filename
stringlengths
9
127
text
stringlengths
133
11k
2402.11960v1.pdf
DB-LLM: Accurate Dual-Binarization for Efficient LLMs Hong Chen1*, Chengtao Lv1*, Liang Ding2, Haotong Qin1, Xiabin Zhou4, Yifu Ding1, Xuebo Liu3, Min Zhang3, Jinyang Guo1, Xianglong Liu1†, Dacheng Tao2 1Beihang University2The University of Sydney 3Harbin Institute of Technology, Shenzhen4Jiangsu University {18373205, ...
2210.13382.pdf
Published as a conference paper at ICLR 2023 EMERGENT WORLD REPRESENTATIONS : EXPLORING A SEQUENCE MODEL TRAINED ON A SYNTHETIC TASK Kenneth Li∗ Harvard UniversityAspen K. Hopkins Massachusetts Institute of TechnologyDavid Bau Northeastern University Fernanda Vi ´egas Harvard UniversityHanspeter Pfister Harvard Universi...
1809.04281.pdf
MUSIC TRANSFORMER : GENERATING MUSIC WITH LONG -TERM STRUCTURE Cheng-Zhi Anna Huang∗Ashish Vaswani Jakob Uszkoreit Noam Shazeer Ian Simon Curtis Hawthorne Andrew M. Dai Matthew D. Hoffman Monica Dinculescu Douglas Eck Google Brain ABSTRACT Music relies heavily on repetition to build structure and meaning. Self-referenc...
NeurIPS-2022-training-language-models-to-follow-instructions-with-human-feedback-Paper-Conference.pdf
Training language models to follow instructions with human feedback Long Ouyang∗Jeff Wu∗Xu Jiang∗Diogo Almeida∗Carroll L. Wainwright∗ Pamela Mishkin∗Chong Zhang Sandhini Agarwal Katarina Slama Alex Ray John Schulman Jacob Hilton Fraser Kelton Luke Miller Maddie Simens Amanda Askell†Peter Welinder Paul Christiano∗† Jan ...
2305.12132.pdf
Can Public Large Language Models Help Private Cross-device Federated Learning? Boxin Wang3∗, Yibo Jacky Zhang4, Yuan Cao2, Bo Li3, H. Brendan McMahan1, Sewoong Oh1, Zheng Xu1, Manzil Zaheer2 1Google Research,2Google Deepmind,3UIUC,4Stanford Abstract We study (differentially) private federated learning (FL) of language ...
2201.02867v3.pdf
Deep Generative Modeling for Volume Reconstruction in Cryo-Electron Microscopy Claire Donnat1+, Axel Levy2,3, Fr´ed´eric Poitevin3, Ellen Zhong4, and Nina Miolane5*+ 1University of Chicago, Department of Statistics, Chicago, Illinois, USA 2Stanford University, Department of Electrical Engineering, Stanford, CA, USA 3LC...
2304.02034.pdf
Effective Theory of Transformers at Initialization Emily Dinan,∗Sho Yaida,†and Susan Zhang‡ Meta AI Meta Platforms, Inc.§ We perform an effective-theory analysis of forward–backward signal propagation in wide and deep Transformers, i.e., residual neural networks with multi-head self-attention blocks and multilayer percep...
2307.12950.pdf
RLCD: R EINFORCEMENT LEARNING FROM CONTRAST DISTILLATION FOR LANGUAGE MODEL ALIGNMENT Kevin Yang1,2Dan Klein1Asli Celikyilmaz2Nanyun Peng3Yuandong Tian2 1UC Berkeley,2Meta AI,3UCLA {yangk,klein}@berkeley.edu,{aslic,yuandong}@meta.com,violetpeng@cs.ucla.edu ABSTRACT We propose Reinforcement Learning from Contrast Distil...
2206.14486.pdf
Beyond neural scaling laws: beating power law scaling via data pruning Ben Sorscher∗ ∗1Robert Geirhos∗2Shashank Shekhar3 Surya Ganguli1,3§Ari S. Morcos3§ ∗equal contribution 1Department of Applied Physics, Stanford University 2University of Tübingen 3Meta AI (FAIR) §Joint senior authors Abstract Widely observed neural ...
2305.16381.pdf
DPOK: Reinforcement Learning for Fine-tuning Text-to-Image Diffusion Models Ying Fan˚,1,2, Olivia Watkins3, Yuqing Du3, Hao Liu3, Moonkyung Ryu1, Craig Boutilier1, Pieter Abbeel3,Mohammad Ghavamzadeh1,Kangwook Lee2,Kimin Lee˚,1 ˚Equal technical contribution 1Google Research2University of Wisconsin-Madison3UC Berkeley A...
10.1016.j.cell.2024.01.036.pdf
Article Structure of the plant plastid-encoded RNA polymerase Graphical abstract Highlights dStructure of the chloroplast transcription complex dFifteen nuclear-encoded subunits encase the plastid- encoded polymerase dSubunits PAP1 and PAP2 interact with the DNA and themRNA, respectively dStructure-guided insights into...
