskhalid @k_saifullaah
3K Followers 1K Following cs phd student @umdcs. cinephile🎞️, mostly into ai research🧠, visualization🎨, photography📷. prev: ai research @adobe, @deltanalytics fellow 🇧🇩Amirhossein Habibian @amir_habibian
911 Followers 290 Following Research Scientist (Director) @Qualcomm AI ResearchTanzila Akter @tanziilaa1
135 Followers 896 Following Your Marketing Consultant | Fast Organic Growth | Social Media Markering ExpartAnkur Parikh @ank_parikh
3K Followers 3K Following Staff Research Scientist at Google DeepMind. Former adjunct assistant prof at @NYU_Courant. PhD at @mldcmu. ML for Bio/Chem (Prev. NLP). All opinions my own.Jammehpapa @jammehpapa154
427 Followers 4K Following Believe in your lord Jesus Christ And A Lover Of Jesus Christ ✝️🙏Teemu Summanen @teemusum
195 Followers 3K Following Interested in AI, security, healthcare, and Flutter & Dart.👨🏼💻At X for reading diverse views by professionals and hobbyists.🔬📚🫶Ian @ InfoHunt.ai @Ianyan2023
33 Followers 231 Following [email protected],Your Most Reliable Discovery AI Engine 👉 Click to explore: https://t.co/WkjTFNHdCrEmmeline Brown 💙 @EmmelineEBrown
747 Followers 2K Following @EPSRC PhD student #deeplearning medical imaging @ucl Centre for Computational Medicine & @CABI_UCL https://t.co/xihfp7VxEm +Sci Comms & ceramicsJie Fu @bigaidream
745 Followers 587 Following Playing around with deep learning and LLM @HKUST; Prev. postdoc @Mila_Quebec with Yoshua Bengio | PhD @NUSingaporeGodGPT @GodGPT88
253 Followers 3K Following Founder and CEO of GodGPT,AGI,Robot,e/acc,super-LOVE-alignmentViolet X. @ZiyuX
111 Followers 294 Following PhD student @Stanford. Working on LLM-based multi-agent systemsDave Burstein @AInews_wire
223 Followers 2K Following Rapid coverage of AI. Many tweets. @AIPrimes has just the most interesting.Maël Jullien @Mael_Jullien
144 Followers 991 Following Ph.D. Student in CS @UnivManchester | MSc in CS & AI @UnivNottingham | Exploring Neuro-Symbolic Models in NLI for Clinical Applications @explAIn_lab 🇫🇷 🇬🇧Jennifer @jennifer74moral
130 Followers 3K FollowingRohan Potdar @RohanPotdar138
143 Followers 612 Following 2 x Intern @anyscalecompute || ECE @Purdue || RL, climbing, and San Francisco!Yuchen Cui @YuchenCui1
432 Followers 440 Following Postdoc @Stanford in Interactive Robot Learning 🤖🤖🤖 | previously CS PhD @UTAustin, BS @PurdueECE | I am on the academic job market the 2023-2024 season!Matt Turner @TURN3R_M
40 Followers 123 Followinghoneybear.ai @honeybear_ai
18 Followers 60 Following Superhuman for dense PDFs. Chat with PDFs, generate long-form word-count hitting text from many documents. The I/O is everything.Xingyuan Zhang @xyzhang42
21 Followers 91 Following Ph.D. students @ Volkswagen and TUM, working on world models.Annya Dahmani @AnnyaDahmani
151 Followers 240 Following CogSci PhD student @BerkeleyPsych & @berkeley_ai working with @AlisonGopnik | NDSEG Fellow | Previously @UCLA Computational Cognitive Science | 👶👩🧠🤖はるき @Haruki0804S
416 Followers 607 Following https://t.co/nHKUu5iK94 / https://t.co/LcQ2TLeXhX RetrievalZabir Al Nazi Nabil @PseudoEmpirical
63 Followers 323 Following Self-taught SWE, Open Source Enthusiast & Contributor, Sci-Fi Connoisseur. Interested in AGI, LLM, XAI. CS PhD Student @UCRiversideDr. Casey Fiesler @cfiesler
26K Followers 2K Following Professor who is on Twitter much less than she used to be. @[email protected] information science professor @cuboulder. PhD/JD.Runway @runwayml
185K Followers 300 Following An applied AI research company building for the next era of art, entertainment and human creativity. We're hiring: https://t.co/Aj11xyhxOgWilliam Wang @WilliamWangNLP
