text stringlengths 64 6.93k |
|---|
,Unnamed: 0.1,TweetID,AuthorID,AuthorName,Tweets,arxiv_link,Abstract,Title,Thread_length,Tweets_coarse,year,month,tweet_length |
0,31,1463733877021700096,1237589672609476611,Zhimei Ren,"Excited to introduce our new paper on sensitivity analysis of individual treatment effects: <LINK>. Predictive inference of counterfactuals and ITEs, robust to unmeasured confounding. Joint work with @YingJin531 and Emmanuel Candès. (1/n) @YingJin531 2/ Previous ... |
Inference Approach",5,"['Excited to introduce our new paper on sensitivity analysis of individual treatment effects: <LINK>. \nPredictive inference of counterfactuals and ITEs, robust to unmeasured confounding. Joint work with @YingJin531 and Emmanuel Candès. (1/n)', '@YingJin531 2/ \nPrevious work (@lihua_lei_stat... |
1,172,1230779292386254849,174298756,Adel Bibi,"I'm very excited about this new work in collaboration with @MotassimF, Hasan Hammoud, Mohamed Gaafar and Bernard Ghanem. We study the decision boundaries through the lens of tropical geometry in a single hidden layer neural network. <LINK> <LINK> @MotassimF We revisit the ... |
Perspective",2,"[""I'm very excited about this new work in collaboration with @MotassimF, Hasan Hammoud, Mohamed Gaafar and Bernard Ghanem.\n\nWe study the decision boundaries through the lens of tropical geometry in a single hidden layer neural network.\n\n<LINK> <LINK>"", '@MotassimF We revisit the lottery ticket h... |
2,101,1468936846927450117,871354083151667200,Tomek Korbak,"How to fine-tune conditional language models with minimal catastrophic forgetting? A thread about our new #NeurIPS21 #CtrlGen workshop paper with @hadyelsahar, @germank and @MarcDymetman. <LINK> 1/7 <LINK> @hadyelsahar @germank @MarcDymetman Imagine you have a ... |
Gradients",8,"['How to fine-tune conditional language models with minimal catastrophic forgetting? A thread about our new #NeurIPS21 #CtrlGen workshop paper with @hadyelsahar, @germank and @MarcDymetman. <LINK> 1/7 <LINK>', ""@hadyelsahar @germank @MarcDymetman Imagine you have a @GitHubCopilot-like Python LM that yo... |
3,58,1417870039378583554,1360892702,Ekin Dogus Cubuk,"New paper on ML & physics at ICML! Learn2Hop: Learned Optimization on Rough Landscapes With Applications to Atomic Structural Optimization We adapt learned optimizers for atomic structural optimization, and compare to baselines from physics. abs: <LINK> <LINK> Fi... |
4,60,1471052426413322240,1273473804438667265,Haritz Puerto,"New preprint!! How to combine pretrained QA agents to create a multi-agent model that can solve any question? Check our paper ""MetaQA: Combining Expert Agents for Multi-Skill Question Answering"" Paper: <LINK> w/ @gozde_gul_sahin and @IGurevych at @UKPLab #NL... |
5,27,1443225724559806467,356676252,John Regan,"New paper on the @arxiv - <LINK> Some of the most massive first stars in the Universe were red and very bright! Tyrone Woods led a project where we looked at the emission spectrum from rapidly accreting massive stars and the great news is that @JWSTObserver <LINK> should b... |
III Formation Through Long-Term Stochastic Color Variations",2,"['New paper on the @arxiv - <LINK> Some of the most massive first stars in the Universe were red and very bright! Tyrone Woods led a project where we looked at the emission spectrum from rapidly accreting massive stars and the great news is that @JWSTObs... |
6,79,1260533396310196225,138395156,Professor Sondipon Adhikari,"Our new paper from a collaboration between @SU_engIMPACT and U. of British Columbia entitled ""Machine learning based #digitaltwins for dynamical systems with multiple time-scales"" is now available for download from <LINK>. Please send your comments/sugge... |
time-scales",1,"['Our new paper from a collaboration between @SU_engIMPACT and U. of British Columbia entitled ""Machine learning based #digitaltwins for dynamical systems with multiple time-scales"" is now available for download from <LINK>. Please send your comments/suggestions. <LINK>']",20,05,268 |