99_on_recovering_higher_order_int.pdf
ONRECOVERING HIGHER -ORDER INTERACTIONS FROM PROTEIN LANGUAGE MODELS Darin Tsui & Amirali Aghazadeh School of Electrical and Computer Engineering Georgia Institute of Technology Atlanta, GA 30332, USA {darint,amiralia }@gatech.edu ABSTRACT Protein language models leverage evolutionary information to perform state-of- t...
langegabelriedmiller2011chapter.pdf
Batch Reinforcement Learning Sascha Lange, Thomas Gabel, and Martin Riedmiller Abstract Batch reinforcement learning is a subfield of dynamic programming-based reinforcement learning. Originally defined as the task of learning the best possible policy from a fixed set of a priori-known transition samples, the (batch) algo...
2210.15097.pdf
Contrastive Decoding: Open-ended Text Generation as Optimization Xiang Lisa Li1, Ari Holtzman2, Daniel Fried3, Percy Liang1, Jason Eisner4, Tatsunori Hashimoto1, Luke Zettlemoyer2,5, Mike Lewis5 Stanford University1, University of Washington2, Carnegie Mellon University3, Johns Hopkins University4, FAIR5 xlisali@stanfo...
3639-the-effects-of-reward-misspeci.pdf
THEEFFECTS OF REWARD MISSPECIFICATION : MAPPING AND MITIGATING MISALIGNED MODELS Alexander Pan CaltechKush Bhatia UC BerkeleyJacob Steinhardt UC Berkeley ABSTRACT Reward hacking—where RL agents exploit gaps in misspecified reward functions—has been widely observed, but not yet systematically studied. To un- derstand ho...
2401.12187.pdf
WARM: On the Benefits of Weight Averaged Reward Models Alexandre Ramé, Nino Vieillard, Léonard Hussenot, Robert Dadashi, Geoffrey Cideron, Olivier Bachem, Johan Ferret Google DeepMind Aligning large language models (LLMs) with human preferences through reinforcement learning (RLHF) can lead to reward hacking, where LLM...
2305.16183.pdf
Passive learning of active causal strategies in agents and language models Andrew K. Lampinen Google DeepMind London, UK lampinen@deepmind.comStephanie C. Y. Chan Google DeepMind London, UK scychan@deepmind.comIshita Dasgupta Google DeepMind London, UK idg@deepmind.com Andrew J. Nam Stanford University Stanford, CA ajh...
2001.08361.pdf
Scaling Laws for Neural Language Models Jared Kaplan∗ Johns Hopkins University, OpenAI jaredk@jhu.eduSam McCandlish∗ OpenAI sam@openai.com Tom Henighan OpenAI henighan@openai.comTom B. Brown OpenAI tom@openai.comBenjamin Chess OpenAI bchess@openai.comRewon Child OpenAI rewon@openai.com Scott Gray OpenAI scott@openai.co...
10.1038.s41467-023-38539-w.pdf
Article https://doi.org/10.1038/s41467-023-38539-w A method for restoring signals and revealing individual macromolecule states incryo-ET, REST Haonan Zhang1,2,3,Y a nL i1,3,Y a n a nL i u1,2, Dongyu Li1,2,L i nW a n g1, Kai Song1, Keyan Bao1& Ping Zhu1,2 Cryo-electron tomography (cryo-ET) is widely used to explore the...
1801.10198.pdf
Published as a conference paper at ICLR 2018 GENERATING WIKIPEDIA BY SUMMARIZING LONG SEQUENCES Peter J. Liu∗, Mohammad Saleh∗, Etienne Pot†, Ben Goodrich, Ryan Sepassi, Łukasz Kaiser, Noam Shazeer Google Brain Mountain View, CA {peterjliu,msaleh,epot,bgoodrich,rsepassi,lukaszkaiser,noam }@google.com ABSTRACT We show t...
10.1126.science.abo7201.pdf
RESEARCH ARTICLE SUMMARY◥ CORONAVIRUS Open science discovery of potent noncovalent SARS-CoV-2 main protease inhibitors Melissa L. Boby †, Daren Fearon †, Matteo Ferla †, Mihajlo Filep †, Lizbé Koekemoer †, Matthew C. Robinson †, The COVID Moonshot Consortium, John D. Chodera *, Alpha A. Lee *, Nir London *, Annette von...
2306.16410.pdf
Towards Language Models That Can See: Computer Vision Through the LENS of Natural Language William Berrios†Gautam Mittal†§Tristan Thrush†§ Douwe Kiela†§Amanpreet Singh† †Contextual AI;§Stanford University Abstract We propose LENS , a modular approach for tackling computer vision problems by leveraging the power of la...
2005.00341.pdf
Jukebox: A Generative Model for Music Prafulla Dhariwal* 1Heewoo Jun* 1Christine Payne* 1Jong Wook Kim1Alec Radford1Ilya Sutskever1 Abstract We introduce Jukebox, a model that generates music with singing in the raw audio domain. We tackle the long context of raw audio using a multi- scale VQ-V AE to compress it to dis...
1905.01969v4.pdf
Published as a conference paper at ICLR 2020 Poly-encoders :architectures and pre -training strategies for fast and accurate multi -sentence scoring Samuel Humeau∗, Kurt Shuster∗, Marie-Anne Lachaux, Jason Weston Facebook AI Research {samuelhumeau,kshuster,malachaux,jase }@fb.com Abstract The use of deep pre-trained tr...