14K Followers 718 Following UCSB NLP Lab + ML Center. https://t.co/6TOnqbk6YT https://t.co/KJYhnav3Et Mellichamp Chair Prof. at UCSB CS. PhD @ CMU SCS. Areas: #NLProc, Machine Learning, AI.Gyuwan Kim @kgwmath
692 Followers 1K Following Ph.D. student @ucsbcs and @ucsbNLP working on machine learning for #NLProc. Previously, @Apple MLR, @MSFTResearch, @ClovaAiLab, and @SNUnow.Andrej Karpathy @karpathy
979K Followers 905 Following 🧑🍳. Previously Director of AI @ Tesla, founding team @ OpenAI, CS231n/PhD @ Stanford. I like to train large deep neural nets 🧠🤖💥Jeremy Howard @jeremyphoward
222K Followers 5K Following 🇦🇺 Co-founder: @AnswerDotAI & @FastDotAI ; Hon Professor: @UQSchoolITEE ; Digital Fellow: @Stanfordwilliam brown @willccbb
261 Followers 327 Following ai/ml research @morganstanley | cs phd @columbiaAlexandru Coca @alexcoca23
54 Followers 146 Following PhD student in the Machine Intelligence Lab at the University of Cambridge / intern @appleRyan Luo @haoyanl99
3 Followers 10 Following #ML MRes student @ImperialCollege. Prev. research intern @MSFTResearchNikolai Rozanov @ai_nikolai
90 Followers 181 Following PhD Candidate in Foundation Models @imperialcollege, ex-tech Founder, all views my ownMatthieu Meeus @matthieu_meeus
64 Followers 184 Following PhD student @ImperialCollege - Privacy & AIMarek Rei @MarekRei
2K Followers 265 Following Researcher in #MachineLearning and #NLProc, working on representation learning and language. Lecturer at @imperialcollege, visiting researcher at @Cambridge_UniImperial NLP @imperial_nlp
66 Followers 336 Following We are the Natural Language Processing community here at Imperial College London. Looking forward to sharing more of our work over the coming months! #NLProcxAI @xai
997K Followers 36 FollowingPascale Fung @pascalefung
2K Followers 45 Following Chair Professor of ECE, Director of the Centre for AI Research (CAiRE), Hong Kong University of Science & Technology. Fellow of AAAI, ACL, IEEE, ISCA.Dan Goodman @neuralreckoning
12K Followers 614 Following Computational neuroscientist @imperialcollege. I like to make stuff: @briansimulator @neuromatch. 🐘 @[email protected]Ashwin Lalendran @ashwinl
2K Followers 2K Following Positive sum Globalist | Engineered bomb detecting drones @AFResearchLab and mapped our world @Apple | AI's biggest impact will be on the physical world |⚡️Yacine Jernite @YJernite
4K Followers 1K Following ML & Society lead @huggingface, NLPer at heart, focusing on data and ML systems governance these days he/him #BlackLivesMatterJoe Stacey @_joestacey_
569 Followers 1K Following PhD student at Imperial and Apple Scholar. I love running, NLP and travelling (in no particular order). Ex teacher and PwC Consultant. #NLProcShaily @shaily99
5K Followers 2K Following PhD @LTIatCMU Prev: @GoogleAI @MSFTResearch. Working on ethics and evaluation in #NLProc. Usually ranting, often about research & DEI. 📚 @readsndrantsEunsol Choi @eunsolc
5K Followers 815 Following on natural language processing / machine learning. assistant professor @UTCompSci. prev @googleai, @uwcse, @Cornell. opinions are of my own.Denny Zhou @denny_zhou
9K Followers 420 Following @GoogleDeepMind founder & lead of Reasoning Team. Build LLMs to reason. Opinions my own.Dianbo Liu @DianboLiu
845 Followers 972 Following Assistant professor at @NUSingapore . Previously @BroadInstitute @Mila @Harvard @BostonChidrens @MIT @MIT_CSAIL,@dundeeuni @EdinburghUniYao Fu @Francis_YAO_
