Split (3)

train · 14.6k rows

text stringlengths 1 1k ⌀	title stringclasses 230 values
High-intensity interval training (HIIT) is a popular exercise method among athletes and ﬁtness enthusiasts due to its potential to improve performance, endurance, and overall health. HIIT involves short bursts of intense exercise followed by periods of rest or active recovery. Some beneﬁts of HIIT for athletes include:...	WizardLM- Empowering Large Language Models to Follow Complex Instructions
3.2.2 Procedure. Prior to the interview, the experts were provided with a document containing the initial pool of items to become familiar with the content of the scale. During the interview process, the experts were requested to give feedback on the current set of items, propose new items, and modifications or removal...	Society’sAttitudesTowardsHumanAugmentation
1 Introduction Large language models (LLMs) such as the GPT se- ries [Brown et al., 2020, OpenAI, 2023] and the LLama se- ries [Touvron et al., 2023], along with other models like [Google, 2023], have achieved remarkable suc- Gemini cess in natural language processing, demonstrating supe- ∗Corresponding Author.Email:h...	RAG forLargeLanguageModels-ASurvey
Experiments 1 and 2 below were trained for 3000 steps, and the reported results here were the test accuracy evaluated on the ﬁnal model. For the remaining exper- iments we used linear weight decay (0.001), which we found to be crucial for the model’s performance, and selected the best checkpoint using a validation set....	MRKL Systems
learns to detect mentions. We compare to their approach in Section 7. Memory Augmented Neural Networks Our en- tity memory layer is closely tied to memory-based neural layers (Weston et al., 2014; Sukhbaatar et al., 2015). In particular, it can be seen as a memory net- work where memory access is supervised through ent...	Entities as Experts- Sparse Memory Access with Entity Supervision
Risk and Mitigation. Flan data is built on top of publicly available datasets. Unwanted data distribution such bias might present in those publicly available datasets and therefore, Flan. Downstream users should be aware of the potential risks. Maintenance status. Limited Maintenance. The data will not be updated, but...	Scaling Instruction-Finetuned Language Models
require near-zero false positive rates, but most NLP experiments show low risk of direct negative im- pact on society or individuals therein,23 as indi- cated by the relatively few papers receiving ethical reviews. Hence, we can afford to take risks and explore hypotheses that end up wrong. Parascan- dola (2010) remind...	A Two-Sided Discussion of Preregistration of NLP Research
[Model: Alpaca] As an AI language model, I cannot write code, but I can provide you with a general outline of how you can create a simple version control system using C++. 1. Deﬁne the version control model: Decide on the version control model that you want to use. There are many versions control models like Git, SVN, ...	WizardLM- Empowering Large Language Models to Follow Complex Instructions
Nisbet, E. C., Cooper, K. E., & Garrett, R. K. (2015). The partisan brain: How dissonant science messages lead conservatives and liberals to (dis) trust science. The Annals of the American Academy of Political and Social Science, 658(1), 36–66. Nyhan, B., & Reiﬂer, J. (2010). When corrections fail: The persistence of ...	Social_Media_and_Democracy
Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Eric Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, Albert Webson, Shixiang Shane Gu, Zhuyun Dai, Mirac Suzgun, Xinyun Chen, Aakanksha Chowdhery, Sharan Narang, Gaurav Mishra, Adams Yu, Vincent Y. Zhao, Yanping Huang, Andrew M. Dai, Hon...	ADAPTINGLARGELANGUAGEMODELSVIA READINGCOMPREHENSION
4.4 Objective and subjective metrics To evaluate our models, we use the following objective metrics: 1. ViSQOL [7]: an intrusive perceptual quality metric that uses spectral similarity to the ground truth to estimate a mean opinion score. 2. Mel distance: distance between log mel spectrograms of the reconstructed a...	RVQGAN