2401.18079.pdf
KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization Coleman Hooper chooper@berkeley.edu UC BerkeleySehoon Kim sehoonkim@berkeley.edu UC BerkeleyHiva Mohammadzadeh hiva@berkeley.edu UC Berkeley Michael W. Mahoney mmahoney@stat.berkeley.edu ICSI, LBNL, UC BerkeleyYakun Sophia Shao ysshao@b...
2305.15717.pdf
The False Promise of Imitating Proprietary LLMs Arnav Gudibande∗ UC Berkeley arnavg@berkeley.eduEric Wallace∗ UC Berkeley ericwallace@berkeley.eduCharlie Snell∗ UC Berkeley csnell22@berkeley.edu Xinyang Geng UC Berkeley young.geng@berkeley.eduHao Liu UC Berkeley hao.liu@berkeley.eduPieter Abbeel UC Berkeley pabbeel@ber...
2306.02707.pdf
Orca: Progressive Learning from Complex Explanation Traces of GPT-4 Subhabrata Mukherjee∗†, Arindam Mitra∗ Ganesh Jawahar, Sahaj Agarwal, Hamid Palangi, Ahmed Awadallah Microsoft Research Abstract Recent research has focused on enhancing the capability of smaller models through imitation learning, drawing on the output...
109_how_well_do_generative_protein.pdf
HOW WELL DO GENERATIVE PROTEIN MODELS GENERATE ? Han Spinner Department of Systems Biology Harvard Medical SchoolAaron W. Kollasch Department of Systems Biology Harvard Medical SchoolDebora S. Marks Department of Systems Biology Harvard Medical School ABSTRACT Protein design relies critically on the generation of plaus...
Pursuing-structural-biology-in-China-cell.pdf
Leading Edge Conversations Pursuing structural biology in China In November 2023, structural biologists from different countries and different disciplines gathered at the Cell Symposium: Structural biology from the nanoscale to cellular mesoscale to discuss recent breakthroughs,including structures of proteins and macr...
HyvO00-icatut.pdf
Indep enden t Comp onen t Analysis/: A T utorialAap o Hyv /ärinen and Erkki OjaHelsinki Univ ersit y of T ec hnologyLab oratory of Computer and Information ScienceP /.O/. Bo x /5/4/0/0/, FIN/-/0/2/0/1/5 Esp o o/, Finlandaapo/.hyvarinen/@hut/.fi/, erkki/.oja/@hut/.fihttp/:////www/.cis/.hut/.fi//pro ject s//ic a//A v ers...
2211.06738.pdf
arXiv:2211.06738v1 [cs.AI] 12 Nov 2022Formalizing the presumption of independence Paul Christiano, Eric Neyman, Mark Xu Alignment Research Center Abstract Mathematical proof aims to deliver confident conclusions, but a ver y similar process of deduction can be used to make uncertain estimates that are open t o revisio...
1805.00899.pdf
AI safety via debate Geoffrey Irving∗Paul Christiano OpenAIDario Amodei Abstract To make AI systems broadly useful for challenging real-world tasks, we need them to learn complexhumangoalsandpreferences. Oneapproachtospecifyingcomplexgoalsaskshumansto judge during training which agent behaviors are safe and useful, but ...
2401.10020.pdf
Self-Rewarding Language Models Weizhe Yuan1,2Richard Yuanzhe Pang1,2Kyunghyun Cho2 Xian Li1Sainbayar Sukhbaatar1Jing Xu1Jason Weston1,2 1Meta2NYU Abstract We posit that to achieve superhuman agents, future models require super- human feedback in order to provide an adequate training signal. Current approaches commonly ...
2401.12192.pdf
Text Embedding Inversion Attacks on Multilingual Language Models Yiyi Chen Heather Lent Johannes Bjerva Department of Computer Science, Aalborg University, Denmark {yiyic, hcle, jbjerva}@cs.aau.dk Abstract Representing textual information as real- numbered embeddings has become the norm in NLP. Moreover, with the rise ...
2211.07793.pdf
EXTREME GENERATIVE IMAGE COMPRESSION BY LEARNING TEXT EMBEDDING FROM DIFFUSION MODELS A P REPRINT Zhihong Pan, Xin Zhou, Hao Tian Baidu Research (USA) ABSTRACT Transferring large amount of high resolution images over limited bandwidth is an important but very challenging task. Compressing images using extremely low bit...
gu-dissertation-augmented.pdf
MODELING SEQUENCES WITH STRUCTURED STATE SPACES A DISSERTATION SUBMITTED TO THE DEPARTMENT OF DEPARTMENT OF COMPUTER SCIENCE AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY Albert Gu June 2023
2108.05540.pdf
Unsupervised Corpus Aware Language Model Pre-training for Dense Passage Retrieval Luyu Gao and Jamie Callan Language Technologies Institute Carnegie Mellon University {luyug, callan}@cs.cmu.edu Abstract Recent research demonstrates the effective- ness of using fine-tuned language mod- els (LM) for dense retrieval. Howev...
1501.05014.pdf
Experimental Simulation of Closed Timelike Curves Martin Ringbauer1,2∗, Matthew A. Broome1,2, Casey R. Myers1, Andrew G. White1,2and Timothy C. Ralph2 1Centre for Engineered Quantum Systems,2Centre for Quantum Computer and Communication Technology, School of Mathematics and Physics, University of Queensland, Brisbane, ...