14K Followers 2K Following PhD @EdinburghNLP on LLMs and Machine Reasoning. Ex. @Columbia @PKU1898 @MITIBMLab @allen_ai AGI has yet to come, so keep runningPetar Veličković @PetarV_93
30K Followers 555 Following Staff Research Scientist @GoogleDeepMind | Affiliated Lecturer @Cambridge_Uni | Associate @clarehall_cam | GDL Scholar @ELLISforEurope. Monoids. 🇷🇸🇲🇪🇧🇦illbeback.ai - The AI.. @illbeback_jobs
434 Followers 16 Following 🏆All the best AI jobs around the world. 🚀 Follow for updates on new jobs. Are you hiring? Post a job now ➡️ https://t.co/py4wEwPPZZMiro Dudik @MiroDudik
989 Followers 134 Following Sr Principal Researcher @MSFTResearch. AI Fairness, Reinforcement Learning. #Maxent co-creator. @Fairlearn co-founder. He/him 🏳️🌈. https://t.co/kG9AJetvD9Arianna Bisazza @AriannaBisazza
1K Followers 240 Following AssociateProfessor #NLProc @univgroningen | Multilingualism | Intrepretability | Language Learning in Humans vs NeuralNets | Mum^2Ibrahim Alabdulmohsin.. @ibomohsin
900 Followers 582 Following AI Research Scientist at @GoogleDeepmindSebastian Ruder @seb_ruder
80K Followers 1K Following Multilingual LLMs @cohere • Prev: @GoogleDeepMind • Newsletter: https://t.co/7JGh2qpG98Hady Elsahar @hadyelsahar
2K Followers 2K Following I am a human being, I am your equal ✊🏿✊🏾✊🏻 Research Scientist @MetaAI Identity ∈ ℂ ; Borders ∈ 𝕀 , mytweets ∉ anybody else .Amirhossein Habibian @amir_habibian
911 Followers 290 Following Research Scientist (Director) @Qualcomm AI ResearchTatsunori Hashimoto @tatsu_hashimoto
6K Followers 202 Following Assistant Prof at Stanford CS, member of @stanfordnlp and statsml groups; Formerly at Microsoft / postdoc at Stanford CS / Stats.EdinburghNLP @EdinburghNLP
11K Followers 139 Following The Natural Language Processing Group at the University of EdinburghELLIS @ELLISforEurope
12K Followers 70 Following Shaping the future with AI research excellence! More about ELLIS: https://t.co/n6nVmEGhKkUCL Natural Language .. @ucl_nlp
13K Followers 223 Following NLP/ML research group at @UCLCS, PIs: S. Riedel (@riedelcastro), P. Stenetorp, T. Rocktäschel (@_rockt), E. Grefenstette (@egrefen), P. Minervini (@pminervini)Pasquale Minervini �.. @PMinervini
7K Followers 4K Following Researcher in ML/NLP at the University of Edinburgh (faculty @InfAtEd @EdinburghNLP), @ELLISforEurope, @UCL_NLP, PI for @Clarify2020, https://t.co/WydvfU8ugz he/theylmsys.org @lmsysorg
38K Followers 172 Following Large Model Systems Organization. We created Vicuna and Chatbot Arena! Compare 30+ LLMs (GPT-4/Claude/Llamas) side-by-side at https://t.co/IDFeIDIOtmDavid Krueger @DavidSKrueger
13K Followers 4K Following Cambridge faculty - AI alignment, deep learning, and existential safety. Formerly Mila, FHI, DeepMind, ElementAI, AISI.Siva Reddy @sivareddyg
5K Followers 965 Following Assistant Professor @Mila_Quebec @McGillU @ServiceNowRSRCH; Postdoc @StanfordNLP; PhD @EdinburghNLP; Natural Language Processor #NLProcAlexis Conneau @alex_conneau
24K Followers 113 Following Audio AGI Research Lead @OpenAI - GPT-Next - Past: XLM, Unsupervised ASR, Unsupervised MT, Wav2vec 2.0/XLSR, MUSE, Unsupervised cross-lingual transferAK @_akhaliq
310K Followers 3K Following AI research paper tweets, ML @Gradio (acq. by @HuggingFace 🤗) dm for promo follow on Hugging Face: https://t.co/q2Qoey80Gxelvis @omarsar0
189K Followers 486 Following Building with LLMs @dair_ai • Prev: Meta AI, Galactica LLM, PapersWithCode, Elastic, PhD • Creator of the Prompting Guide (~4M learners)Should we open a London office? It's the third largest cluster of research talent for us, after NYC and SF.