7,16,1290085946101178368,1196266674954985472,Nirmal Raj,"New paper with @GrahamKribs and @davemckeen on hunting new forces in Nature by breaking up protons! <LINK> Thread follows. <LINK> Everyone knows the Russian doll-like assembly of matter: atom > nucleus > protons & neutrons > quarks & gluons. This knowled... |
8,30,1221714698812477440,882303076505456642,Timon Emken,"Fresh off the press: Our new #paper on next-generation's direct sub-GeV #DarkMatter searches with semiconductors on today's @arxiv. <LINK> We study future experiments' discovery reaches using frequentist and #MonteCarlo methods. <LINK> @arxiv The paper's emphasis... |
semiconductor detectors",4,"[""Fresh off the press: Our new #paper on next-generation's direct sub-GeV #DarkMatter searches with semiconductors on today's @arxiv.\n\n<LINK>\n\nWe study future experiments' discovery reaches using frequentist and #MonteCarlo methods. <LINK>"", ""@arxiv The paper's emphasis is on the im... |
9,38,1133537573366747136,863828060600139776,Dr. Deep Anand,"check out our new paper on the motions of galaxies in the local Universe! (<LINK>) <LINK> mini-summary: The Milky Way lies in a thin plane (the Local Sheet) right near a large void (the Local Void). Here we test the idea that voids expand by measuring the pec... |
10,54,1151326890331971585,4750366201,Bryan Ostdiek,Very pleased to announce our new paper using #MachineLearning to tell what stars are accreted onto the Milky Way vs those that were born in situ in the @ESAGaia data. See what we find in the data tomorrow! <LINK> @ESAGaia With @linoush95 @SheaGKosmo @AndrewWetzel @astr... |
11,126,1154295187121876997,841703218979733505,Andrei Igoshev,In our recent research <LINK> using @ESAGaia we find that the timescale of common envelope differs for hot subdwarf stars and CVs. Mass ejection happens very fast (~100 years) for sdBs and it takes much longer for cataclysmic variables (~10000 years).,https:/... |
companions",1,['In our recent research <LINK> using @ESAGaia we find that the timescale of common envelope differs for hot subdwarf stars and CVs. Mass ejection happens very fast (~100 years) for sdBs and it takes much longer for cataclysmic variables (~10000 years).'],19,07,251 |
12,99,1372116012309549058,187336449,Luis Welbanks,"Advanced modeling frameworks are essential to reliably extract atmospheric properties from the data of today and tomorrow. Check out our new paper introducing Aurora, a next-generation generalized retrieval framework. Today on the arXiv <LINK> <LINK> @V_Parmentier I'm ... |
Spectra",2,"['Advanced modeling frameworks are essential to reliably extract atmospheric properties from the data of today and tomorrow. Check out our new paper introducing Aurora, a next-generation generalized retrieval framework. Today on the arXiv <LINK> <LINK>', ""@V_Parmentier I'm glad you like the logo! All cre... |
13,65,1070144193362870272,954465907539152897,Dr. Doctor,"New paper from our team analyzing our Dark Energy Camera data from optical follow-up of GW170814! Check it out! <LINK> Spoiler: We didn't find any associated candidates, but we have done the most comprehensive search for optical emission from binary-black-hole me... |
with the Dark Energy Camera",2,"['New paper from our team analyzing our Dark Energy Camera data from optical follow-up of GW170814! Check it out!\n<LINK>', ""Spoiler: We didn't find any associated candidates, but we have done the most comprehensive search for optical emission from binary-black-hole mergers in terms o... |
14,53,1296506144022769666,29997510,Neel Dey,"It never ceases to surprise just how versatile diffeomorphic registration is. In a new #MICCAI OMIA workshop paper with Guillaume Gisbert, we denoise clinical-grade #OCT images where repeat scans with large nonlinear deformations are commonly acquired: <LINK> <LINK> A hybrid... |
