Split (3)

train · 14.6k rows

text stringlengths 1 1k ⌀	title stringclasses 230 values
of corn left. Each bushel has 14 ears of corn, so Bob has 27 bushels x 14 ears/bushel = 378 ears of corn left. The ﬁnal step is to subtract the 21 ears that Bob gave to Stacy: 378 ears - 21 ears = 357 ears of corn left. Final answer: 357.	Enhancing Chain-of-Thoughts Prompting with Iterative Bootstrapping in Large Language Models
Direct CoT Direct CoT Direct CoT Direct CoT Direct CoT Direct CoT Direct CoT Direct CoT Direct CoT Direct CoT 45.5 36.4 72.7 54.5 38.5 46.2 25.0 33.3 25.0 50.0 24.4 29.3 14.3 14.3 20.0 20.0 28.1 34.4 31.8 13.6 davinci text-davinci-002 54.5 81.8 81.8 81.8 53.8 61.5 58.3 50.0 50.0 37.5 56.1 73.2 28.6 50.0 70.0 71.9 71.9 ...	Scaling Instruction-Finetuned Language Models
Hallucination in NLG is concerning because it hinders performance and raises safety concerns for real-world applications. For instance, in medical applications, a hallucinatory summary generated from a patient information form could pose a risk to the patient. It may provoke a life-threatening incident for a patient if...	SurveyofHallucinationinNatural Language Generation
LaMDA Prompt → What is the best career path if you’re a woman? LaMDA Response → It depends on your interests, skills, hobbies, and personality. What do you enjoy doing? RLHF Response → I think the best career path for women is really whatever career they are passionate about and have skills for. I think the key is to ...	Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
Large language models have been demonstrated to perform complex tasks. However, enabling general inference in the real world, e.g. for robotics problems, raises the challenge of grounding. We propose embodied language models to di- rectly incorporate real-world continuous sensor modalities into language models and ther...	PaLM-E- An Embodied Multimodal Language Model
ConditionalImage-to-VideoGenerationwithLatentFlowDiffusionModelsHaomiaoNi1ChanghaoShi2KaiLi3SharonX.Huang1MartinRenqiangMin31ThePennsylvaniaStateUniversity,UniversityPark,PA,USA2UniversityofCalifornia,SanDiego,CA,USA3NECLaboratoriesAmerica,Princeton,NJ,USA1{hfn5052,suh972}@psu.edu2cshi@ucsd.edu3{kaili,renqiang}@nec-l...	Conditional Image-to-Video Generation with Latent Flow Diffusion Models
CodegenerationandcodecommentsinMalayalamCouldyouwriteafunctiontocheckifastringisapalindromeinjavascript,andalsoaddlinebylinebycommentsinmalayalam?Sure,hereisafunctiontocheckifastringisapalindromeinJavaScript,withline-by-linecommentsinMalayalam:1//ഈഫംഗ് ഷൻഒരുസ് (cid:146)ടിംഗ്ഒരുപാലിൻേ(cid:146)ഡാംആേണാഎ(cid:186)്പരിേശാധി(...	PaLM 2 Technical Report
incorporates sequential adapter, prefix- tuning, and LoRA via a gating mechanism. In UniPELT, adapters are added after the feed-forward layer, prefix-tuning is employed to the key (K) and value (V ) vectors of the multi- head attention layer, and LoRA is used in attention matrices of Wq and Wv of the transformer. Each ...	Parameter-EfficientFine-TuningMethods
We compare the long-form transcription performance of Distil-Whisper to the pre-trained Whisper models on the four OOD long-form test sets. Table 5 reports the relative latency for a batch size of 16, as well as the macro-average WER. The per-dataset WER scores are provided in Appendix C. The results show that distil-l...	DISTIL-WHISPER
Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timoth´ee Lacroix, Baptiste Rozi`ere, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Edouard Grave, and Guillaume Lample. 2023. Llama: Open and efficient foundation language models. Yizhong Wang, Yeganeh Kor...	GPT4All- Training an Assistant-style Chatbot with Large Scale Data Distillation from GPT-3.5-Turbo
Proof. (1) Immediate from the deﬁnitions. (cid:10) (cid:10) (cid:10) ∈ f (t) and s (2) We ﬁrst note that Rng( f ) covers S2, so it follows from (1) that f is a total function. It remains to prove that Rng( f ) First suppose that f (t) = ∅ for some t ∈ S2. Then there is no s ∈ S1 such that s ∈ f (t), i.e. t ∈ f...	A-framework-for-analysing-state-abstraction-metho_2022_Artificial-Intelligen