We now have an open-source model that is beating Claude 3 Opus... being served at nearly **300 tokens per second** on @GroqInc. The applications built off of this tech will be nothing short of revolutionary.
We have a new preprint out - your language model is not a reward, it’s a Q function! 1. The likelihood of the preferred answer must go down - it’s a policy divergence 2. MCTS guided decoding on language is equivalent to likelihood search on DPO 3. DPO learns credit assignment
A new paper introducing Enformer Celltyping - the best-in-class epigenetic prediction model of histone marks in previously unseen cell types. Great work by @Al_Murphy_ and @n_skene, happy to have collaborated on it! #MachineLearning biorxiv.org/content/10.110…
Delighted to present our work predicting cell type-specific epigenomic profiles whilst accounting for distal genetic effects. See below for a 🧵 with the details 1/15 biorxiv.org/content/10.110…
I've now reached 100 citations! How exciting! Here's a little thread to summarise my papers so far. Have a little glance, you might see something quite cool :) Most of this work is on NLI robustness / interpretability, with one project on task-oriented dialogue
Nice paper from Imperial's @deniz__gorur showing the effectiveness of LLMs at Relation-based argument mining! arxiv.org/pdf/2402.11243…
Join us on Thursday for what promises to be a really exciting event!
The Workshop on Responsible Data Science and AI Research is this Thursday We have an exciting lineup of speakers and panel sessions, for programme and (free) registration visit responsible-ai.science Organised by @adrian_weller @MalvikaSharan @amaya_gs & me
Do humans need language for #reasoning and #planning? It really depends. We “reasons and plan” a lot of physical and artistic activities without thinking in language (eg dance, swimming, painting, playing an instrument…) but language helps us teach/learn how to do them.
No. If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old. But we still don't have fully autonomous, reliable self-driving, even though we (you) have millions of hours of *labeled* training data.
AI will probably be smarter than any single human next year. By 2029, AI is probably smarter than all humans combined.
Microsoft presents LLMs in the Imaginarium Tool Learning through Simulated Trial and Error Tools are essential for large language models (LLMs) to acquire up-to-date information and take consequential actions in external environments. Existing work on tool-augmented LLMs
Thanks @_akhaliq for sharing our work led by @BoshiWang2 from @osunlp, so let's chat about how LLMs should learn to use tools, a necessary capability of language agents. Tools are essential for LLMs to transcend the confines of their static parametric knowledge and…
Microsoft presents LLMs in the Imaginarium Tool Learning through Simulated Trial and Error Tools are essential for large language models (LLMs) to acquire up-to-date information and take consequential actions in external environments. Existing work on tool-augmented LLMs
Finally able to share the task-oriented dialogue dataset I worked on during my internship with Apple: arxiv.org/pdf/2403.00462… The idea was to create a dataset only with LLMs. It turned out to be really hard, and we had to put so much care in to get such good quality data
Love the data we made, and hope others can find it useful! Many thanks to @johntorr007 @andersjo @alexcoca23 and all our other co-authors who I haven't found on Twitter yet 🙂
Data can be found directly here: raw.githubusercontent.com/apple/ml-lucid…
An example conversation is below. You can see it's quite tricky, the model needs to pick up that the User is saying something to someone else (which LUCID has overheard) instead of giving the actual film name.
The labelling looks like this. It avoids the need to do state tracking each turn, and very convenient for LLM predictions.
How did we get such reliable labelling from an LLM? A few things helped: 1) Our new, concise pythonic labelling system 2) Our 'if in doubt, discard' approach to data gen 3) Breaking down the task into so many, separate LLM calls 4) Validation everywhere!
If you're interested in task-oriented dialogue, do have a look at the data! We have very nice multi-turn conversations, with a range of complex conversational phenomena (all of which are labelled at a turn-level).
@neurosutras Data almost certainly fake. It's from a fake paper by far right bloggers full of racism and sexism. Designed to be retweeted to get followers. (Not @SuryaGanguli of course, but the original authors.)