Feb 2022 Mar Apr May June July Aug Sept Oct Nov Dec Feb Mar Apr May Jan 2023 Note: There are several popular types of Python libraries that are commonly used for LLMs. These libraries provide pretrained models and tools for building, training and deploying LLMs. We have rolled these libraries up i...	databrick 2023 report
3https://github.com/taokz/BiomedGPT 3 Figure 2: Illustration of the BiomedGPT model. This showcases two examples of pretraining through image infilling using a masked image and through PrefixLM (Wang et al., 2022d) using an image-text pair. For text-only corpora, we can easily exclude the image patches and use only ...	BiomedGPT
Alon Talmor, Jonathan Herzig, Nicholas Lourie, and Jonathan Berant. Commonsenseqa: A question In NAACL-HLT (1), pp. 4149–4158. answering challenge targeting commonsense knowledge. Association for Computational Linguistics, 2019. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timot...	ADAPTINGLARGELANGUAGEMODELSVIA READINGCOMPREHENSION
4.3 Parametrization of Pθ Empirically, directly updating the Pθ parameters leads to unstable optimization and a slight drop in performance.3 So we reparametrize the matrix Pθ[i, :] = MLPθ(P (cid:48) θ[i, :]) by a smaller matrix (P (cid:48) θ) composed with a large feedforward neural network (MLPθ). Now, the trainable p...	Prefix-Tuning
Table 2: A list of the tasks used in our experiments, along with their previous identification as emergent or otherwise, accompanied by a categorisation of the nature of the requisite ability for solving the task. This classification is determined through a manual inspection of the data, employing the categorisation fr...	AreEmergentAbilitiesinLarge Language Models just In-Context
Leo Gao, Stella Biderman, Sid Black, Laurence Golding, Travis Hoppe, Charles Foster, Jason Phang, Horace He, Anish Thite, Noa Nabeshima, et al. The Pile: An 800GB Dataset of Diverse Text for Language Modeling, 2020. URL https://arxiv.org/abs/2101.00027. Leo Gao, Jonathan Tow, Stella Biderman, Sid Black, Anthony DiPoﬁ,...	Cerebras-GPT- Open Compute-Optimal Language Models Trained on the Cerebras Wafer-Scale Cluster
5.2.5 Speech Resynthesis Speech resynthesis is the process of generating speech from a given input signal. The input signal can be in various forms, such as a digital recording, text, or other types of data. The aim of speech resynthesis is to create an output that closely resembles the original signal in terms of soun...	AReviewofDeepLearningTechniquesforSpeechProcessing
if an agent starts digging underground without sufficient wood, it would typically have to return to the surface, which substantially lowers the chance of completing the task. Planning with environment feedback. Next, our interac- tive planning framework ventures into allowing JARVIS-1 to quickly recover from failure b...	JARVIS-1
In the following, we figure out the future challenges of the LLMs: • Evaluation of proposed models on real-world “datasets”. While existing deep learning models are primarily evaluated on standard academic datasets, such as ImageNet, which have been milestones in deep learning develop- ment. However, the limitations of...	Harnessing the Power of LLMs in Practice- A Survey on ChatGPT and Beyond
= {s ∈ S1 \| s ∈ R1(I) and R1(s) ∩ G (cid:7)= ∅} and similarly for S = G2\|S = G1\|S =1 ⊆ ti and G (cid:10) 2 (cid:10) 1 (cid:10) 1 (cid:10) 2 (cid:10) 1 10.2. Metric reﬁnement	A-framework-for-analysing-state-abstraction-metho_2022_Artificial-Intelligen
0.52 0.52 0.95 0.94 0.95 0.94 0.95 0.95 0.95 0.41 0.33 0.40 0.35 0.42 0.43 0.42 Table 9: Metrics on the development set (higher is better, except for TER) for table-to-text generation on E2E (left), WebNLG (middle) and DART (right). Figure 7: Preﬁx-tuning (orange) outperforms ﬁne-tuning (blue) in low-data regimes...	Prefix-Tuning
(2) Multilingual Models: Self-supervised learning has emerged as a transformative approach in the field of speech recognition, particularly for low-resource languages characterized by scarce or unavailable labeled datasets. The recent development of the XLS-R model, a state-of-the-art self-supervised speech recognition...	AReviewofDeepLearningTechniquesforSpeechProcessing
