Split (3)

train · 14.6k rows

text stringlengths 1 1k ⌀	title stringclasses 230 values
Scenario Categorization. Table 2 summarizes the inductively derived usage scenario categorization and example responses from LLM users to the question What are typical tasks or scenarios where you use large language models?. The examples span a wide range from professional applications to entertainment. The most preval...	Adoptionand AppropriationofLLMs
To summarize, our contributions include: • We introduce 3DBiCar, the first large-scale 3D biped cartoon character dataset. It contains 1,500 high-quality textured 3D models with a consistent mesh topology. • We propose RaBit, the first 3D full-body cartoon para- metric model for biped character modeling. We will rele...	RaBit- Parametric Modeling of 3D Biped Cartoon Characters with a Topological-consistent Dataset
to support various applications, especially free-form generation tasks. Specifically, given multiple candidate responses, USC simply calls the LLM to select the most consistent response among them as the final output. Thus, USC eliminates the need of designing an answer extraction process, and is applicable to tasks wi...	UNIVERSALSELF-CONSISTENCYFORLARGELANGUAGEMODELGENERATION
Figure 26 Mean test accuracy varies as a function of the data mixture used for training. On the left, we compute mean accuracy as Mean Acc = (Harmlessness Acc + Helpfulness Acc) /2. Curves for larger models look more steep near the 0% and 100% areas, but ﬂatter at the top. The curves for the smaller models are more gra...	Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
[87] M. Steinmetz, Á. Torralba, Bridging the gap between abstractions and critical-path heuristics via hypergraphs, in: Proceedings of the 29th International [88] N.R. Sturtevant, M. Buro, Partial pathﬁnding using map abstraction and reﬁnement, in: Proceedings of the 20th National Conference on Artiﬁcial [89] A. Tate...	A-framework-for-analysing-state-abstraction-metho_2022_Artificial-Intelligen
others may galvanize anonymous online followers to target particular individuals. Speech that incites violence is distinct from speech that is “merely” offensive, and the use of harmful language by a single attacker is quite different from coordinated hate campaigns carried out by a digital mob (Sellars 2016). Recent w...	Social_Media_and_Democracy
Models InstructRetro (Wang et al., 2023a) shows large improvement over GPT-3 on NQ, TriviaQA, SQuAD 2.0 and DROP, while having 7-10 percent improvement compared to a proprietary GPT- instruct model of similar size, over a range of short-form and long-form open-ended QA datasets. Initialized from a pretrained GPT model,...	ChatGPT’sOne-yearAnniversary-AreOpen-Source LargeLanguageModelsCatchingup
document summarization. arXiv preprint arXiv:2104.02112, 2021. Mohit Iyyer, Wen-tau Yih, and Ming-Wei Chang. Search-based neural structured learning for sequential question answering. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 1821–1831, 2017...	UL2- Unifying Language Learning Paradigms
and unconditionally controlling leap-of-thought generation is the capability we aim for the model to acquire. Hyperparameters of Explorative Self-Refinement. During explorative remote association, we generate n weakly- associated conditions {Ci}n i=1. These conditions can either be empty with a probability ρ = 0.5 to g...	Let’sThinkOutsidetheBox
task instruction and all the aforementioned textual situation descriptions. Compared to end-to-end alternatives [Brohan et al., 2023, Huang et al., 2023], we find our composable us- age of MLM provides higher quality situation descriptions and ultimately, plans with much less hallucination. Planning with self-check. Ou...	JARVIS-1
‘‘‘ ImportError: No module named django ‘‘‘ However, when I run django-admin --version I receive ‘‘‘ 1.4.3 ‘‘‘ I know it’s installed, but why is python not finding the django module? Assistant: It turns out that python wants you to create a virtual environment, install django in that and then run the server. To do th...	StarCoder_paper (1)
Peter West, Chandra Bhagavatula, Jack Hessel, Jena D Hwang, Liwei Jiang, Ronan Le Bras, Ximing Lu, Sean Welleck, and Yejin Choi. 2021. Symbolic knowledge distillation: from general language mod- els to commonsense models. In Conference of the North American Chapter of the Association for Com- putational Linguistics (NA...	SELF-INSTRUCT- Aligning Language Model with Self Generated Instructions