Application to Optical Coherence Tomography",2,"['It never ceases to surprise just how versatile diffeomorphic registration is.\n\nIn a new #MICCAI OMIA workshop paper with Guillaume Gisbert, we denoise clinical-grade #OCT images where repeat scans with large nonlinear deformations are commonly acquired: <LINK> <LINK... |
15,34,1420111695364640768,1305574724743974912,Neeraja Gupta,"📢New Working Paper📢 (w. @AlistairEcon & @rigotti_luca) Social scientists now have multiple populations from which they can collect data: which one should they use? The answer depends on cost, convenience, and data quality: The Experimenter's Dilemma. <LINK>... |
16,220,1512029224571445256,1669743397,Michael Alexander Riegler,Are visual explanations of #AI algorithms helpful for #medical doctors? Check our preprint where we conducted a study involving a large number of #gastroenterologists <LINK> @Strumke @sravmd @ThomasdeLange1 @AndreaStoras @vlbthambawita @simula_research,htt... |
intrinsic versus extrinsic explanations",1,['Are visual explanations of #AI algorithms helpful for #medical doctors? Check our preprint where we conducted a study involving a large number of #gastroenterologists <LINK> @Strumke @sravmd @ThomasdeLange1 @AndreaStoras @vlbthambawita @simula_research'],22,04,252 |
17,8,1445350214127206404,1280041792122093568,Yi-Ling Chung,"New #ArgMining #NLProc paper 🤗 🚨Multilingual Counter Narrative Type Classification 🚨 is now out: <LINK> Joint work with @m_guerini and @ragerri A short thread: @m_guerini @ragerri Multilingual and diverse Counter Narratives (CNs) are needed to fight online ... |
18,25,1333403358917496838,887278045761077248,Dmytro Mishkin 🇺🇦,"Efficient Initial Pose-graph Generation for Global SfM <LINK> Our new paper without deep learning invoved! Tl;dr: Bag of tricks for 7x speed-up camera pose graph generation (29 hours for 402 130 image pairs vs 202 hours originally) Details in the thread ... |
19,39,862243144808767488,3716338821,Mikko Tuomi,"New paper of @phillippro: ""Agatha: disentangling periodic signals from correlated noise in a periodogram framework"" <LINK> <LINK> .@phillippro The freely available software ""Agatha"" can calculate useful moving periodograms to study time-invariance of periodic signals... |
periodogram framework",4,"['New paper of @phillippro: ""Agatha: disentangling periodic signals from correlated noise in a periodogram framework"" <LINK> <LINK>', '.@phillippro The freely available software ""Agatha"" can calculate useful moving periodograms to study time-invariance of periodic signals. https://t.co/... |
20,39,1376456668293296128,4249537197,Christian Wolf,New paper by @JannySteeven of our group: We forecast trajectories after observing an initial seq and show that any latent representation of size >= 2m+1 with linear and stable dynamics has a solution. Joint work with V. Andrieu + M. Nadri (LAGEPP). <LINK> <LINK>,ht... |
21,224,1408158879775485953,1124967058414825472,Sindhana Pannir-Sivajothi,My first paper in grad school 🤩 Thanks for being a wonderful advisor @ucsd_yuen! I learned a lot from you @JorgeACamGA and @LamQ310. Our first paper together @ShubhamSinha029 ❤️ We find that polariton condensation can change reaction yields and r... |
22,15,1189536300338044929,4639078397,John Wise,"New paper day! Focusing on the growth of a massive black hole at z=7.5, we find that UV radiation from the SMBH and nuclear stellar cluster regulates runaway star formation within 30 pc, promoting early rapid growth of the SMBH. Led by Ji-hoon Kim (SNU) <LINK> <LINK> Our ... |
Massive Black Holes: Stellar Feedback and Quasar Growth",2,"['New paper day! Focusing on the growth of a massive black hole at z=7.5, we find that UV radiation from the SMBH and nuclear stellar cluster regulates runaway star formation within 30 pc, promoting early rapid growth of the SMBH. Led by Ji-hoon Kim (SNU) <L... |
23,231,1374332704662167557,996420531334369281,Andres Karjus,"Preprint ""Conceptual similarity and communicative need shape colexification"" w/ @DrAlgernon @SimonKirby Tianyu Wang @kennysmithed: <LINK>. We carry out 4 artificial language experiments (incl a self-repl) to test 2 hypotheses from a crosslinguistic study 1/... |