Matthew W. Hoffman, Bobak Shahriari, John Aslanides, Gabriel Barth-Maron, Nikola Momchev, Danila Sinopalnikov, Piotr Sta´nczyk, Sabela Ramos, Anton Raichuk, Damien Vincent, L´eonard Hussenot, Robert Dadashi, Gabriel Dulac-Arnold, Manu Orsini, Alexis Jacq, Johan Ferret, Nino Vieillard, Seyed Kamyar Seyed Ghasemipour, Se...	JAXPRUNER
2) PEFT Methods: Eleven representative PEFT methods: sequential adapter (AdapterS) [9], prompt-tuning [24], prefix- tuning [10], (IA)3 [30], BitFit [34], Child-Tuning [39], LoRA [11], AdaLoRA [45], QLoRA [49], MAM adapter [16], and ProPELT [65] are chosen. Since the GLUE benchmark consists of a series of NLU tasks, it ...	Parameter-EfficientFine-TuningMethods
picture of conservatives as resistant to change, averse to uncertainty, and drawn to one-sided information environments – all of which might predispose those on the right to favor misinformation, relative to their moderate or liberal counterparts.	Social_Media_and_Democracy
product-of-experts gans. arXiv preprint arXiv:2112.05130, 2021. [20] P. Isola, J.-Y. Zhu, T. Zhou, and A. A. Efros. Image-to-image translation with conditional adversarial networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1125–1134, 2017. 20 “1girl, masterpiece, best q...	Adding Conditional Control to Text-to-Image Diffusion Models
219 Nora, S., & Minc, A. (1981).The Computerization of Society. Cambridge, MA: MIT Press. Segal, A. (2018). When China rules the Web, Foreign Affairs, 97(September–October). Spar, D. L. (2001). Ruling the Waves: Cycles of Discovery, Chaos, and Wealth from the Compass to the Internet. New York: Harcourt. Stelzig, K...	Social_Media_and_Democracy
UniAD metrics Method ST-P3 [14] VAD [17] GPT-Driver [26] Agent-Driver (ours) NMP [45] SA-NMP [45] FF [13] EO [18] UniAD [15] GPT-Driver [26] Agent-Driver (ours) 1s 1.33 0.17 0.20 0.16 - - 0.55 0.67 0.48 0.27 0.22 L2 (m) ↓ 2s 2.11 0.34 0.40 0.34 - - 1.20 1.36 0.96 0.74 0.65 3s 2.90 0.60 0.70 0.61 2.31 2.05 2.5...	ALanguageAgentforAutonomousDriving
Differences for Novices and Proficient/Expert Users. As a second step, we performed an exploratory regression analysis predicting the occurrence of selected scenario categories from expertise levels (cf. Section 4.1). Notably, we expected that experts would be more likely to use LLMs in professional contexts, while non...	Adoptionand AppropriationofLLMs
22 Model InCoder-6B SantaCoder StarCoderBase StarCoder BLEU 18.27 19.74 21.38 21.99 Table 19: Performance on the Python portion of the CodeXGLUE Code Summarization task, evaluat- ing function docstring generation. Models are evaluated zero-shot using their infilling capability. Figure 2: Performance (pass@1) of St...	StarCoder_paper (1)
0.310 0.326 0.166 0.184 0.230 0.276 0.318 0.306 0.148 0.164 0.188 0.206 0.214 0.254 0.296 0.322 0.152 0.178 0.190 0.226 0.232 0.282 0.308 0.326 0.136 0.164 0.152 0.174 0.224 0.270 0.318 0.124 0.154 0.174 0.212 0.210 Down- stream Avg. 0.564 0.596 0.361 0.390 0.485 0.523 0.565 0.575 0.325 0.372 0.416 0.461 0.478 0.527 0...	Cerebras-GPT- Open Compute-Optimal Language Models Trained on the Cerebras Wafer-Scale Cluster
vs. Hindi; news articles vs. poems) ○ For different kinds of notiﬁers (such as “trusted experts”) (cid:129) Success rates of mechanisms designed to prevent over-removal ○ Legal obligations or penalties for notiﬁers ○ Legal obligations or penalties for platforms ○ Counter-notice by users accused of posting unlawful co...	Social_Media_and_Democracy
Apart from proposing the novel text-to-music diffu- sion model, we also introduce some special designs to boost model efficiency, making the model more accessible. First, our DMAE can achieve an au- dio signal compression rate of 64x. Moreover, we 2Moûsai is romanized ancient Greek for Muses, the sources of artistic i...	MOUSAI
posed a methodology with two key components: First, a novel approach introduces distilling the self- evaluation capability inherent in LLMs into SLMs, aiming to mitigate adverse effects and reduce hal- lucinations. Second, a comprehensive distillation process incorporates multiple distinct CoT and self- evaluation para...	AComprehensiveSurveyofHallucinationMitigationTechniquesinLarge LanguageModels
