154 Hits in 1.6 sec

Program Synthesis with Large Language Models [article]

Jacob Austin, Augustus Odena, Maxwell Nye, Maarten Bosma, Henryk Michalewski, David Dohan, Ellen Jiang, Carrie Cai, Michael Terry, Quoc Le, Charles Sutton
2021 arXiv   pre-print
Maarten Bosma created the MBPP dataset, checked for duplication of MBPP data in the training dataset, and gave feedback on the paper.  ... 
arXiv:2108.07732v1 fatcat:ynefbmkupfhftnbulnm72ufzyq

Determinanten van deelname aan een resocialisatieprogramma in Nederlandse penitentiaire inrichtingen

Anouk Bosma, Maarten Kunst, Anja Dirkzwager, Paul Nieuwbeerta
2014 Tijdschrift voor Criminologie  
Questionnaire (CVTRQ)(Ward e.a., 2004, zoals in Bosma e.a., 2013.  ...  Het totale percentage van niet-deelname en uitval lag daarmee in deze periode op 60 procent (Bosma e.a., 2013) . Deze beperkte mate van deelname heeft verschillende consequenties.  ... 
doi:10.5553/tvc/0165182x2014056003003 fatcat:hbieemceafeq7fiyqbusf74yfy

Show Your Work: Scratchpads for Intermediate Computation with Language Models [article]

Maxwell Nye, Anders Johan Andreassen, Guy Gur-Ari, Henryk Michalewski, Jacob Austin, David Bieber, David Dohan, Aitor Lewkowycz, Maarten Bosma, David Luan, Charles Sutton, Augustus Odena
2021 arXiv   pre-print
Large pre-trained language models perform remarkably well on tasks that can be done "in one pass", such as generating realistic text or synthesizing computer programs. However, they struggle with tasks that require unbounded multi-step computation, such as adding integers or executing programs. Surprisingly, we find that these same models are able to perform complex multi-step computations -- even in the few-shot regime -- when asked to perform the operation "step by step", showing the results
more » ... f intermediate computations. In particular, we train transformers to perform multi-step computations by asking them to emit intermediate computation steps into a "scratchpad". On a series of increasingly complex tasks ranging from long addition to the execution of arbitrary programs, we show that scratchpads dramatically improve the ability of language models to perform multi-step computations.
arXiv:2112.00114v1 fatcat:qk5u752x5bdvneqemsuoy7eqke

Splice-Site Mutations: A Novel Genetic Mechanism of Crigler-Najjar Syndrome Type 1

Shailaja Gantla, Conny T.M. Bakker, Bishram Deocharan, Narsing R. Thummala, Jeffry Zweiner, Maarten Sinaasappel, Jayanta Roy Chowdhury, Piter J. Bosma, Namita Roy Chowdhury
1998 American Journal of Human Genetics  
In recent studies, a large number of different Crigler-Najjar syndrome-causing mutations in the exons encoding the UGT1A1 gene have been reported (Bosma et al. 1992a (Bosma et al. , 1992b Ritter et al  ...  All five exons constituting the coding region of bilirubin-UGT 1 and the flanking intronic regions were amplified by PCR, as described elsewhere (Bosma et al. 1992a) .  ... 
doi:10.1086/301756 pmid:9497253 pmcid:PMC1376950 fatcat:6bmjn4b4xfhijabu75ghghps3u

Persistent unconjugated hyperbilirubinemia after liver transplantation due to an abnormal bilirubin UDP-glucuronosyltransferase gene promotor sequence in the donor

Peter L.M. Jansen, Piter J. Bosma, Conny Bakker, Simon P.M. Lems, Maarten J.H. Slooff, Elizabeth B. Haagsma
1997 Journal of Hepatology  
.; Bosma, P.J.; Bakker, C.; Lems, S.P.M.; Slooff, M.J.H.; Haagsma, E.B. with persistent unconjugated hyperbilirubinemia.  ...  Background/Aims: Gilbert's syndrome is genetically characterized by an extra TA element in the TATAAbox of the promotor region upstream of the bilirubin UDP-glucuronosyltransferase ( UGTlA) coding region (Bosma  ... 
doi:10.1016/s0168-8278(97)80272-3 pmid:9252066 fatcat:a4odusx5m5h4hpn6qnco2vfvja

Recidivism After a Prison-Based Treatment Program: a Comparison Between a Treatment and Control Group Using Proportional Weighting Within Strata

Anouk Q. Bosma, Maarten J. J. Kunst, Anja J. E. Dirkzwager, Paul Nieuwbeerta
2020 Journal of Developmental and Life-Course Criminology  
Bosma et al. A. Q. Bosma et al.  ...  Bosma et al.  ... 
doi:10.1007/s40865-020-00139-3 fatcat:jo4yxdzbaffdxjxq7r3olvd67e

Finetuned Language Models Are Zero-Shot Learners [article]