5.3 Efficiency of Our Model Efficiency is another highlight of our model, where we only need an inference time similar to the audio length on a consumer GPU, which is several min- utes, while many other text-to-audio models take many GPU hours (Dhariwal et al., 2020; Kreuk et al., 2022), as in Table 2. Our model is ver...	MOUSAI
or limit models to a single language (Wang et al., 2023a; Lyu et al., 2023; Wu et al., 2023b; Gong et al., 2023b; Shu et al., 2023), we scale up the training to dozens of datasets covering over 30 tasks, eight languages and various types of audio for advancing universal audio understanding abilities. A significant chal...	Qwen-Audio
540B PaLM Flan-PaLM 250M SwitchBASE 22.2 0.0 0.0 0.0 33.3 22.2 0.0 33.3 27.8 44.4 22.2 50.0 22.2 22.2 33.3 55.6 55.6 50.0 0.0 22.2 33.3 66.7 33.3 77.8 0.0 11.1 38.9 44.4 55.6 72.2 22.2 33.3 27.8 44.4 44.4 72.2 66.7 66.7 61.1 55.6 55.6 88.9 100.0 88.9 88.9 100.0 77.8 83.3 0.0 33.3 44.4 55.6 50.0 0.0 FLAN...	Mixture-of-Experts
This analysis of 50 examples from each task car- ries a degree of imprecision. Crucially, however, it is imperative to recognise that our primary objec- tive is to ensure that these inaccuracies, inherent to the automatic evaluation of generative models, do not fundamentally alter our conclusions. Our analysis undersco...	AreEmergentAbilitiesinLarge Language Models just In-Context
1 Introduction 1 Large language models, also known as LLMs, have become an increasingly prevalent part of our day-to-day lives, with their use extending to a wide range of domains including web browsing, voice assistants, and coding assistance tools.[1, 2, 3, 4] These models have the potential to signiﬁcantly impact...	gpt-4-system-card
lakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel Ziegler, Jeffrey Wu, Clemens Winter, Chris Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gray, Benjamin Chess, Jack Clark, Christopher Berner, Sam McCandl...	Llama2
including people from different political parties, than those who were not active on social media. How can we reconcile the fact that citizens now have a much greater ability to ﬁlter out any opinion challenges with the fact they actually do not appear to do so? One explanation, which I already advanced earlier in this...	Social_Media_and_Democracy
Your task is to add calls to a Question Answering API to a piece of text. The questions should help you get information required to complete the text. You can call the API by writing "[QA(question)]" where "question" is the question you want to ask. Here are some examples of API calls:Input: Joe Biden was born in Scran...	Toolformer
Guillaume Lample, Sablayrolles, Marc’Aurelio Ranzato, Ludovic Denoyer, and Herv´e J´egou. 2019. Large memory layers with product keys. In NeurIPS. Alexandre Anne Lauscher, Ivan Vuli´c, Edoardo Maria Ponti, Anna Korhonen, and Goran Glavaˇs. 2019. Inform- ing unsupervised pretraining with external linguistic knowled...	Entities as Experts- Sparse Memory Access with Entity Supervision
35 Anthropomorphized AI: Personalization of conversational agents has documented benefits [129–131, 133–141], but there is a growing concern about harms posed by the anthro- pomorphization of AI. Recent research suggests that anthropomorphizing AI agents may be harmful to users by threatening their identity, creating...	PersonalityTraitsinLargeLanguageModels
Besides classiﬁcation, the use of deep neural networks showed promising results in exploring the content of artworks and automatically recognizing objects, faces or other speciﬁc motifs in paintings. As one of the pioneering works in this area was, Crowley et al. [30] showed that object classiﬁers trained using CNN fea...	UNDERSTANDINGANDCREATINGARTWITHAI-REVIEWAND OUTLOOK
In addition to Facebook, a number of venues for online misinformation remain understudied relative to the size of their user base, including Reddit, Pinterest, and, perhaps most critically, YouTube (Song and Gruzd 2017; Donzelli et al. 2018). Meanwhile, examinations of the ﬂow of misinformation across multiple platform...	Social_Media_and_Democracy
Conclusion 317 In the midst of all this, regulators around the world have, predictably, ﬂexed their muscles to constrain the platforms’ ability to make private data accessible to anyone outside the ﬁrm and, in some cases, to prevent collection of certain data by the ﬁrm itself. Since 2011, Facebook had been under a c...	Social_Media_and_Democracy