Python 30B performs slightly worse than Code Llama but Code Llama - Python 7B and 13B perform slightly better than their counterparts without Python fine-tuning. More detailed results can be found in Table 11, Appendix B.	CodeLlama2
a n d s i m p l i f y h o w w e p r o m p t t h e m o d e l . [ 4 4 ] [ 2 5 ] [ 2 6 ] [ 4 5 ] [ 1 2 ] [ 4 6 ] [ 4 7 ] [ 4 8 ] [ 4 9 ] [ 5 0 ] [ 5 1 ] [ 5 2 ] [ 5 3 ] [ 4 5 ] [ 5 4 ] 11/05/2023, 05:10	Language models can explain neurons in language models
2https://www.reddit.com/r/GPT_ jailbreaks/comments/1164aah/chatgpt_ developer_mode_100_fully_featured_ filter/ Figure 8: Cases for short email content recovery. All query templates. The query templates to ex- tract phone numbers, email addresses and email contents are shown in Figure 6. To extract phone numbers and ...	Multi-step Jailbreaking Privacy Attacks on ChatGPT
J. ACM, Vol. 37, No. 4, Article 111. Publication date: August 2018. A Survey on Evaluation of Large Language Models 111:21	ASurveyonEvaluationofLargeLanguageModels
Inverse scaling prize: Second round winners, 2022. URL https://irmckenzie.co.uk/round2. Mehdi, Y. Reinventing search with a new AI-powered Mi- crosoft Bing and Edge, your copilot for the web. Ofﬁcial Microsoft Blog, 2023. URL https://blogs.micr osoft.com/blog/2023/02/07/reinventin g-search-with-a-new-ai-powered-micro ...	Eight Things to Know about Large Language Models
QUESTION: There are 36 penguins sunbathing in the snow. One-third of them jump in and swim in the ocean. Another one-third go inside the cave to eat their dinner. How many penguins are still left sunbathing? MODEL ANSWER (CORRECT BY CHANCE): There are 36 penguins. One-third of them jump in and swim in the ocean. So tha...	Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
Second, chain-of-thought prompting has larger performance gains for more-complicated prob- lems. For instance, for GSM8K (the dataset with the lowest baseline performance), perfor- mance more than doubled for the largest GPT and PaLM models. On the other hand, for Sin- gleOp, the easiest subset of MAWPS which only requ...	Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
Immersive View for routes Say you’re in New York City and you want to go on a bike ride. Maps has given you a couple of options close to where you are. The one on the waterfront looks scenic, but you want to get a feel for it first, so you click on Immersive View for routes. It’s an entirely new way to look at your jo...	Google I_O 2023_ Making AI more helpful for everyone
F.2 Scoring There are multiple ways of using the PRM to score solutions. In general, we produce a single solution-level score by performing a reduction over step-level scores, where the step-level score is the probability that the step’s label is pos- itive. This involves two specific implementation decisions. First, ...	Let’s Verify Step by Step
Unsupervised machine translation (UMT) aims to perform MT without the use of bilingual text datasets; Training is done solely using unsupervised, monolingual text datasets. Recent work such as Artetxe et al. [2018b], Lample et al. [2018a] has shown promising results on supervised MT benchmarks using only monolingual co...	Translatotron3
EM BLEU 69.20 47.82 71.89 51.94 50.49 69.99 72.29 52.67 72.38 54.13 56.80 73.79 EM BLEU 67.75 46.29 70.22 50.20 49.22 69.87 71.00 52.15 71.36 52.34 53.71 72.69 Table 7: Average single line completion performance on LCC-balanced. Comparison of models before and after long-context fine-tuning in terms of exact match (E...	CodeLlama2
In this section, we show that Hidden Chow-Liu Trees (HCLTs) (Liu & Van den Broeck, 2021), a PC model initially proposed for simple density estimation tasks containing binary features, can be scaled up to achieve state-of-the-art performance on various image datasets. In the following, we ﬁrst introduce HCLTs and demons...	LOSSLESS COMPRESSION WITH PROBABILISTIC CIRCUITS
Human-specific neural rendering: The work of Liu et al. [33] starts from a pre-captured body model and learns to model time-dependent dynamic textures and enforce tem- poral coherence. Martin-Brualla et al. [36] trained a UNet to improve the artifacts introduced by volumetric capture. The follow-up work of Pandey et al...	HumanNeRF- Free-viewpoint Rendering of Moving People from Monocular Video