2310.18168.pdf
PERSONAS AS A WAY TO MODEL TRUTHFULNESS IN LANGUAGE MODELS Nitish Joshi1∗Javier Rando2∗Abulhair Saparov1Najoung Kim3He He1 1New York University2ETH Zurich3Boston University {nitish}@nyu.edu {jrando}@ethz.ch ABSTRACT Large Language Models (LLMs) are trained on vast amounts of text from the internet, which contains both ...
1712.03346.pdf
Variational auto-encoding of protein sequences Sam Sinai∗ Harvard University samsinai@g.harvard.eduEric Kelsic†‡ Harvard Medical School eric kelsic@hms.harvard.edu George M. Church§†‡ Harvard Medical School church labadmin@hms.harvard.eduMartin A. Nowak∗‡¶ Harvard University martin nowak@harvard.edu Abstract Proteins a...
2309.16797.pdf
PROMPTBREEDER : SELF-REFERENTIAL SELF-IMPROVEMENT VIAPROMPT EVOLUTION Chrisantha Fernando, Dylan Banarse, Henryk Michalewski, Simon Osindero, Tim Rockt ¨aschel Google DeepMind {chrisantha,dylski,henrykm,osindero,rocktaschel }@google.com ABSTRACT Popular prompt strategies like Chain-of-Thought Prompting can dramatically...
2404.12253v1.pdf
Toward Self-Improvement of LLMs via Imagination, Searching, and Criticizing Ye Tian∗, Baolin Peng∗, Linfeng Song∗, Lifeng Jin, Dian Yu, Haitao Mi†, Dong Yu Tencent AI Lab, Bellevue, WA {yaptian,baolinpeng,lfsong,lifengjin,yudian,haitaomi}@global.tencent.com Abstract Despite the impressive capabilities of Large Language...
2005.10242.pdf
Understanding Contrastive Representation Learning through Alignment and Uniformity on the Hypersphere Tongzhou Wang1Phillip Isola1 Abstract Contrastive representation learning has been out- standingly successful in practice. In this work, we identify two key properties related to the con- trastive loss: (1) alignment (...
Improving-Memory-Search-through-Model-Based-Cue-Selection.pdf
IMPROVING MEMORY SEARCH 1 . Improving Memory Search through Model-Based Cue Selection Charlotte A. Cornell1, Kenneth A. Norman2, Thomas L. Griffiths2,3, and Qiong Zhang1,4 1Psychology Department, Rutgers University–New Brunswick 2Psychology Department, Princeton University 3Computer Science Department, Princeton Univer...
tr00-004.pdf
/CC /D6/CP/CX/D2/CX/D2/CV /C8/D6/D3 /CS/D9/CR/D8/D7 /D3/CU /BX/DC/D4 /CT/D6/D8/D7 /CQ /DD /C5/CX/D2/CX/D1/CX/DE/CX/D2/CV /BV/D3/D2 /D8/D6/CP/D7/D8/CX/DA /CT/BW/CX/DA /CT/D6/CV/CT/D2/CR/CT/BZ/BV/C6/CD /CC/CA /BE/BC/BC/BC/B9/BC/BC/BG/BZ/CT/D3/AB/D6/CT/DD /BX/BA /C0/CX/D2 /D8/D3/D2/BZ/CP/D8/D7/CQ /DD /BV/D3/D1/D4/D9/D8/CP...
2212.04356.pdf
Robust Speech Recognition via Large-Scale Weak Supervision Alec Radford* 1Jong Wook Kim* 1Tao Xu1Greg Brockman1Christine McLeavey1Ilya Sutskever1 Abstract We study the capabilities of speech processing systems trained simply to predict large amounts of transcripts of audio on the internet. When scaled to 680,000 hours ...
Rombach-High-Resolution-Image-Synthesis-With-Latent-Diffusion-Models-CVPR-2022-paper.pdf
High-Resolution Image Synthesis with Latent Diffusion Models Robin Rombach1∗Andreas Blattmann1∗Dominik Lorenz1Patrick Esser Bj¨orn Ommer1 1Ludwig Maximilian University of Munich & IWR, Heidelberg University, Germany Runway ML https://github.com/CompVis/latent-diffusion Abstract By decomposing the image formation proc...
2402.09668.pdf
How to Train Data-Efficient LLMs Noveen Sachdeva1 2Benjamin Coleman1Wang-Cheng Kang1Jianmo Ni1Lichan Hong1Ed H. Chi1 James Caverlee1 3Julian McAuley2Derek Zhiyuan Cheng1 Abstract The training of large language models (LLMs) is expensive. In this paper, we study data-efficient approaches for pre-training LLMs, i.e., tec...
mapreduce.pdf
MapReduce: Simplied Data Processing onLargeClusters JeffreyDean andSanjay Ghema wat jeff@google.com, sanjay@google.com Google,Inc. Abstract MapReduce isaprogramming model andanassoci- ated implementation forprocessing andgenerating large data sets. Users specify amap function thatprocesses a key/valuepairtogenerate as...
2311.00208.pdf
Transformers as Recognizers of Formal Languages: A Survey on Expressivity Lena Strobl Umeå University lena.strobl@umu.seWilliam Merrill New York University willm@nyu.eduGail Weiss EPFL gail.weiss@epfl.ch David Chiang University of Notre Dame dchiang@nd.eduDana Angluin Yale University dana.angluin@yale.edu Abstract As t...