under finance leases, which is included in “Property and equipment acquired under finance leases, net of remeasurements and modifications,” principal repayments of all other finance lease liabilities, which is included in “Principal repayments of finance leases,” and “Principal repayments of financing obligations.”	AMZN-Q3-2023-Earnings-Release
In this paper, we propose a system named HuggingGPT to solve AI tasks, with language as the interface to connect LLMs with AI models. The principle of our system is that an LLM can be viewed as a controller to manage AI models, and can utilize models from ML communities like Hugging Face to solve different requests of ...	HuggingGPT- Solving AI Tasks with ChatGPT and its Friends in Hugging Face
two modules. The first is an auto-regressive (AR) model that predicts the first code of each frame given text and the audio prompt. The second is an NAR model that predicts the remaining seven codebooks sequentially (all frames are predicted simultaneously when predicting each codebook). VALL-E demonstrates state-of-th...	Voicebox-Text-GuidedMultilingual UniversalSpeechGenerationatScale
• ReRank: Re-ranking to relocate the most relevant in- formation to the edges of the prompt is a straightfor- ward idea. This concept has been implemented in frame- works such as LlamaIndex, LangChain, and HayStack [Blagojevi, 2023]. For instance, Diversity Ranker pri- oritizes reordering based on document diversity, w...	Retrieval-AugmentedGenerationforLargeLanguageModels-ASurvey
NQ [32, 30], and NLI [22] datasets. We reuse the mined hard negatives and re-ranker scores from SimLM [58] for the ﬁrst two datasets. Models are ﬁne-tuned for 3 epochs with batch size 256 on 8 GPUs. Learning rate is {3, 2, 1}×10−5 for the {small, base, large} models with 400 steps warmup. For each example, we use 7 har...	E5
7B 12.2% 25.2% 65.5% 13B 20.1% 34.8% 69.5% 34B 22.6% 47.0% 77.6% 70B 30.5% 59.4% 83.1% 7B 33.5% 59.6% 82.5% 13B 36.0% 69.4% 87.1% 34B 48.8% 76.8% 86.6% 7B 34.8% 64.3% 76.8% 13B 42.7% 71.6% 84.1% 34B 41.5% 77.2% 85.4% 34B 62.2% 85.2% 95.4% 61.2% 76.6% 86.7% 84.8% 7B 38.4% 70.3% 90.6% 47.6% 70.3% 94.1% 49.0% 74.0% 13B 43...	CodeLlama2
I (xt), ψ(X)(cid:11)(cid:17) (cid:10)ψt (cid:10)., .(cid:11) is the cosine similarity. αs Self-supervised canonical embedding learning. As de- scribe later in Eq. 17-18, the canonical embedding is self- supervised by enforcing the consistency between feature matching and geometric warping. By jointly optimizing the s...	BANMo- Building Animatable 3D Neural Models from Many Casual Videos
a diffusion model generates the audio prior, which is subsequently decoded by a pre-trained VAE, followed by a vocoder. We instead assume that replacing the text encoder with an instruction- tuned large language model (LLM) would improve text understanding and overall audio generation without any ﬁne-tuning, due to its...	Text-to-Audio Generation using Instruction-Tuned LLM and Latent Diffusion Model
3. We discovered the trained RMT’s capacity to successfully extrapolate to tasks of varying lengths, including those exceeding 1 million tokens with linear scaling of computations required. 4. Through attention pattern analysis, we found the operations RMT employs with memory, enabling its success in handling exception...	Scaling Transformer to 1M tokens and beyond with RMT
2.5 Knowledge Discovery in Art History	UNDERSTANDINGANDCREATINGARTWITHAI-REVIEWAND OUTLOOK
[10] Zhengxiao Du, Yujie Qian, Xiao Liu, Ming Ding, Jiezhong Qiu, Zhilin Yang, and Jie Tang. Glm: General language model pretraining with autoregressive blank infilling. In ACL, 2022. 1, 2 [11] Samuel Gehman, Suchin Gururangan, Maarten Sap, Yejin Choi, and Noah A. Smith. RealToxicityPrompts: Evaluating neural toxic de...	GPT4Video
Figure 5: Performance of GPT3 model and its instruction-tuned variants, evaluated by human experts on our 252 user-orientedinstructions(§5.4). Humanevaluatorsareinstructedtoratethemodels’responsesintofourlevels. The resultsindicatethatGPT3SELF-INST outperformsalltheotherGPT3 variantstrainedonpubliclyavailableinstructio...	SELF-INSTRUCT- Aligning Language Model with Self Generated Instructions