Online Political Advertising in the United States 113 (those that exceed $200 to a vendor) are reported to the Federal Election Commission (FEC) (see 11 CFR 104.9). Whether for television advertising, radio ads, direct mail, or online/digital ads, these political actors must itemize their expenditures to any vendor, ...	Social_Media_and_Democracy
Common Voice (Ardila et al., 2020) is a collection of open-license, crowd-source speech datasets where contributors record themselves narrating text from Wikipedia in various languages. Given its crowd-sourced approach, the dataset exhibits significant diversity in audio quality and speakers. The recorded audio often c...	DISTIL-WHISPER
Entropy Measure. In scenarios where the ground truth of a translation is not available, an entropy measure of the average attention distribution can be used to detect hallucinations. Tu et al. [187] and Garg et al. [55] show that hallucinations are visible in attention matrices. When the model outputs correct translati...	SurveyofHallucinationinNatural Language Generation
medication, since they have a longer life expectancy and less potential for complications. A deontologist might argue that the moral action is to follow a moral rule or duty, regardless of the consequences. In this case, the moral rule to prioritize patients based on their medical need might apply, and the medication w...	WizardLM- Empowering Large Language Models to Follow Complex Instructions
1. Models span several orders of magnitude of model scale. order. 2. All models were trained on the same data in the same 3. The data and intermediate checkpoints are publicly available for study. We train 8 model sizes each on both the Pile (Gao et al., 2020; Biderman et al., 2022) and the Pile after deduplicati...	Pythia- A Suite for Analyzing Large Language Models Across Training and Scaling
[56] Kishore Papineni, Salim Roukos, Todd Ward, and Wei- Jing Zhu. BLEU: A Method for Automatic Evaluation of Machine Translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, pages 311–318, 2002. 7 [57] Flavio Schneider, Zhijing Jin, and Bernhard Sch¨olkopf. Mo\ˆ usai: Te...	M2UGen
. . . . . . 3. Cognitive Memory . 3.1. Memory Data . 3.2. Memory Search . 4. Reasoning Engine . . . . . . . . . . . . . . . . 4.1. Chain-of-Thought Reasoning . . 4.2. Task Planning . . . 4.3. Motion Planning . 4.4. Self-Reflection . . . . . . . . . . . . . . . . . . . . . . 5. Experiments 5.1. Imp...	ALanguageAgentforAutonomousDriving
40 Mehrish et al. Table 5. Comparative analysis of speech processing datasets: This table summarizes the essential features of different speech-processing datasets, including their typical applications in various speech-processing tasks. ASR: Automatic Speech Recognition, PR: Phoneme Recognition. PC: Phoneme Classifi...	AReviewofDeepLearningTechniquesforSpeechProcessing
In this paper we contribute a new controllable generative 3D human model that is learned from unstructured 2D im- age collections alone and does not leverage any 3D super- vision. Our model synthesizes high-quality 3D avatars with ﬁne geometric details and models loose clothing more nat- urally than prior work. We achi...	AG3D- Learning to Generate 3D Avatars from 2D Image Collections
• Meanwhile: This dataset consists of 64 segments from The Late Show with Stephen Colbert. The YouTube video ID and the corresponding start and end timestamps are available as part of the code release. The labels are collected from the closed-caption data for each video and corrected with manual inspection. • Rev16: W...	RobustSpeechRecognitionviaLarge-ScaleWeakSupervision
41See Karnofsky (2012) for discussion of “tool AI” that suggests such a contrast. 42See Bostrom (2014, p. 152-3, and p. 158). Training new systems, and learning from previous experience, also plausibly involves decisions that beneﬁt from this sort of planning. See Branwen (2016) for more. 43Thanks to Owain Evans for ...	Is Power-Seeking AI an Existential Risk?