Jason Wei, Maarten Bosma, Vincent Y. Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M. Dai, Quoc V. Le
2022 arXiv   pre-print
Jason Wei, Maarten Bosma, and Quoc V. Le wrote most of the paper. Jason Wei, Maarten Bosma, and Nan Du obtained the zero and few-shot baselines.  ...  AUTHOR CONTRIBUTIONS Maarten Bosma conceived the original idea and implemented the first version of FLAN.  ... 
arXiv:2109.01652v5 fatcat:lmnxjwtddnbbtnxqedb5uf57ye

Mucosal vaccine delivery of antigens tightly bound to an adjuvant particle made from food-grade bacteria

Maarten L. van Roosmalen, Rolf Kanninga, Mohamed El Khattabi, Jolanda Neef, Sandrine Audouy, Tjibbe Bosma, Anneke Kuipers, Eduard Post, Anton Steen, Jan Kok, Girbe Buist, Oscar P. Kuipers (+2 others)
2006 Methods  
Bosma, submitted for publication). Binding of PA-containing proteins to GEM particles Protein anchor fusion proteins quickly bind to GEM particles when added together in solution.  ...  Bosma et al. [31] have shown the versatility and Xexibility of this system by functional display of two enzymes, -amylase and -lactamase, in diVerent ratios onto the surface of GEM particles.  ... 
doi:10.1016/j.ymeth.2005.09.015 pmid:16414272 fatcat:oi2cuxekonhazeomd4geoliaje

A multiplex assay for the quantification of antibody responses in Staphylococcus aureus infections in mice

Sanne van den Berg, M. Gabriela Bowden, Tjibbe Bosma, Girbe Buist, Jan Maarten van Dijl, Willem J. van Wamel, Corné P. de Vogel, Alex van Belkum, Irma A.J.M. Bakker-Woudenberg
2011 JIM - Journal of Immunological Methods  
Bosma (BiOMaDe Technology, Groningen, The Netherlands). ClfB, IsdA, IsdH, FnbpA, and FnbpB were expressed and purified as described previously (Verkaik et al., 2009a) .  ... 
doi:10.1016/j.jim.2010.12.013 pmid:21185300 fatcat:vappozumajfkdc2zdtihhjb6ce

GLaM: Efficient Scaling of Language Models with Mixture-of-Experts [article]

Nan Du, Yanping Huang, Andrew M. Dai, Simon Tong, Dmitry Lepikhin, Yuanzhong Xu, Maxim Krikun, Yanqi Zhou, Adams Wei Yu, Orhan Firat, Barret Zoph, Liam Fedus (+15 others)
2021 arXiv   pre-print
Scaling language models with more data, compute and parameters has driven significant progress in natural language processing. For example, thanks to scaling, GPT-3 was able to achieve strong results on in-context learning tasks. However, training these large dense models requires significant amounts of computing resources. In this paper, we propose and develop a family of language models named GLaM (Generalist Language Model), which uses a sparsely activated mixture-of-experts architecture to
more » ... cale the model capacity while also incurring substantially less training cost compared to dense variants. The largest GLaM has 1.2 trillion parameters, which is approximately 7x larger than GPT-3. It consumes only 1/3 of the energy used to train GPT-3 and requires half of the computation flops for inference, while still achieving better overall zero-shot and one-shot performance across 29 NLP tasks.
arXiv:2112.06905v1 fatcat:sg2j6rsqubhmpp5wtavu636ooe

Who is providing HIV diagnostic testing? Comparing HIV testing by general practitioners and sexual health centres in five regions in the Netherlands, 2011−2018

Saskia J Bogers, Denise E Twisk, Loes M Beckers, Hannelore M Götz, Bram Meima, Michelle Kroone, Elske Hoornenborg, Alewijn Ott, Marleen N Luning-Koster, Nicole H T M Dukers-Muijrers, Christian J P A Hoebe, Carolina J G Kampman (+4 others)
2021 Sexually Transmitted Infections  
ObjectivesGeneral practitioners (GPs) and sexual health centres (SHCs) are the main providers of HIV testing and diagnose two-thirds of HIV infections in the Netherlands. We compared regional HIV testing and positivity by GPs versus SHCs to gain insight into strategies to improve HIV testing, to enable timely detection of HIV infections.MethodsLaboratory data (2011–2018) on HIV testing by GPs and SHCs in five Dutch regions with varying levels of urbanisation were evaluated. Regional HIV testing
more » ... rates per 10 000 residents ≥15 years (mean over period and annual) were compared between providers using negative binomial generalised additive models and additionally stratified by sex and age (15–29 years, 30–44 years, 45–59 years, ≥60 years). χ2 tests were used to compare positivity percentage between the two groups of providers.ResultsIn the study period, 505 167 HIV tests (GP 36%, SHC 64%) were performed. The highest HIV testing rates were observed in highly urbanised regions, with large regional variations. The HIV testing rates ranged from 28 to 178 per 10 000 residents by GPs and from 30 to 378 per 10 000 by SHCs. Testing rates by GPs were lower than by SHCs in three regions and comparable in two. In all regions, men were tested less by GPs than by SHCs; for women, this varied by region. Among those aged 15–29 years old, GPs' testing rates were lower than SHCs', while this was reversed in older age categories in four out of five regions. The overall mean HIV positivity was 0.4%. In contrast to other regions, positivity in Amsterdam was significantly higher among individuals tested by GPs than by SHCs.ConclusionsThis retrospective observational study shows that besides SHCs, who perform opt-out testing for key groups, GPs play a prominent role in HIV testing, especially in non-key populations, such as women and older individuals. Large regional variation exists, requiring region-specific interventions to improve GPs' HIV testing practices.
doi:10.1136/sextrans-2021-055109 pmid:34315804 fatcat:xpaofygji5c6rnz3yopvqaqyce