2402.04833.pdf
Long Is More for Alignment: A Simple but Tough-to-Beat Baseline for Instruction Fine-Tuning Hao Zhao1Maksym Andriushchenko1Francesco Croce1Nicolas Flammarion1 Abstract There is a consensus that instruction fine-tuning of LLMs requires high-quality data, but what are they? LIMA (NeurIPS 2023) and AlpaGa- sus (ICLR 2024)...
1801.05134.pdf
Understanding the Disharmony between Dropout and Batch Normalization by Variance Shift Xiang Li1Shuo Chen1Xiaolin Hu2Jian Yang1 Abstract This paper first answers the question “why do the two most powerful techniques Dropout and Batch Normalization (BN) often lead to a worse performance when they are combined together?” ...
2305.13301.pdf
TRAINING DIFFUSION MODELS WITH REINFORCEMENT LEARNING Kevin Black∗1Michael Janner∗1Yilun Du2Ilya Kostrikov1Sergey Levine1 1University of California, Berkeley2Massachusetts Institute of Technology {kvablack, janner, kostrikov, sergey.levine}@berkeley.edu yilundu@mit.edu ABSTRACT Diffusion models are a class of flexible ...
2306.04488.pdf
Rewarded soups: towards Pareto-optimal alignment by interpolating weights fine-tuned on diverse rewards Alexandre Rame1∗, Guillaume Couairon1,2†, Mustafa Shukor1†, Corentin Dancette1†,Jean-Baptiste Gaya1,2†,Laure Soulier1,Matthieu Cord1,3 1Sorbonne Université, CNRS, ISIR, Paris, France2Meta AI3Valeo.ai Abstract Foundat...
2210.03057.pdf
LANGUAGE MODELS ARE MULTILINGUAL CHAIN -OF-THOUGHT REASONERS Freda Shi1,2,∗Mirac Suzgun1,3,∗Markus Freitag1Xuezhi Wang1 Suraj Srivats4Soroush Vosoughi4Hyung Won Chung1Yi Tay1 Sebastian Ruder1Denny Zhou1Dipanjan Das1Jason Wei1 1Google Research2Toyota Technological Institute at Chicago 3Stanford University4Dartmouth Coll...
2306.17806.pdf
Stay on topic with Classifier-Free Guidance Guillaume V . Sanchez* Hexaglobe EleutherAI gsanchez@hexaglobe.comHonglu Fan* University of Geneva EleutherAI honglu.fan@unige.chAlexander Spangher* Information Sciences Institute University of Southern California spangher@usc.edu Elad Levi Sightful eladlevico@gmail.comPawan ...
2310.10638v5.pdf
Published as a conference paper at ICLR 2024 IN-CONTEXT PRETRAINING : LANGUAGE MODELING BEYOND DOCUMENT BOUNDARIES Weijia Shi1,2Sewon Min1,2Maria Lomeli1Chunting Zhou1 Margaret Li1,2Gergely Szilvasy1Rich James1Xi Victoria Lin1 Noah A. Smith2,3Luke Zettlemoyer1,2Scott Yih1Mike Lewis1 1Meta AI2University of Washington3Al...
2305.15348.pdf
READ: Recurrent Adaptation of Large Transformers Sid Wang John Nguyen Ke Li Carole-Jean Wu Meta AI {yuwang2020,ngjhn,kli26,carolejeanwu}@meta.com Abstract Fine-tuning large-scale Transformers has led to the explosion of many AI applica- tions across Natural Language Processing and Computer Vision tasks. However, fine-t...
2309.10668.pdf
Language Modeling Is Compression Grégoire Delétang*1, Anian Ruoss*1, Paul-Ambroise Duquenne2, Elliot Catt1, Tim Genewein1, Christopher Mattern1, Jordi Grau-Moya1, Li Kevin Wenliang1, Matthew Aitchison1, Laurent Orseau1, Marcus Hutter1and Joel Veness1 *Equal contributions,1Google DeepMind,2Meta AI & Inria It has long be...
2404.16710v1.pdf
LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding Mostafa Elhoushi1,†,∗,Akshat Shrivastava1,†,∗,Diana Liskovich2,†,Bram Wasti2,Basil Hosmer1, Liangzhen Lai3,Anas Mahmoud4,Bilge Acun1,Saurabh Agrawal6,Ahmed Roman7,Ahmed A Aly3,Beidi Chen1,5,Carole Jean-Wu1 1FAIR at Meta,2GenAI at Meta,3Reality Labs ...
2212.14024.pdf
DEMONSTRATE –SEARCH –PREDICT : Composing retrieval and language models for knowledge-intensive NLP Omar Khattab1Keshav Santhanam1Xiang Lisa Li1David Hall1 Percy Liang1Christopher Potts1Matei Zaharia1 Abstract Retrieval-augmented in-context learning has emerged as a powerful approach for addressing knowledge-intensive t...
L08_expressivity.pdf
Expressive Variational Autoencoders John Thickstun The Gaussian VAE parameterizes the prior r(z), conditional likelihood p(x|z), and posterior approximation q(x|z) with with Gaussian distributions. The in-expressivity of these Gaussian models can make it difficult to capture the distribution p(x); complaints about the “b...