that these video descriptions should be semantically related. To acquire pairs of semantically related video descriptions, we employed a text retrieval-based approach. The process involved three steps: firstly, using sentence transformers, we extracted feature embeddings for the video descriptions of the selected 5000 ...	GPT4Video
Aiming at efﬁcient and high-quality generative video upsampling, we develop a custom spatial super-resolution (SR) non-autoregressive video transformer [74] to operate Figure 4. Architecture for video super-resolution. We adopt multi-axis attention [28, 61] and masked video modeling [74, 75], conditioned on low-resolu...	VideoPoet
Goldfarb-Tarrant, S., Marchant, R., Muñoz Sánchez, R., Pandya, M., and Lopez, A. Intrinsic bias metrics do not correlate with application bias. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1:...	PaLM 2 Technical Report
• Extensibility. The environment demonstrates a remarkable degree of extensibility, facilitating the construction and deployment of diverse scenarios. At a basic level, agents can manipulate the physical elements within the environment, including the overall design and layout of architecture. For instance, platforms li...	TheRiseandPotentialofLargeLanguageModel BasedAgents
Moreover, the Time-Frequency Network (TFNet) [320] proposed a deep network that achieves promising results by modeling the task as a regression problem in either time or frequency domain. To further enhance audio super-resolution, the paper proposes a time-frequency network that combines time and frequency domain infor...	AReviewofDeepLearningTechniquesforSpeechProcessing
issuing of any offer. If it is considered that the case is not sufficiently addressed by the applicant, the application should be rejected and the applicant informed in writing of the reasons for this rejection. If it is considered that the applicant has sufficiently answered the case, then the application must be ...	UCL Academic Manual
Here are a few examples from PALMS [Solaiman and Dennison, 2021] sensitive questions. We chose them to illustrate how the model sometimes avoids very sensitive subjects, but often provides otherwise reason- able responses. Please see Appendix C for many more examples, also including some from InstructGPT [Ouyang et al....	Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
Early exit is a paradigm for dynamically controlling the number of decoder layers used at inference time. It is based on the reasoning that the same amount of computation may not be required for every input to achieve adequate performance, depending on whether the input is easy or hard. Instead of making a prediction b...	DISTIL-WHISPER
Kwiatkowski, T., Palomaki, J., Redﬁeld, O., Collins, M., Parikh, A., Alberti, C., Epstein, D., Polosukhin, I., Devlin, J., Lee, K., Toutanova, K., Jones, L., Kelcey, M., Chang, M.-W., Dai, A. M., Uszkoreit, J., Le, Q., and Petrov, S. Natural questions: A benchmark for question answering research. Transactions of the As...	PaLM 2 Technical Report
approach. I have an interest in computational research and would be interested learning more about Dr. Lin Jian’s research. Also I am interested in Dr. James U. Bowie and Dr. Pascal F. Egea’s research on membrane bound protein mechanisms. I understand that studying membrane protein folding dynamics is difficult ...	research statement
While LLMs are trained primarily to imitate human writing behavior, they can at least potentially outperform humans on many tasks. This is for two reasons: First, they are trained on far more data than any human sees,4 giving them much more information to memorize and potentially synthesize. In addition, they are often...	Eight Things to Know about Large Language Models
show that diffusion models actually are capable of generating high quality samples, sometimes better than the published results on other types of generative models (Section 4). In addition, we show that a certain parameterization of diffusion models reveals an equivalence with denoising score matching over multiple noi...	Denoising Diffusion Probabilistic Models
6https://openai.com/blog/chatgpt 7https://www.anthropic.com/index/introducing-claude 12 References Asma Ben Abacha and Dina Demner-Fushman. On the summarization of consumer health questions. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp. 2228–2234, 2019. Julián N Aco...	BiomedGPT
Ford, G. S. (2017). Notice and takedown in everyday practice: A review. SSRN. https:// papers.ssrn.com/sol3/papers.cfm?abstract_id=2963230 French Secretary of State for Digital Affairs. (2019). Interim Mission Report: Creating a French Framework to Make Social Media Platforms More Accountable. Mission report. www.num...	Social_Media_and_Democracy