Chain of Thought Prompting Elicits Reasoning in Large Language Models [article]

Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed Chi, Quoc Le, Denny Zhou
Although scaling up language model size has reliably improved performance on a range of NLP tasks, even the largest models currently struggle with certain reasoning tasks such as math word problems, symbolic manipulation, and commonsense reasoning. This paper explores the ability of language models to generate a coherent chain of thought -- a series of short sentences that mimic the reasoning process a person might have when responding to a question. Experiments show that inducing a chain of
more » ... ught via prompting can enable sufficiently large language models to better perform reasoning tasks that otherwise have flat scaling curves.
doi:10.48550/arxiv.2201.11903 fatcat:noe2p22fljcp5l5m3crauyaipu

Fundamental Physics with the Square Kilometre Array [article]

A. Weltman, H. Padmanabhan, S. Riemer-Sørensen, S. Andrianomena, E. Athanassoula, D. Bacon, R. Barkana, G. Bertone, C. Bonvin, A. Bosma, M. Brüggen, C. Burigana (+32 others)
2019 arXiv   pre-print
These are not straightforward, and a debate is ongoing on the validity of using stellar velocity dispersions to settle this issue (see for example Bosma, 2017, and references therein) .  ...  in constraining the properties and evolution of DM halos since their initial prediction, particularly in the observation of HI rotation curves well beyond the optical radius of galaxies (see, e.g., Bosma  ... 
arXiv:1810.02680v2 fatcat:55m2mrjcezc65iingvhijvbxzy

Fundamental physics with the Square Kilometre Array

A. Weltman, P. Bull, S. Camera, K. Kelley, H. Padmanabhan, J. Pritchard, A. Raccanelli, S. Riemer-Sørensen, L. Shao, S. Andrianomena, E. Athanassoula, D. Bacon (+41 others)
2020 Publications Astronomical Society of Australia  
These are not straightforward, and a debate is ongoing on the validity of using stellar velocity dispersions to settle this issue (see e.g., Bosma 2017, and references therein).  ...  One can find detailed accounts of various ways one can constrain the large-scale geometry of the Universe from observations in two recent reviews (Clarkson & Maartens 2010; Clarkson 2012) .  ... 
doi:10.1017/pasa.2019.42 fatcat:dclqf7mzcrglpphsdo6ohbd4dm

LaMDA: Language Models for Dialog Applications [article]

Romal Thoppilan, Daniel De Freitas, Jamie Hall, Noam Shazeer, Apoorv Kulshreshtha, Heng-Tze Cheng, Alicia Jin, Taylor Bos, Leslie Baker, Yu Du, YaGuang Li, Hongrae Lee (+48 others)
2022 arXiv   pre-print
We present LaMDA: Language Models for Dialog Applications. LaMDA is a family of Transformer-based neural language models specialized for dialog, which have up to 137B parameters and are pre-trained on 1.56T words of public dialog data and web text. While model scaling alone can improve quality, it shows less improvements on safety and factual grounding. We demonstrate that fine-tuning with annotated data and enabling the model to consult external knowledge sources can lead to significant
more » ... ments towards the two key challenges of safety and factual grounding. The first challenge, safety, involves ensuring that the model's responses are consistent with a set of human values, such as preventing harmful suggestions and unfair bias. We quantify safety using a metric based on an illustrative set of human values, and we find that filtering candidate responses using a LaMDA classifier fine-tuned with a small amount of crowdworker-annotated data offers a promising approach to improving model safety. The second challenge, factual grounding, involves enabling the model to consult external knowledge sources, such as an information retrieval system, a language translator, and a calculator. We quantify factuality using a groundedness metric, and we find that our approach enables the model to generate responses grounded in known sources, rather than responses that merely sound plausible. Finally, we explore the use of LaMDA in the domains of education and content recommendations, and analyze their helpfulness and role consistency.
arXiv:2201.08239v3 fatcat:mijuot4z3bdzxa6qplbjrucrem
« Previous Showing results 1 — 15 out of 154 results