experimental study",6,"['Preprint ""Conceptual similarity and communicative need shape colexification"" w/ @DrAlgernon @SimonKirby Tianyu Wang @kennysmithed: <LINK>. We carry out 4 artificial language experiments (incl a self-repl) to test 2 hypotheses from a crosslinguistic study 1/5 <LINK>', ""2/5 ....Xu et al 202... |
24,63,1329360036766904321,720027280051957761,Oliver Newton,"New paper is out! It feels good to finally get this one out the door. Here, we build on previous work to constrain the properties of warm dark matter models using the satellite galaxies of the MW. We also end up with a lower bound on the MW halo mass <LINK> 1/... |
galaxies of the Milky Way",7,"['New paper is out! It feels good to finally get this one out the door.\n\nHere, we build on previous work to constrain the properties of warm dark matter models using the satellite galaxies of the MW. We also end up with a lower bound on the MW halo mass\n\n<LINK>\n1/7', ""In the first ... |
25,128,1455808495224229891,21902101,Jim Geach,"Check out our new paper on arXiv today ""Realistic galaxy image simulation via score-based generative models"" <LINK> GANs get a lot of glory, but score-based generative methods can produce superior results compared to GANs. Here we use a denoising diffusion probabilistic ... |
26,7,1488421960824537089,948528424926220288,Johannes Røsok Eskilt,"Excited that my new paper is out! <LINK> I'm extremely grateful for the help I've received from so many people! In the paper, I look at the frequency dependence of the cosmic birefringence signal found in Planck data, and I include the LFI for the first... |
HFI Planck Data Release 4",3,"[""Excited that my new paper is out! <LINK> I'm extremely grateful for the help I've received from so many people! In the paper, I look at the frequency dependence of the cosmic birefringence signal found in Planck data, and I include the LFI for the first time"", 'I sample the birefring... |
27,49,1263152573126672385,119837224,Jason Baldridge,"Happy to share a new ACL paper: Li, He, Zhou, Zhang and Baldridge ""Mapping Natural Language Instructions to Mobile UI Action Sequences"". We build datasets and models for learning to take UI actions based on NL instructions. <LINK> <LINK> We evaluate the end-to-end ... |
28,214,1381898128618573829,1199686509328289793,Jean-Baptiste Cordonnier,"I am happy to present our work at #CVPR2021 from my internship with Google Brain Berlin. We study a CNN inefficiency: every part of the image is processed by same network at same resolution. Instead we extract only ""relevant"" patches at full re... |
29,9,1077234804578533376,554869994,Yonatan Belinkov,"Interested in understanding neural networks for #NLProc ? Looking for reading material for your winter break? Check out our new paper, ""Analysis Methods in Neural Language Processing: A Survey"", to appear in TACL. Preprint: <LINK> Website: <LINK> In the paper, we: ... |
30,24,1520435076194082816,947751551258583040,Yoav Levine,"1/2 New paper: ""Sub-Task Decomposition Enables Learning in Seq2seq Tasks"". We prove that unlearnable composite problems can be learned if the network is trained to first predict intermediate steps towards the solution, and only then the solution itself <LINK> ... |
31,40,1064862803679432704,5850692,Aaron Roth,"Excited about a new paper with @datasciwell and @zstevenwu : ""How to Use Heuristics For Differential Privacy"", here: <LINK> We give oracle efficient algorithms for private learning and synthetic data generation, and leave a bunch of interesting open problems.",https://arx... |
32,119,1456167325732904968,1107605372905377792,Aku Venhola🇺🇦,"New paper in ArXiv today about low surface brightness dwarfs in the Fornax cluster: <LINK> In this paper, we applied MTO to find new galaxies from the FDS data. This way we were able to extend our previous FDS dwarf catalog to 821 dwarfs. Below some key fi... |