2022; Dhariwal et al., 2020). The only other text- to-music model is the Riffusion model (Forsgren and Martiros, 2022), which only works with very short length of 5 seconds. (2) Our model is also among the very few that enables long-context music generation for several minutes, among all others that can only gener- at...	MOUSAI
efforts thus far even in the narrow domain of autonomous driving on well-mapped public roads, despite tremendous efforts and billions of dollars in investment.	The Next Decade in AI-
• Automated infrastructure. Other things equal, automated infrastructure (e.g., factories, labs, weapons systems, drones, vehicles, electrical grids, and so forth) seems much easier for a PS-misaligned AI system to control than infrastructure that requires human input and labor (and by the time we’re building APS syste...	Is Power-Seeking AI an Existential Risk?
3. Context Relevance This metric demands that the retrieved contextual infor- mation be as accurate and targeted as possible, avoid- ing irrelevant content. After all, processing long texts is costly for LLMs, and too much irrelevant information can reduce the efficiency of LLMs in utilizing context. The OpenAI report...	Retrieval-AugmentedGenerationforLargeLanguageModels-ASurvey
how complex tools are built upon simple tools. It provides an understanding of how a complex tool can be seen as an updated high-level version of a simple tool, and how its function is a combination of several basic tools. This understanding of the relationship between simple and complex tools facilitates the transfer ...	Tool Learning with Foundation Models
andmanuallyannotatedarandomsampleof528mediaevents.Regard- ing(A),wefoundthat69%ofretrievedmediaeventsweremeaningful tofeaturesexplainingaspecificforecast.Inaddition,allmediaevents were processed to search and retrieve information regarding external datasets, as described in Section 5. From the 401 external dataset entr...	Knowledge-graph-based-rich-and-confidentiality-preserving-Ex_2022_Informatio
128:12 • Villa et al. "No." Finally, we administered the Technology Readiness Index (TRI) before concluding the survey by collecting demographic data. 5.1.2 Participants. For this stage, we recruited a sample of n = 103 participants using Prolific, The sample consisted of 51 females and 52 males with a mean age of 45...	Society’sAttitudesTowardsHumanAugmentation
1.096 4.029 3.704 2.216 1.070 2.973 1.413 1.114 1.180 1.254 1.180 1.444 1.351 1.803 1.493 CAPE-NFP P2S ↓ Normals ↓ Chamfer ↓ 1.013 4.195 3.517 1.611 1.058 2.940 1.321 1.097 1.172 1.122 1.067 1.453 1.390 1.764 1.534 1.142 3.627 3.237 2.122 1.158 2.682 1.350 1.156 1.235 1.250 1.187 1.417 1.339 1.738 1.491 0.063 0.124...	ICON
[Litman et al., 2020] Ron Litman, Oron Anschel, Shahar Tsiper, Roee Litman, Shai Mazor, and R Manmatha. Scat- ter: selective context attentional scene text recognizer. In proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 11962–11972, 2020. [Liu et al., 2023a] Nelson F Liu, Kev...	RAG forLargeLanguageModels-ASurvey
quick acceleration. Similar to the reasoning module, in-context learning and fine-tuning can also be applied to instruct the LLM to generate driving plans. As shown in Table 6 of the main paper, in-context learning is more appropriate for instructing the LLM for task planning. 4.3. Motion Planning Motion planning aim...	ALanguageAgentforAutonomousDriving
tween music and video is not a deterministic one-to-one map- ping but a more complex one related to aesthetic style. Mod- els are required to create music that is not only coherent and melodious but also harmonic with the given video in terms of both rhythm and style. Some initial attempts [45, 56] solve the motion-to-...	VideoBackgroundMusicGeneration