2311.11944v1.pdf
FINANCE BENCH : A New Benchmark for Financial Question Answering Pranab Islam1∗Anand Kannappan1Douwe Kiela2,3 Rebecca Qian1Nino Scherrer1Bertie Vidgen1 1Patronus AI2Contextual AI3Stanford University Abstract FINANCE BENCH is a first-of-its-kind test suite for evaluating the performance of LLMs on open book financial qu...
2403.09636.pdf
Dynamic Memory Compression: Retrofitting LLMs for Accelerated Inference Piotr Nawrot*Q VAdrian Ła ´ncucki*Q KMarcin ChochowskiQDavid TarjanQEdoardo M. PontiV QNVIDIAKUniversity of WrocławVUniversity of Edinburgh Abstract Transformers have emerged as the backbone of large language models (LLMs). However, genera- tion re...
1610.03518v1.pdf
Transfer from Simulation to Real World through Learning Deep Inverse Dynamics Model Paul Christiano, Zain Shah, Igor Mordatch, Jonas Schneider, Trevor Blackwell, Joshua Tobin, Pieter Abbeel, and Wojciech Zaremba OpenAI, San Francisco, CA, USA Abstract — Developing control policies in simulation is often more practical ...
2302.03764.pdf
Sketchy: Memory-efficient Adaptive Regularization with Frequent Directions Vladimir Feinberg1Xinyi Chen1 2Y. Jennifer Sun2Rohan Anil1Elad Hazan1 2 Abstract Adaptive regularization methods that exploit more than the diagonal entries exhibit state of the art performance for many tasks, but can be pro- hibitive in terms of...
1608.04471.pdf
Stein Variational Gradient Descent: A General Purpose Bayesian Inference Algorithm Qiang Liu Dilin Wang Department of Computer Science Dartmouth College Hanover, NH 03755 {qiang.liu, dilin.wang.gr}@dartmouth.edu Abstract We propose a general purpose variational inference algorithm that forms a natural counterpart of gr...
1812.11118.pdf
Reconciling modern machine learning practice and the bias-variance trade-off Mikhail Belkina, Daniel Hsub, Siyuan Maa, and Soumik Mandala aThe Ohio State University, Columbus, OH bColumbia University, New York, NY September 12, 2019 Abstract Breakthroughs in machine learning are rapidly changing science and society, yet...
2002.05616.pdf
Learning the Stein Discrepancy for Training and Evaluating Energy-Based Models without Sampling Will Grathwohl1Kuan-Chieh Wang1J¨orn-Henrik Jacobsen1David Duvenaud1Richard Zemel1 Abstract We present a new method for evaluating and train- ing unnormalized density models. Our approach only requires access to the gradient...
2304.14802.pdf
ResiDual: Transformer with Dual Residual Connections Shufang Xie‡†, Huishuai Zhang†, Junliang Guo†, Xu Tan†∗, Jiang Bian† Hany Hassan Awadalla†,Arul Menezes†,Tao Qin†,Rui Yan‡∗ †Microsoft Research†Microsoft Azure Translation ‡Gaoling School of Artificial Intelligence, Renmin University of China {shufangxie,ruiyan}@ruc.e...
2403.07816.pdf
Branch-Train-MiX: Mixing Expert LLMs into a Mixture-of-Experts LLM Sainbayar Sukhbaatar ,Olga Golovneva ,Vasu Sharma ,Hu Xu,Xi Victoria Lin ,Baptiste Rozière ,Jacob Kahn,Daniel Li,Wen-tau Yih ,Jason Weston ,Xian Li FAIR at Meta We investigate efficient methods for training Large Language Models (LLMs) to possess capabi...
2209.15634.pdf
A General Framework for Sample-Efficient Function Approximation in Reinforcement Learning Zixiang Chen‡∗Chris Junchi Li⋄∗Angela Yuan‡∗Quanquan Gu‡Michael I. Jordan⋄,† Department of Computer Sciences, University of California, Los Angeles‡ Department of Electrical Engineering and Computer Sciences, University of Californi...
2205.13147.pdf
Matryoshka Representation Learning Aditya Kusupati∗†⋄, Gantavya Bhatt∗†, Aniket Rege∗†, Matthew Wallingford†, Aditya Sinha⋄, Vivek Ramanujan†, William Howard-Snyder†, Kaifeng Chen⋄, Sham Kakade‡, Prateek Jain⋄and Ali Farhadi† †University of Washington,⋄Google Research,‡Harvard University {kusupati,ali}@cs.washington.ed...
2307.15043.pdf
Universal and Transferable Adversarial Attacks on Aligned Language Models Andy Zou1,2, Zifan Wang2, Nicholas Carlini3, Milad Nasr3, J. Zico Kolter1,4, Matt Fredrikson1 1Carnegie Mellon University,2Center for AI Safety, 3Google DeepMind,4Bosch Center for AI Abstract Because “out-of-the-box” large language models are cap...
2207.10551.pdf
Scaling Laws vs Model Architectures : How does Inductive Bias Influence Scaling? Yi Tay∗Mostafa Dehghani∗Samira Abnar Hyung Won Chung William Fedus Jinfeng Rao Sharan Narang Vinh Q. Tran Dani Yogatama†Donald Metzler Google Research & DeepMind† {yitay,dehghani}@google.com Abstract There have been a lot of interest in the...