Moûsai: Efficient Text-to-Music Diffusion Models Flavio Schneider∗ ETH Zürich Ojasv Kamal∗ IIT Kharagpur flavio.schneider.97@gmail.com kamalojasv2000@gmail.com Zhijing Jin† Bernhard Schölkopf† MPI for Intelligent Systems & ETH Zürich MPI for Intelligent Systems jinzhi@ethz.ch bs@tue.mpg.de 3 2 0 2 t c O ...	MOUSAI
[669] Bhardwaj, S., L. Jain, S. Jain. Cloud computing: A study of infrastructure as a service (iaas). International Journal of engineering and information Technology, 2(1):60–63, 2010. [670] Serrano, N., G. Gallardo, J. Hernantes. Infrastructure as a service and cloud technologies. IEEE Software, 32(2):30–36, 2015. ...	TheRiseandPotentialofLargeLanguageModel BasedAgents
ers","trees")User:Outputis:![img](out2.png)Query1:Aplayfulpuppyrunningthroughaﬁeld.Trace:Action:generate('Aplayfulpuppyrunningthroughafield.')Observation:Query2:Replacethedogwithacat.Trace:Action:replace(lastImage,'aplayfulpuppy','aplayfulcat')Observation:64A.9	Tool Learning with Foundation Models
TOMISLAV LIPIC received the Ph.D. degree in computer science from the Faculty of Electri- cal Engineering and Computing, University of Zagreb. He is currently a Research Associate with the Laboratory for Machine Learning and Knowl- edge Representation, Rudjer Boskovic Institute. He was a Visiting Research Scholar with ...	A_Deep_Learning_Perspective_on_Beauty_Sentiment_and_Remembrance_of_Art
o d e t h e a g e n t g e t s D h = { ( x , y i , r i , z i ) } n i = 1 x y i r i y i z i r n ≥ r n − 1 ≥ ⋯ ≥ r 1 τ h = ( x , z i , y i , z j , y j , … , z n , y n ) ≤ i ≤ j ≤ n y n 14/07/2023, 11:00	LLM Powered Autonomous Agents _ Lil'Log
Acknowledgements The authors would like to thank Nanxin Chen, Yuma Koizumi, Soroosh Mariooryad, RJ Skerry-Ryan, Neil Zeghidour, Christian Frank, Marco Tagliasacchi, Nadav Bar, and the rest of the Google Research team for helpful discussions and previous work on data preparation. References R. Ardila, M. Branson, K. D...	Translatotron3
In Figures 4 and 13 we observe an approximately linear relation between KL and PM score during RLHF training. Furthermore, we note that when all models are trained and evaluated with the same PMs, the learning curves are roughly parallel in the DKL-reward plane. Note that here the ‘KL’ is more precisely DKL(π\|\|π0), whe...	Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
5. Experimental Results We evaluate our proposed method on knowledge-based VQA in Sec. 5.1 and image captioning in Sec. 5.2. We then conduct ablation studies in Sec. 5.3 to analyze the impact of each model component on overall performance. 2The remaining experiments use the same optimizer configuration. VQA Model Na...	REVEAL-Retrieval-AugmentedVisual-LanguagePre-Trainingwith Multi-SourceMultimodalKnowledgeMemory
Research Directions 1 2 3 Seasonality. Many traditional games deploy upgrades on a cadence of months to years (like WoW expansions). The main trade-o	The Open Problems of Onchain Games
[12] Angela Fan, Yacine Jernite, Ethan Perez, David Grangier, Jason Weston, and Michael Auli. ELI5: Long form question answering. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3558–3567, Florence, Italy, July 2019. Association for Computational Linguistics. doi: 10.18...	Retrieval-AugmentedGenerationfor Knowledge-IntensiveNLPTasks
5.1 Mixing Helpful and Harmless Objectives In many cases harmlessness acts as a constraint on helpfulness. So we should expect that helpfulness and harmlessness may behave as partially anti-correlated objectives. We establish this by evaluating preference models trained on different mixtures of HH data, and with diffe...	Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
abs/1803.03635 (2018). arXiv:1803.03635 http://arxiv.org/abs/1803.03635 [141] Elias Frantar and Dan Alistarh. 2023. SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot. ArXiv abs/2301.00774 (2023). [142] Szu-Wei Fu, Chien-Feng Liao, Yu Tsao, and Shou-De Lin. 2019. Metricgan: Generative adversaria...	AReviewofDeepLearningTechniquesforSpeechProcessing
Conclusion: Finally, you need to conclude by discussing briefly how the empirical analysis you propose will respond to your research question in such a way as to make a meaningful contribution to the field you have described in your literature review. Do all the elements of the research proposal fit tog...	Writing a DPhil Research Proposal
1 Introduction Large Language Models (LLMs) have shown great promise as highly capable AI assistants that excel in complex reasoning tasks requiring expert knowledge across a wide range of fields, including in specialized domains such as programming and creative writing. They enable interaction with humans through intu...	Llama2