galaxies in the Fornax cluster",5,"['New paper in ArXiv today about low surface brightness dwarfs in the Fornax cluster:\n<LINK>\nIn this paper, we applied MTO to find new galaxies from the FDS data. This way we were able to extend our previous FDS dwarf catalog to 821 dwarfs. Below some key findings:', '- Massive dw... |
33,167,1279017657178763270,893109085,Edouard Grave,"New work w/ @gizacard (Gautier Izacard): how much do generative models for open domain QA benefit from retrieval? A lot! Retrieving 100 passages, we get 51.4 EM on NaturalQuestions, 67.6 EM on TriviaQA. 1/3 Paper: <LINK> <LINK> Our main finding: generative models are ... |
Question Answering",3,"['New work w/ @gizacard (Gautier Izacard): how much do generative models for open domain QA benefit from retrieval? A lot! Retrieving 100 passages, we get 51.4 EM on NaturalQuestions, 67.6 EM on TriviaQA. 1/3\nPaper: <LINK> <LINK>', 'Our main finding: generative models are great at combining in... |
34,49,1187007570432679936,876274407995527169,David Madras,"New paper on ArXiv! “Detecting Extrapolation with Local Ensembles” (w/ @alexdamour and @james_c_atwood)! <LINK> TL;DR: We present a post-hoc method for approximating the variance of an ensemble, which is useful for estimating prediction reliability. Ensemble tr... |
35,4,1366647976974639106,3094610676,Pranav Rajpurkar,"Common practice for training medical imaging AI models is to use labels extracted from radiology reports. This assumes that report labels are good proxies for image labels. BUT are they? No! 😲 New paper🔎 <LINK> @saahil9jain @AkshaySmit @mattlungrenMD 1/9 <LINK> O... |
Labels and Image Labels",10,"['Common practice for training medical imaging AI models is to use labels extracted from radiology reports.\n\nThis assumes that report labels are good proxies for image labels. BUT are they?\n\nNo! 😲\n\nNew paper🔎 <LINK>\n\n@saahil9jain @AkshaySmit @mattlungrenMD \n\n1/9 <LINK>', 'Over... |
36,10,1510918086509137921,1144657431382958082,Jannis Kurtz,"If you are interested in #robustoptimization with discrete uncertainty sets maybe you want to check our new short paper: <LINK> With Marc Goerigk we study the problem of selecting start scenarios for iterative scenario generation methods for RO problems 1/n O... |
37,292,1321004483988623360,1085279811692621825,Luca Matrà,"And in another great work led by @Sebastromarino today, we find yet another radially wide planetesimal belt with a ⭐️GAP⭐️ around HD206893, hinting at a third companion orbiting the star at 74 au outer to the brown dwarfs HD206893B and C! What a system! <LINK> ... |
38,220,1446391470236254209,1244601028466655232,Hurault Samuel,"<LINK> We propose a new convergent Plug&Play (PnP) scheme with novel theoretical guarantees and state-of-the-art image restoration performance. We train a deep denoiser as an exact gradient descent step on a functional parameterized by a deep neural network... |
39,19,1067918096046927872,7352832,Theo Weber,"New paper on using counterfactual inference for RL! <LINK>. In contrast with ‘classical’ model-based RL questions of type 'what would happen if', counterfactual reasoning's 'what would have happened if' simulates low-bias alternative outcomes of real experience.",https://ar... |
40,98,1273519079593312257,481539448,Richard Alexander,"New paper, led by @PhysicsUoL PhD student (and soon-to-be @QMPlanets postdoc) @GBallabio. Giulia created a new method to understand how emission lines can measure the properties of disc winds, so we can figure out how planet-forming discs lose mass. <LINK> This is ... |
41,34,1519788983852711937,886443925,Philipp Siedler,"New paper on Multi-Agent #ReinforcementLearning (MARL) solving a wild-fire management resource distribution task utilising collaboration, auto-curricula training and an openended environment. Accepted at @GamificationMAS workshop @iclr_conf: <LINK> 1/2 <LINK> Huge th... |
Graph Neural Network Communication Layer for Open-ended Wildfire-Management |