2212.14024v2.pdf
DEMONSTRATE –SEARCH –PREDICT : Composing retrieval and language models for knowledge-intensive NLP Omar Khattab1Keshav Santhanam1Xiang Lisa Li1David Hall1 Percy Liang1Christopher Potts1Matei Zaharia1 Abstract Retrieval-augmented in-context learning has emerged as a powerful approach for addressing knowledge-intensive t...
2302.12441.pdf
MUX-PLMs: Data Multiplexing for High-throughput Language Models Vishvak Murahari1Ameet Deshpande1Carlos E. Jimenez1 Izhak Shafran2Mingqiu Wang2Yuan Cao2Karthik Narasimhan1 1Princeton University2Google Brain murahari@cs.princeton.edu Abstract The widespread adoption of large language models such as ChatGPT and Bard has ...
10.1038.s41467-021-25756-4.pdf
ARTICLE Efficient generative modeling of protein sequences using simple autoregressive models Jeanne Trinquier1,2, Guido Uguzzoni3,4, Andrea Pagnani3,4,5, Francesco Zamponi2& Martin Weigt1✉ Generative models emerge as promising candidates for novel sequence-data driven approaches to protein design, and for the extractio...
2306.03078.pdf
SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression Tim Dettmers∗ † University of WashingtonRuslan Svirschevski∗ HSE University & YandexVage Egiazarian∗ HSE University & Yandex Denis Kuznedelev∗ Yandex & SkoltechElias Frantar IST AustriaSaleh Ashkboos ETH ZurichAlexander Borzunov HSE Univer...
1706.03741.pdf
Deep Reinforcement Learning from Human Preferences Paul F Christiano OpenAI paul@openai.comJan Leike DeepMind leike@google.comTom B Brown nottombrown@gmail.com Miljan Martic DeepMind miljanm@google.comShane Legg DeepMind legg@google.comDario Amodei OpenAI damodei@openai.com Abstract For sophisticated reinforcement lear...
karakida19a.pdf
Universal Statistics of Fisher Information in Deep Neural Networks: Mean Field Approach Ryo Karakida Shotaro Akaho Shun-ichi Amari AIST, Japan AIST, Japan RIKEN CBS, Japan Abstract The Fisher information matrix (FIM) is a fundamental quantity to represent the char- acteristics of a stochastic model, including deep neur...
2310.06816.pdf
Text Embeddings Reveal (Almost) As Much As Text John X. Morris, Volodymyr Kuleshov, Vitaly Shmatikov, Alexander M. Rush Department of Computer Science Cornell University Abstract How much private information do text em- beddings reveal about the original text? We investigate the problem of embedding inver- sion, recons...
1908.10084v1.pdf
Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks Nils Reimers and Iryna Gurevych Ubiquitous Knowledge Processing Lab (UKP-TUDA) Department of Computer Science, Technische Universit ¨at Darmstadt www.ukp.tu-darmstadt.de Abstract BERT (Devlin et al., 2018) and RoBERTa (Liu et al., 2019) has set a new state-...
2402.00854.pdf
SymbolicAI: A framework for logic-based approaches combining generative models and solvers Marius–Constantin Dinu∗ †Claudiu Leoveanu–Condrei‡Markus Holzleitner† Werner Zellinger§Sepp Hochreiter† Abstract We introduce SymbolicAI , a versatile and modular framework employing a logic-based approach to concept learning and...
1907.10786.pdf
Interpreting the Latent Space of GANs for Semantic Face Editing Yujun Shen1, Jinjin Gu2, Xiaoou Tang1, Bolei Zhou1 1The Chinese University of Hong Kong2The Chinese University of Hong Kong, Shenzhen {sy116, xtang, bzhou }@ie.cuhk.edu.hk, jinjingu@link.cuhk.edu.cn Original Pose Age Gender Eyeglasses Figure 1: Manipulatin...
2107.13163.pdf
arXiv:2107.13163v3 [cs.LG] 30 Mar 2023Statistically Meaningful Approximation: a Case Study on Approximating Turing Machines with Transform ers Colin Wei Yining Chen Tengyu Ma Department of Computer Science Stanford University {colinwei,cynnjjs,tengyuma}@cs.stanford.edu March 31, 2023 Abstract A common lens to theoret...
1906.08237.pdf
XLNet: Generalized Autoregressive Pretraining for Language Understanding Zhilin Yang∗1, Zihang Dai∗12, Yiming Yang1, Jaime Carbonell1, Ruslan Salakhutdinov1, Quoc V . Le2 1Carnegie Mellon University,2Google AI Brain Team {zhiliny,dzihang,yiming,jgc,rsalakhu}@cs.cmu.edu, qvl@google.com Abstract With the capability of mo...
2206.05895.pdf
Latent Diffusion Energy-Based Model for Interpretable Text Modeling Peiyu Yu1 2Sirui Xie1Xiaojian Ma1 2Baoxiong Jia1 2Bo Pang3 Ruiqi Gao4Yixin Zhu5 6Song-Chun Zhu1 2 5 6 7 8Ying Nian Wu7 Abstract Latent space Energy-Based Models ( EBM s), also known as energy-based priors, have drawn grow- ing interests in generative m...