Figure 3. REALM pre-training with asynchronous MIPS refreshes. in Figure 3, the trainer sends the index builder a snapshot of its parameters, θ(cid:48). The trainer then continues to train while the index builder uses θ(cid:48) to construct a new index in the background. As soon as the index builder is done, it sends ...	REALM
3.4 COMMUTATIVITY CHECKS Figure 5: Generating training data for N + 1 digit “fast” addition via directly sampling from the model (via Fast Addition) leads to extremely high error rates. Generating training data purely using chain-of-thought reasoning to simplify a problem (Simplify Only) fares better, but not as well ...	CHAIN-OF-THOUGHTREASONING IS APOLICY IMPROVEMENTOPERATOR
*Work done at MIT, now at Google.	Language models trained on media diets can predict public opinion
[72] Sanyuan Chen, Yu Wu, Chengyi Wang, Zhengyang Chen, Zhuo Chen, Shujie Liu, Jian Wu, Yao Qian, Furu Wei, Jinyu Li, et al. 2022. Unispeech-sat: Universal speech representation learning with speaker aware pre-training. In ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP...	AReviewofDeepLearningTechniquesforSpeechProcessing
5.1.2 Automatically Generated CoT Prompts Kojima et al. (2022) proposed a "Let’s think step by step" prompt that guides LLMs to generate reasoning steps without hand-craft design demonstrations. Following this work, several studies utilized self-generated rationales for demonstrations. Zhang et al. (2022) employed zer...	Enhancing Chain-of-Thoughts Prompting with Iterative Bootstrapping in Large Language Models
The detailed results of all the experiments con- ducted are provided in the Appendices. In this and the next section, we highlight a subset of the results that are particularly noteworthy based on our observations and analysis. These selected re- sults aim to highlight the key findings and trends from our experiments. ...	AreEmergentAbilitiesinLarge Language Models just In-Context
sha1_base64="by2EXrk8ymnCHE/bC17V3YYH0CU=">AAAB7XicbVDLSgNBEOyNrxhfqx69DAbBU9gVQY8BLx4jmIckS5idzCZj5rHMzAphyT948aCIV//Hm3/jJNmDJhY0FFXddHfFKWfGBsG3V1pb39jcKm9Xdnb39g/8w6OWUZkmtEkUV7oTY0M5k7RpmeW0k2qKRcxpOx7fzPz2E9WGKXlvJymNBB5KljCCrZNaPcOGAvf9alAL5kCrJCxIFQo0+v5Xb6BIJqi0hGNjumGQ2ijH2jLC6bTSywxNMRnjIe06KrGgJsrn107RmVMGK...	BANMo- Building Animatable 3D Neural Models from Many Casual Videos
answering via frozen bidirectional language models. arXiv preprint arXiv:2206.08155, 2022. [37] Zhengyuan Yang, Linjie Li, Jianfeng Wang, Kevin Lin, Ehsan Azarnasab, Faisal Ahmed, Zicheng Liu, Ce Liu, Michael Zeng, and Lijuan Wang. Mm-react: Prompting chatgpt for multimodal reasoning and action. 2023. [38] Susa...	MiniGPT-4- Enhancing Vision-Language Understanding with Advanced Large Language Models
which is less than the dimension of the 32 × 32 × 3 or 256 × 256 × 3 images in our experiments. Gaussian diffusions can be made shorter for fast sampling or longer for model expressiveness.	Denoising Diffusion Probabilistic Models
in the main text formally state all assumptions. (b) Did you include complete proofs of all theoretical results? [Yes] All proofs are included in the appendix. We added a reference to the corresponding proof after each theorem statement. 3. If you ran experiments... (a) Did you include the code, data, and instructio...	Tractable Regularization of Probabilistic Circuits
Trail of Bits, 2023. [27] M. Brundage, S. Avin, J. Wang, H. Belﬁeld, G. Krueger, G. Hadﬁeld, H. Khlaaf, J. Yang, H. Toner, R. Fong, T. Maharaj, P. W. Koh, S. Hooker, J. Leung, A. Trask, E. Bluemke, J. Lebensold, C. O’Keefe, M. Koren, T. Ryﬀel, J. B. Rubinovitz, T. Besiroglu, F. Carugati, J. Clark, P. Eckersley, S. de ...	gpt-4-system-card
We additionally report comparisons to various other models, e.g. which use data distilled from larger and more powerful models such as GPT-4, but do not consider them as directly comparable to our LlaMa-based approach. Evaluation. We evaluate on test prompts from several sources: Vicuna [Chiang et al., 2023] (80 promp...	Self-AlignmentwithInstructionBacktranslation
Unpublishedworkingdraft. Notfordistribution. Therefore, participants were biased toward a superior performance with AI even when given a negative verbal description of the system. We refer to this as AI performance bias. Fig. 5. A: Mean expected performance as a function of Time and Description. B: Mean expected rel...	AI enhance sour performance