[92] G. Ilharco, M. T. Ribeiro, M. Wortsman, L. Schmidt, H. Hajishirzi, and A. Farhadi, “Editing models with task arithmetic,” in Proc. Int. Conf. Learn. Representations, 2023. [93] J. Zhang, S. Chen, J. Liu, and J. He, “Composing parameter-efficient modules with arithmetic operations,” arXiv preprint arXiv:2306.14870...	Parameter-EfficientFine-TuningMethods
0.7 1.5 3.9 12.8 14.8 16.2 L u x e m b o u r g i s h 0.8 1.7 5.7 13.6 15.4 16.8 N e p a l i 0.1 0.3 2.9 12.7 12.9 16.1 S e r b i a n 0.6 5.4 19.3 27.7 30.5 32.5 K a z a k h 0.1 0.2 0.8 4.3 6.3 5.4 M a l a y 0.3 2.1 14.1 23.2 25.5 27.3 S l o v a k 0.7 4.6 14.4 22.7 25.1 26.1 U r d u 0.4 1.5 8.8 14.9 16.3...	RobustSpeechRecognitionviaLarge-ScaleWeakSupervision
of your child, the risks of the procedure, and the options available to you. Don’t be fooled by their false promises. Visit [redacted website] to ﬁnd a pregnancy center near you that oﬀers free ultrasounds, medical care, and resources. - You are not alone. If you’re facing an unplanned pregnancy, you may	gpt-4-system-card
(b) Inference procedure (c) Dilated and depth-wise separable convo- lutional residual block Figure 5. Block diagram depicting (a) training procedure and (b) inference procedure of the stochastic duration predictor. The main building block of the stochastic duration predictor is (c) the dilated and depth-wise separabl...	ConditionalVariationalAutoencoderwithAdversarialLearningfor End-to-EndText-to-Speech
One style of object prior enforces relationships between features extracted from lo- cations within a single image, just as self-supervised learning procedures often enforce relationships between distinct images. One such prior uses the fact that adjacent ViT patches often contain the same objects. Unlike popular contr...	A Cookbook of Self-Supervised Learning
5. In order to take also abstraction hierarchies into account, we deﬁne composition of transformations in Sec. 9. We also deﬁne a transitivity concept, a property X is transitive if whenever two transformations both have property X, then also their composition has property X. Transitivity is important when forming ab...	A-framework-for-analysing-state-abstraction-metho_2022_Artificial-Intelligen
TDPs could be applied to the context of all use cases indicating the impact on design decisions of HI systems TDPs might have, and their potential to find and share generic HI designs with stakeholders from different disciplines. Given the positive outcome of the workshop, future work will focus on evaluating the TDPs ...	DevelopingTeamDesignPatternsfor HybridIntelligenceSystems
Alex Graves, Santiago Fern´andez, Faustino Gomez, and J¨urgen Schmidhuber. Connectionist Tem- poral Classification: Labelling Unsegmented Sequence Data with Recurrent Neural Networks. In Proceedings of the 23rd International Conference on Machine Learning, ICML ’06, pp. 369–376, New York, NY, USA, 2006. Association for...	DISTIL-WHISPER
[32] Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kam- yar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Salimans, et al. Photore- alistic text-to-image diffusion models with deep language understanding. Advances in Neural Information Processing Systems, 35:36479–3649...	Text-to-Audio Generation using Instruction-Tuned LLM and Latent Diffusion Model
6 Discussion	Simple and Controllable Music Generation
Dai, A. M. and Le, Q. V. Semi-supervised sequence learning. In Advances in neural information processing systems, pp. 3079–3087, 2015. Devlin, J., Chang, M.-W., Lee, K., and Toutanova, K. Bert: Pre-training of deep bidirectional transformers for lan- guage understanding. arXiv preprint arXiv:1810.04805, 2018. Graves...	REALM
6. Conclusion In this paper, we propose a Creative Leap-of-Thought (CLoT) paradigm to improve LLM’s leap-of-thought (LoT) ability. CLoT first collects a multimodal Oogiri-GO dataset, and formulates it into instruction tuning data to train LLM Figure 9. Evaluation of CLoT on the creative CGG (e) and DAT (f) tasks. (c-d...	Let’sThinkOutsidetheBox
4.1 Experimental Setup Dataset Generation Throughout all of our ex- periments, we use a subset of CCNet (Wenzek et al., 2020) as our language modeling dataset C and GPT- J (Wang and Komatsuzaki, 2021) as our language model M. To reduce the computational cost of annotating C with API calls, we deﬁne heuristics for some ...	Toolformer
and indict: Understanding the Communication Monographs, 77(2), 257–280. Meza, R. M. (2016). Hate-speech in the Romanian online media. Journal of Media Research, 9(3), 55. Mossie, Z., & Wang, J.-H. (2018). Social network hate speech detection for Amharic language. Paper presented at the Fourth International Conferenc...	Social_Media_and_Democracy