Resource Distribution",2,"['New paper on Multi-Agent #ReinforcementLearning (MARL) solving a wild-fire management resource distribution task utilising collaboration, auto-curricula training and an openended environment. Accepted at @GamificationMAS workshop @iclr_conf: <LINK> 1/2 <LINK>', 'Huge thanks to @serkancabi,... |
42,88,1448717928262615040,1316159755254075392,Moya Chen,New paper - Teach E2E task-oriented models use of new APIs by generating synthetic dialogues with a user simulator; self-train on the successful ones with maybe a sprinkle of examples. Repeat. Profit. Paper: <LINK> See thread below for results + highlights. <LINK... |
Dialogue",1,['New paper - Teach E2E task-oriented models use of new APIs by generating synthetic dialogues with a user simulator; self-train on the successful ones with maybe a sprinkle of examples. Repeat. Profit. \n\nPaper: <LINK>\nSee thread below for results + highlights. <LINK>'],21,10,265 |
43,156,1333736004759523335,426509606,Yamir Moreno,"New preprint out today ""Induced Percolation on Networked Systems"" (<LINK>). We propose a new percolation framework that accounts for indirect influence, finding a rich variety of 1st-order, 2nd-order, and hybrid phase transitions. w/ @xiangrongwang et al. <LINK>",htt... |
44,178,1455131794169270275,995661858508976129,Xingang Pan,"Our #NeurIPS2021 paper ShadeGAN is released at <LINK>. While most 3D-aware GANs are trained via the multi-view constraint, we propose a multi-lighting constraint that resolves the shape-color ambiguity and leads to more accurate 3D shapes! <LINK> <LINK>",https:... |
Image Synthesis",1,"['Our #NeurIPS2021 paper ShadeGAN is released at <LINK>.\nWhile most 3D-aware GANs are trained via the multi-view constraint, we propose a multi-lighting constraint that resolves the shape-color ambiguity and leads to more accurate 3D shapes!\n<LINK> <LINK>']",21,10,253 |
45,35,1495857044439826433,106464999,Kat Volk,"hey, do you need to know the free inclinations of things in the main classical Kuiper belt? (free inclinations are way more useful than ecliptic inclinations!) Check out our new paper led by @yukun_huang that has everything you need! <LINK> I made a simple animation for a r... |
46,40,1419902117758881792,10666172,Sabine Hossenfelder,"New paper, comments welcome <LINK> <LINK> @anothernoki Recht auf körperliche Unversehrtheit. Wenn Du Dich nicht impfen lassen willst weil es Dir egal ist ob Du an COVID stirbst, ist das deine Sache. Wenn Du andere Leute damit in Gefahr bringst nicht. @anothernoki ... |
averaging problem",4,"['New paper, comments welcome <LINK> <LINK>', '@anothernoki Recht auf körperliche Unversehrtheit. Wenn Du Dich nicht impfen lassen willst weil es Dir egal ist ob Du an COVID stirbst, ist das deine Sache. Wenn Du andere Leute damit in Gefahr bringst nicht.', '@anothernoki was redest Du für einen ... |
47,227,1407419464723935238,301450268,Mireia Montes,"Paper day!! Today we focus on the ""enigmatic"" galaxy ""lacking"" dark matter: NGC1052-DF2 (the first one). You can find the accepted paper here: <LINK> @infantesainz @asborlaff <LINK> First, we investigated the globular cluster system of this galaxy and we found tha... |
dark matter NGC 1052-DF2",14,"['Paper day!! Today we focus on the ""enigmatic"" galaxy ""lacking"" dark matter: NGC1052-DF2 (the first one). You can find the accepted paper here: <LINK>\n@infantesainz @asborlaff <LINK>', 'First, we investigated the globular cluster system of this galaxy and we found that all but two ... |
48,193,1380442667964850179,2739464881,Edward Bryant,"What's this? Large TTVs detected for an extremely low density, long period planet?! Check out the paper - <LINK> - to find out how we did this using data from @NextGenTransits and other telescopes, and what the future holds for the fascinating HIP41378 system <LINK>... |
density exoplanet HIP 41378f",1,"[""What's this? Large TTVs detected for an extremely low density, long period planet?! \n\nCheck out the paper - <LINK> - to find out how we did this using data from @NextGenTransits and other telescopes, and what the future holds for the fascinating HIP41378 system <LINK>""]",21,04,... |