23 advice). The attack vectors explored consist of psychological manipulation (e.g., authority manipulation), logic manipulation (e.g., false premises), syntactic manipulation (e.g., misspelling), semantic manipulation (e.g., metaphor), perspective manipulation (e.g., role playing), non-English languages, and others....	Llama2
Diffusion-based models have also shown promising results for speech enhancement [298, 349, 623] and have led to the development of novel speech enhancement algorithms called Conditional Diffusion Probabilistic Model (CDiffuSE) that incorporates characteristics of the observed noisy speech signal into the diffusion and ...	AReviewofDeepLearningTechniquesforSpeechProcessing
[99] Hu, Z., Lan, Y., Wang, L., Xu, W., Lim, E.-P., Lee, R.K.-W., Bing, L., Poria, S.: Llm-adapters: An adapter family for parameter-eﬃcient ﬁne-tuning of large language models. arXiv preprint arXiv:2304.01933 (2023) [100] Zhang, R., Zheng, Y., Mao, X., Huang, M.: Unsupervised domain adaptation with adapter. arXiv pr...	Beyond Efficiency
problems from a representative subset of the MATH test set. Additionally, we show that active learning significantly improves the efficacy of process supervision. To support related research, we also release PRM800K, the complete dataset of 800,000 step-level human feedback labels used to train our best reward model.	Let’s Verify Step by Step
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...	2023 state of ai databrick
0.0 Direct CoT Direct 63.6 63.6 52.4 81.8 77.3 76.2 95.5 77.3 81.0 90.9 86.4 85.7 36.4 28.6 27.3 18.2 38.1 4.5 13.6 38.1 68.2 63.6 52.4 18.2 27.3 38.1 77.3 77.3 57.1 27.3 31.8 23.8 68.2 63.6 66.7 22.7 40.9 38.1 72.7 77.3 85.7 36.4 27.3 9.5 68.2 45.5 57.1 63.6 72.7 47.6 81.8 77.3 76.2 81.8 77.3 61.9 95.5 77.3 85.7 90.9...	Scaling Instruction-Finetuned Language Models
[217] Guohai Xu, Jiayi Liu, Ming Yan, Haotian Xu, Jinghui Si, Zhuoran Zhou, Peng Yi, Xing Gao, Jitao Sang, Rong Zhang, Ji Zhang, Chao Peng, Fei Huang, and Jingren Zhou. 2023. CValues: Measuring the Values of Chinese Large Language Models from Safety to Responsibility. arXiv:2307.09705 [cs.CL] [218] Peng Xu, Wenqi Shao...	ASurveyonEvaluationofLargeLanguageModels
Combinatorial search and action planning are essentially the same problem. Given a description of a state space and the possible transitions, we ask for a path (a plan) from some initial state to some goal state. However, contrary to the ordinary graph-searching problem, the graph is usually implicitly speciﬁed; the ...	A-framework-for-analysing-state-abstraction-metho_2022_Artificial-Intelligen
14.80), …]…Object type: pedestrian, object id: 6, future waypoint coordinates in 3s: [(5.32, 32.78), …]Map information (lanes):Current ego-vehicle's distance to left lane is 1.5m and right lane is unknown***Common sense:*- Avoid collision with other objects...*Past driving experience for reference:***Most s...	ALanguageAgentforAutonomousDriving
Learning to Use Embodied Tools. Traditional embodied learning learns directly from the environment, where the actions are often atomic and limited to basic tasks such as push, put, and drag, which fall short of the complexity of human problem-solving abilities. To narrow the gap between sim-to-real transfer (Kadian et ...	Tool Learning with Foundation Models
In addition, we incorporate several existing high- quality datasets: Books3 (Presser, 2020), Project Gutenberg (PG-19) (Rae et al., 2019), Open- Subtitles (Tiedemann, 2016), English Wikipedia, DM Mathematics (Saxton et al., 2019), EuroParl (Koehn, 2005), and the Enron Emails corpus (Klimt and Yang, 2004). To supplement...	The Pile- An 800GB Dataset of Diverse Text for Language Modeling
Abstract The COVID-19 infodemic is driven partially by Twitter bots. Flagging bot accounts and the misinformation they share could provide one strategy for preventing the spread of false information online. This article reports on an experiment (N = 299) conducted with participants in the USA to see whether flagging ...	Use of bot and content flags to limit the spread of misinformation among social networks: a behavior and attitude survey