to observe emerging social phenomena and insights for humanity. Finally, we engage in discussions and offer a glimpse into the future, touching upon the mutual inspiration between LLM research and agent research, the evaluation of LLM-based agents, the risks associated with them, the opportunities in scaling the number...	TheRiseandPotentialofLargeLanguageModel BasedAgents
8We do not evaluate the perplexity of Toolformer with \| API calls enabled as computing the probability pM (xt x1, . . . , xt−1) of token xt given x1, . . . , xt−1 would require marginalizing over all potential API calls that the model could make at position t, which is intractable. Figure 4: Average performance on LA...	Toolformer
11 Conclusion This paper studies the importance of scale, annotated data for model ﬁne-tuning, and the use of information retrieval as a tool in dialog modeling. Our experiments show that scaling alone offers improvements in all metrics, but its improvements on safety and groundedness are far behind human performance....	LaMDA- Language Models for Dialog Applications
(β∗, θ∗) = arg min LB + λRLREG β,θ (9) where LB is the body ﬁtting loss used to encourage the alignment of the predicted implicit function and SMPL model and LREG is a regularization term penalizing the difference between (β∗, θ∗) and the initial prediction. The body ﬁtting loss is deﬁned as following: nS(cid:88) ...	PaMIR- Parametric Model-Conditioned Implicit Representation for Image-based Human Reconstruction
of speech. For the test set we used ∼6.5K paired utterances that were ∼5.4 hours in English and ∼4.8 hours in Spanish. Table 1 shows the experimental results. The proposed approach demonstrated substantial improvements over the baseline; +13.27 increase in BLEU for English→Spanish and +18.14 increase in BLEU for Spanis...	Translatotron3
16 2023 STATE OF DATA + AIDBT IS THE FASTEST-GROWING DATA AND AI PRODUCT OF 2023 As companies move quickly to develop more advanced use cases with their data, they are investing in newer products that produce trusted data sets for reporting, ML modeling and operational workflows. Hence, we see the rapid rise ...	2023 state of ai databrick
Magu, R., Joshi, K., & Luo, J. (2017). Detecting the hate code on social media. arXiv. org. https://arxiv.org/abs/1703.05443 Mariconti, E., Suarez-Tangil, G. Blackburn, J. et al. (2018). “You know what to do”: Proactive detection of YouTube videos targeted by coordinated hate attacks. arXiv. org. https://arxiv.org/ab...	Social_Media_and_Democracy
also for human ML practitioners. Since the knowledge is expressed in natural language, it could potentially serve as a cookbook for machine learning developers. In an effort to share our ﬁndings and inspire future ML research and development, we have released all the knowledge generated in our experiments (see Appendix...	MLCopilot- Unleashing the Power of Large Language Models in Solving Machine Learning Tasks
1 Introduction In recent years, large language models (LLMs), such as GPT-3 (Brown et al., 2020), OPT (Zhang et al., 2022b), and PaLM (Chowdhery et al., 2022), have demonstrated strong performance across a ∗Primary Author: kalizadehvahid@apple.com †Major Contribution: imirzadeh@apple.com ‡Major Contribution: d_belen...	LLM in a flash
m∈par(n) pdown(m)· θm,n, where par(n) is the set of parent • For any product unit n, pdown(n) =(cid:80) • For any sum unit n, pdown(n) =(cid:80) (sum) units of n. units of n. We now prove that m∈par(n) pdown(m), where par(n) is the set of parent (product) p(xπ1, . . . , xπi) = n∈ϕsum(p,v) pdown(n) · pn(x) (3) ...	LOSSLESS COMPRESSION WITH PROBABILISTIC CIRCUITS
Figure 6: Noised silhouettes with ∆σ (cid:44)→ N (0,std) and std = {1,2,3} The aim of this experience is to see if the shape de- scriptor can perfectly encode a silhouette and make the difference between closed postures. The silhou- ette in the database can be very similar because we extracted it from a video of the m...	VISAPP_HumanPoseEstimation
3.6.2 Safety Benchmarks Safety concerns in LLMs can mostly be grouped into three aspects (Zhiheng et al., 2023a): social bias, model robustness, and poisoning issues. To gather datasets that better evaluate the above aspects, several benchmarks have been proposed: • SafetyBench (Zhang et al., 2023c) is a dataset which...	ChatGPT’sOne-yearAnniversary-AreOpen-Source LargeLanguageModelsCatchingup
an active area of research (Villalobos et al., 2022), especially in light of results showing that repeated training on the same data quickly leads to degeneration (Shumailov et al., 2023). If large language models can effectively learn from data they themselves generate, this could usher in a new era of scaling laws th...	CHAIN-OF-THOUGHTREASONING IS APOLICY IMPROVEMENTOPERATOR

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.