49,42,1519352654899691520,1158317142447706112,Stephanie Brandl,"You can find our new paper “How Conservative are Language Models? Adapting to the Introduction of Gender-Neutral Pronouns” which we will present at @naaclmeeting now on arxiv: <LINK>. TL,DR: In this paper, we show that gender-neutral pronouns in Danish, En... |
Gender-Neutral Pronouns",6,"['You can find our new paper “How Conservative are Language Models? Adapting to the Introduction of Gender-Neutral Pronouns” which we will present at @naaclmeeting now on arxiv: <LINK>. \nTL,DR: In this paper, we show that gender-neutral pronouns in Danish, English', 'and Swedish are assoc... |
50,50,1519119528835567617,107357817,daisukelab,"Our new paper is out! We explored a simple masked patch modeling w/o augmentation to learn a latent that describes the input spectrogram as it is. “Masked Spectrogram Modeling using Masked Autoencoders for Learning General-purpose Audio Representation” <LINK> We used MAE ... |
General-purpose Audio Representation",2,"['Our new paper is out! We explored a simple masked patch modeling w/o augmentation to learn a latent that describes the input spectrogram as it is.\n“Masked Spectrogram Modeling using Masked Autoencoders for Learning General-purpose Audio Representation”\n<LINK>', 'We used MA... |
51,154,1384512694183661572,65432023,Prof. Costas Andreopoulos,New #neutrino x-section pheno paper by my PhD student Jùlia Tena-Vidal <LINK>. First in a series of papers on the new tunes of the well-known #GENIE MC. Next: Hadronization. Funded by @livuniphysics LIV.DAT. Initial work supported by an @IPPP_Durham associat... |
52,277,1322301181687857152,888216099757490176,Maithra Raghu,"Do Wide and Deep neural networks Learn the Same Things? Paper: <LINK> We study representational properties of neural networks with different depths and widths on CIFAR/ImageNet, with insights on model capacity effects, feature similarity & characteristic erro... |
Network Representations Vary with Width and Depth",2,"['Do Wide and Deep neural networks Learn the Same Things? \nPaper: <LINK>\n\nWe study representational properties of neural networks with different depths and widths on CIFAR/ImageNet, with insights on model capacity effects, feature similarity & characteristi... |
53,67,1061348205634314240,250846437,Nouha Dziri,"Our paper is now on Arxiv! <LINK> with @korymath, @ehsk0 and @ozaiane. We introduce THRED model which can generate diverse and contextual responses. We also introduce two new, easy to implement evaluation metrics and provide a large clean conversational dataset. @bose_jo... |
Attention",2,"['Our paper is now on Arxiv! <LINK> with @korymath, @ehsk0 and @ozaiane. We introduce THRED model which can generate diverse and contextual responses. We also introduce two new, easy to implement evaluation metrics and provide a large clean conversational dataset.', '@bose_joey @korymath @ehsk0 @ozaian... |
54,180,1268103502892498945,12745042,RJB Goudie,"New on arXiv: Yang Liu @MRC_BSU’s paper with me on inference for modularised Bayes models when ""cutting"" feedback from a module, due to possible misspecification Proposes a new algorithm for the cut distribution: Stochastic Approximation Cut Algorithm <LINK> <LINK>",htt... |
Bayesian Models",1,"['New on arXiv: Yang Liu @MRC_BSU’s paper with me on inference for modularised Bayes models when ""cutting"" feedback from a module, due to possible misspecification\n\nProposes a new algorithm for the cut distribution: Stochastic Approximation Cut Algorithm\n\n<LINK> <LINK>']",20,06,265 |
55,140,1389822601741144066,560668519,Neil Houlsby,"New paper from Brain Zurich and Berlin! We try a conv and attention free vision architecture: MLP-Mixer (<LINK>) Simple is good, so we went as minimalist as possible (just MLPs!) to see whether modern training methods & data is sufficient... <LINK> Mixer consists of ML... |