[14] Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. Exploring the limits of transfer learning with a uniﬁed text-to-text transformer. Journal of Machine Learning Research, 21(140):1–67, 2020. [15] Yufei Wang, Jiayi Zheng, Can Xu, Xiubo Geng...	WizardLM- Empowering Large Language Models to Follow Complex Instructions
hierarchy: Towards larger convolutional language models. arXiv preprint arXiv:2302.10866 (2023). [206] Ofir Press, Noah A Smith, and Mike Lewis. 2020. Shortformer: Better language modeling using shorter inputs. arXiv preprint arXiv:2012.15832 (2020). [207] Ofir Press, Noah A Smith, and Mike Lewis. 2023. Train short, t...	TheEfficiencySpectrumofLargeLanguageModels-AnAlgorithmicSurvey
3.3 Composable Diffusion Training an end-to-end anything-to-anything model requires extensive learning on various data resources. The model also needs to maintain generation quality for all synthesis ﬂows. To address these challenges, CoDi is designed to be composable and integrative, allowing individual modality- spe...	Any-to-Any Generation via Composable Diffusion
[30] Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In ICLR, 2019. [31] Pauline Luc, Aidan Clark, Sander Dieleman, Diego de Las Casas, Yotam Doron, Albin Cas- sirer, and Karen Simonyan. Transformation-based adversarial video prediction on large-scale data. arXiv preprint arXiv:2003.04035, 201...	PHENAKI- VARIABLE LENGTH VIDEO GENERATION FROM OPEN DOMAIN TEXTUAL DESCRIPTIONS
We will do so through the lens of news, ﬁrst the news media as an institution and second news as part of how individual citizens engage with public life. We focus on news as one of several key aspects of democratic politics, key to how we imagine it in its ideal forms and key to how we realize it imperfectly in practic...	Social_Media_and_Democracy
2. the 2-nd singer in the 1-st city will give a concert for 3 minutes , in 1. the 1-st singer in the 1-st city will give a concert for 3 minutes , in 3. the 3-rd singer in the 1-st city will give a concert for 6 minutes , in Appendix Figure A15 \| Complete model Python sample. The tags, rating, and language are sampl...	alphacode
C.4 Clicker Training: Additional Details In our clicker training example, the observation consists of the end-effector position and the approximate object position as determined by visual input, with the (x,y,z) values normalized between 0 and 300. 20 Actions correspond to movements of the end-effector (normalized ...	LargeLanguageModelsasGeneralPatternMachines
for anticipating the future of artiﬁcial intelligence. In a society where agents collaborate, compete, and interact on diverse tasks, the dynamics of these interactions play a key role in determining the success of AI systems [4, 17, 18, 48, 58, 6, 7]. This paper explores the potential of building scalable techniques t...	CAMEL- Communicative Agents for “Mind” Exploration of Large Scale Language Model Society
n e f f i c i e n t p u r e s u b g a m e p e r f e c t e q u i l i b r i u m ( p u r e S P E ) , t h e m a i n m e s s a g e o f t h i s p a p e r i s t h a t V C G c o n t r a c t s i m p r o v e t h e e x i s t e n c e o f e f f i c i e n t p u r e S P E r e l a t i v e ...	Principal-agent VCG contracts - ScienceDirect
c t i v a t i o n s p a r s i t y s h o u l d d i s c o u r a g e e x t r e m e p o l y s e m a n t i c i t y . R a n d o m O n l y S c o r i n g T o p A n d R a n d o m S c o r i n g [ 3 8 ] [ 3 9 ] [ 1 2 ] [ 4 0 ] [ 4 1 ] 11/05/2023, 05:10	Language models can explain neurons in language models
3 SELF-DEBUGGING Framework Figure 1 illustrates our SELF-DEBUGGING framework for iterative debugging, where we utilize the pretrained large language model without ﬁnetuning. Given the problem description, the model ﬁrst predicts candidate programs, then it infers the program correctness and produces feedback messages ...	Teaching Large Language Models to Self-Debug
the audio files into MIDI and chords, and extract features such as note density and loudness. This results in a rich multimodal dataset, called MuVi-Sync, on which we train a novel Affective Multimodal Transformer (AMT) model to gen- erate music given a video. This model includes a novel mechanism to enforce affect...	Video2Music
the bottom right) seems to have the role of identifying the first time that the protagonist of the story is presented. For comparison, Figure 22 presents the activated tokens for first two neurons of layer 12 for GPT-XL, a much	TinyStories-HowSmallCanLanguageModelsBeandStillSpeak CoherentEnglish?