2209.13325.pdf
Outlier Suppression: Pushing the Limit of Low-bit Transformer Language Models Xiuying Wei1, 2, Yunchen Zhang2, 4, Xiangguo Zhang2, Ruihao Gong1, 2, Shanghang Zhang3, Qi Zhang2, Fengwei Yu2, Xianglong Liu1∗ 1State Key Lab of Software Development Environment, Beihang University 2SenseTime Research,3Peking University 4Uni...
2308.05660v1.pdf
Thermodynamic Linear Algebra Maxwell Aifer, Kaelan Donatella, Max Hunter Gordon, Thomas Ahle, Daniel Simpson, Gavin Crooks, Patrick J. Coles Normal Computing Corporation, New York, New York, USA Linear algebraic primitives are at the core of many modern algorithms in engineering, science, and machine learning. Hence, a...
2312.17227.pdf
Gradient-based Planning with World Models Jyothir S V1∗Siddhartha Jalagam1∗Yann LeCun1, 2Vlad Sobal1, 2 1New York University2Meta AI {jyothir, scj9994, us441}@nyu.edu yann@cs.nyu.edu Abstract The enduring challenge in the field of artificial intelligence has been the control of systems to achieve desired behaviours. Wh...
10.1038.s41564-023-01584-8.pdf
Nature Microbiology nature microbiologyhttps://doi.org/10.1038/s41564-023-01584-8 Analysis Large language models improve annotation of prokaryotic viral proteins Zachary N. Flamholz   1, Steven J. Biller   2 & Libusha Kelly   1,3 Viral genomes are poorly annotated in metagenomic samples, representing an obstacle ...
2202.03286.pdf
Red Teaming Language Models with Language Models WARNING: This paper contains model outputs which are offensive in nature. Ethan Perez1 2Saffron Huang1Francis Song1Trevor Cai1Roman Ring1 John Aslanides1Amelia Glaese1Nat McAleese1Geoffrey Irving1 1DeepMind,2New York University perez@nyu.edu Abstract Language Models (LMs...
2401.14196.pdf
DeepSeek-Coder: When the Large Language Model Meets Programming - The Rise of Code Intelligence Daya Guo*1, Qihao Zhu∗1,2, Dejian Yang1, Zhenda Xie1, Kai Dong1, Wentao Zhang1 Guanting Chen1, Xiao Bi1, Y. Wu1, Y.K. Li1, Fuli Luo1, Yingfei Xiong2, Wenfeng Liang1 1DeepSeek-AI 2Key Lab of HCST (PKU), MOE; SCS, Peking Unive...
dubey2022pursuit.pdf
RESEA RCH ARTICL E Thepursuit ofhappiness: Areinforcement learning perspective onhabituation and comparisons Rachit Dubey ID 1*,Thomas L.Griffiths2,Peter Dayan ID 3,4 1Department ofComputer Science, Princeton University ,Princeton, New Jersey, United States ofAmerica, 2Department ofPsychology, Prince tonUniversity, Pri...
2312.11671v2.pdf
Evaluating Language-Model Agents on Realistic Autonomous Tasks Megan Kinniment Lucas Jun Koba Sato Haoxing Du Brian Goodrich Max Hasin Lawrence Chan Luke Harold Miles Tao R. Lin Hjalmar Wijk Joel Burget Aaron Ho Elizabeth Barnes∗Paul Christiano† METR (Formerly ARC Evals) Abstract In this report, we explore the ability ...
2310.11589.pdf
ELICITING HUMAN PREFERENCES WITH LANGUAGE MODELS Belinda Z. Li∗ MIT CSAIL bzl@mit.eduAlex Tamkin∗ Anthropic† atamkin@cs.stanford.eduNoah Goodman Stanford ndg@stanford.eduJacob Andreas MIT CSAIL jda@mit.edu ABSTRACT Language models (LMs) can be directed to perform target tasks by using labeled examples or natural langua...
2403.20222v1.pdf
Shallow Cross-Encoders for Low-Latency Retrieval Aleksandr V. Petrov, Sean MacAvaney, and Craig Macdonald University of Glasgow, Glasgow, UK a.petrov.1@research.gla.ac.uk {sean.macavaney;craig.macdonald }@glasgow.ac.uk Abstract. Transformer-based Cross-Encoders achieve state-of-the-art effectivness in text retrieval. H...
2309.10400v3.pdf
Published as a conference paper at ICLR 2024 POSE: E FFICIENT CONTEXT WINDOW EXTENSION OF LLM S VIA POSITIONAL SKIP-WISE TRAINING Dawei Zhu∗♡♠Nan Yang♢Liang Wang♢Yifan Song♡♠Wenhao Wu♡♠ Furu Wei♢Sujian Li♡♠ ♡School of Computer Science, Peking University ♠National Key Laboratory for Multimedia Information Processing, Pe...
2202.04728.pdf
Predicting Human Similarity Judgments Using Large Language Models Raja Marjieh1,*, Ilia Sucholutsky2,*, Theodore R. Sumers2, Nori Jacoby3, Thomas L. Griffiths1,2 1Department of Psychology, Princeton University 2Department of Computer Science, Princeton University 3Computational Auditory Perception Group, Max Planck Inst...