56,39,1331653437252120580,923231130383536128,Eduardo Fonseca,"🔊New paper out! ""Unsupervised Contrastive Learning of Sound Event Representations"". Work done with @Diego_OrtegoH We learn audio representations by contrasting differently augmented views of sound events. paper: <LINK> code (soon): <LINK> (1/5) <LINK> The... |
57,170,1460857303465283589,1093715503762165760,Ilya Valmianski,"1/2 Check out our new extended abstract accepted to #ML4H2021 “Adding more data does not always help: A study in medical conversation summarization with PEGASUS” <LINK> We show empirical limits of PEGASUS fine-tuned to summarize medical dialogue. 2/2 we fi... |
summarization with PEGASUS",2,"['1/2 Check out our new extended abstract accepted to #ML4H2021 “Adding more data does not always help: A study in medical conversation summarization with PEGASUS”\n\n<LINK>\n\nWe show empirical limits of PEGASUS fine-tuned to summarize medical dialogue.', '2/2 we find that performance ... |
58,268,1368962213621415936,68538286,Dan Hendrycks,"To find the limits of Transformers, we collected 12,500 math problems. While a three-time IMO gold medalist got 90%, GPT-3 models got ~5%, with accuracy increasing slowly. If trends continue, ML models are far from achieving mathematical reasoning. <LINK> <LINK> @flori... |
59,15,1323078447787315200,983857052840636417,Rob Corless,"For those interested in AI in math education, @rotmanphilo maybe, please find my new paper with @JamesHDavenport and @euniceyschan and others at <LINK> . We talk about our respective theory and practice; in some ways old-school, the paper is formed by experience... |
60,148,1511260351307821056,1133734070293323776,Hammer Lab ML,"New (wet) paper alert 🥳🌊 on learning and adapting to the #dynamics in #water distribution networks. A #preprint of our work ""SAM-kNN Regressor for Online Learning in Water Distribution Networks"" by Jonathan Jakob and André Artelt now available @arxiv <LI... |
61,40,1309263951780352000,5025111,Will Whitney,"New paper! We propose to measure the quality of learned representations using the complexity of finding a nearly-optimal predictor on a downstream task. Blog: <LINK> Paper: <LINK> Library: <LINK> The methods that people currently use for evaluating learned representations... |
predictors",10,"['New paper! We propose to measure the quality of learned representations using the complexity of finding a nearly-optimal predictor on a downstream task.\n\nBlog: <LINK>\nPaper: <LINK>\nLibrary: <LINK>', 'The methods that people currently use for evaluating learned representations are related: fix an... |
62,8,957899194328600577,16434310,chrislintott,"A new Galaxy Zoo paper out today - <LINK>. Simulated galaxies below 10^11 solar masses don't look like their counterparts in the real Universe. <LINK> @Space_Mog Yeah, in cosmological simulations (Illustris specifically). I'm not too surprised either but finding the actual... |
Illustris simulation",12,"[""A new Galaxy Zoo paper out today - <LINK>. Simulated galaxies below 10^11 solar masses don't look like their counterparts in the real Universe. <LINK>"", ""@Space_Mog Yeah, in cosmological simulations (Illustris specifically). I'm not too surprised either but finding the actual mass at wh... |
63,80,1416965721020526596,1176335023818997760,Madeleine Zurowski,"New paper on the arxiv today! It’s based around assessing the model independent test of DAMA’s results. We look at how the target mass and background of an experiment will influence its ability to observe the apparent annual modulation 1/2 <LINK> And fin... |
64,42,1055080172284768256,51257255,Natasha Jaques,"Excited to release our new multi-agent RL paper showing that when agents receive a social reward for having causal influence over other agents, it leads to enhanced cooperation and emergent communication <LINK> @daniel_bilar @hardmaru Hey, could you point me to the pap... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.