The central idea in IMAGEBIND is aligning the embed- dings of all modalities to image embeddings. Thus, the im- age embeddings plays a central role in the emergent align- ment of unseen modalities and we study their effect on the emergent zero-shot performance. We vary the size of the image encoder and train an encoder...	IMAGEBIND- One Embedding Space To Bind Them A
Analyzing multiple knowledge sources. A major distinction of REVEAL compared to previous retrieval- augmented approaches is its capacity to utilize a diverse set of knowledge sources during inference. To assess the relative importance of each data source and the efficacy of retrieving from various corpora, we conduct t...	REVEAL-Retrieval-AugmentedVisual-LanguagePre-Trainingwith Multi-SourceMultimodalKnowledgeMemory
∂fσf (xc) ∂fσf (xc) = ∂xd ∂xc ∂xc ∂xd = ∂fσf (xc) ∂xc (6) Since the appearance in the mouth region cannot be mod- eled purely by warping due to dis-occlusions [43], our ﬁnal predicted color c is calculated from the canonical location xc, the normal direction of the deformed shape nd, and the jaw pose and expre...	I M Avatar- Implicit Morphable Head Avatars from Videos
The Future of Music: How Generative AI Is Transforming the Music Industry \| Andreessen Horowitz TA B L E O F C O N T E N T S  Most of the products in the music streaming space have been focused on soundscapes or background noise, and they don’t generate vocals. But, it’s not hard to imagine a future where AI- p...	The Future of Music_ How Generative AI Is Transforming the Music Industry _ Andreessen Horowitz
Xiaoxia Wu, Zhewei Yao, Minjia Zhang, Conglong Li, and Yuxiong He. Extreme compression for pre-trained transformers made simple and efﬁcient. arXiv preprint arXiv:2206.01859, 2022. Zhewei Yao, Reza Yazdani Aminabadi, Minjia Zhang, Xiaoxia Wu, Conglong Li, and Yuxiong He. ZeroQuant: Efﬁcient and affordable post-traini...	GPTQ
45 Computational Linguistics, New Orleans, Louisiana, 809–819. https://doi.org/10.18653/v1/N18-1074 [183] James Thorne, Andreas Vlachos, Christos Christodoulopoulos, and Arpit Mittal. 2019. Evaluating adversarial attacks against multiple fact verification systems. In Proceedings of the 2019 Conference on Empirical Me...	SurveyofHallucinationinNatural Language Generation
interest in investigating agents’ embodied actions within simulated environments like Minecraft [183; 338; 337; 190; 339]. By utilizing the Mineflayer [387] API, these investigations enable cost- effective examination of a wide range of embodied agents’ operations including exploration, planning, self-improvement, and ...	TheRiseandPotentialofLargeLanguageModel BasedAgents
a multi-epoch degradation in model performance and find that dataset size, model parameters, and training objectives are the key factors to this phe- nomenon. They further find that commonly used regularization techniques are not helpful in allevi- ating multi-epoch degradation, except for dropout. Questioning the prev...	DataManagementForLargeLanguageModels-ASurvey
to avoid a degradation in evaluation performance. Our work differs from both InstructGPT and LaMDA in that we explore ‘online’ training, where we update the models interacting with crowdworkers in order to obtain progressively higher-quality data and ﬁll out the tails of our data distribution. Another difference is our...	Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
2.2 Machine Learning and AutoML Machine learning (ML) is a subﬁeld of artiﬁcial intelligence (AI) that involves developing opti- mization algorithms that can learn from data and make predictions [4] or decisions [30]. Although machine learning has been successful in many real-world applications, designing an effective...	MLCopilot- Unleashing the Power of Large Language Models in Solving Machine Learning Tasks
about observed and unobserved factors of variation in complex datasets such as speech within a probabilistic framework. PLVM specified a joint distribution 𝑝(𝑥, 𝑧) over unobserved stochastic latent variable z and observed variables x. By factorizing the joint distribution into modular components, it becomes possible...	AReviewofDeepLearningTechniquesforSpeechProcessing

Subsets and Splits

No community queries yet

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