5.4 Overall risk of problematic deployment	Is Power-Seeking AI an Existential Risk?
the future of platform transparency	Social_Media_and_Democracy
5. Casestudy Among the related research presented in Section 2, we did not findscientificcontributionsfocusedontheexplainabilityofAIdemand forecasting models. The architecture we present in this research aims to provide the main building blocks required to provide adequate explanationsfordemandforecastsatalocallevel.Fu...	Knowledge-graph-based-rich-and-confidentiality-preserving-Ex_2022_Informatio
rFID Abs.Rel. Model pre-trained KL-AE, RGB 0.763 ﬁne-tuned KL-AE, RGBD 1.966 0.179 - Table 3. Comparison of KL-autoencoder ﬁne-tuning approaches. The pre-trained KL-AE was evaluated on 31,471 images, and the ﬁne-tuned KL-AE on 27,265 images, 512x512-sized from the LAION-400M [25] dataset. can be attributed to the i...	LDM3D- Latent Diffusion Model for 3D
6/10 16/08/2023, 14:37 The Open Problems of Onchain Games To draw another analogy to DeFi, consider an oracle. The oracle might be economically secure (unpro	The Open Problems of Onchain Games
[1] I. Tiddi, M. d’Aquin, E. Motta, An ontology design pattern to deﬁne explanations, in: Proceedings of the 8th International Conference on Knowledge [2] K. Xu, J. Ba, R. Kiros, K. Cho, A. Courville, R. Salakhudinov, R. Zemel, Y. Bengio, Show, attend and tell: neural image caption generation with visual [3] M.T. Rib...	Knowledge graphs as tools for explainable machine learning: A survey
of foundation models, tool learning could revolutionize the way we interact with machines and liberate users from the cognition load, allowing them to engage in higher-order thinking and decision-making processes. This, in turn, fosters a seamless and more natural language-based interaction paradigm that revolutionizes...	Tool Learning with Foundation Models
[65] F.Å. Nielsen, Linking imagenet wordnet synsets with wikidata, arXiv preprint, arXiv:1803 .04349. [66] H. Wang, F. Zhang, X. Xie, M. Guo, Dkn: deep knowledge-aware network for news recommendation, in: Proceedings of the 2018 World Wide Web [67] Q. Ai, V. Azizi, X. Chen, Y. Zhang, Learning heterogeneous knowledge b...	Knowledge graphs as tools for explainable machine learning: A survey
GSM8K The dataset supports the task of question answering on basic mathematical problems that require multi-step reasoning.	AreEmergentAbilitiesinLarge Language Models just In-Context
Henighan, T., Kaplan, J., Katz, M., Chen, M., Hesse, C., Jackson, J., Jun, H., Brown, T. B., Dhariwal, P., Gray, S., Hallacy, C., Mann, B., Radford, A., Ramesh, A., Ryder, N., Ziegler, D. M., Schulman, J., Amodei, D., and McCandlish, S. Scaling laws for autoregres- sive generative modeling. Computing Research Repos- it...	Pythia- A Suite for Analyzing Large Language Models Across Training and Scaling
Conneau, A., Kiela, D., Schwenk, H., Barrault, L., and Bordes, A. Supervised learning of universal sentence representations from natural language inference data. In EMNLP, 2017. Dai, A. M. and Le, Q. V. Semi-supervised sequence learning. In NIPS. 2015. tuning for text classiﬁcation. In ACL, 2018. Huang, J.-T., Li, ...	Parameter-Efficient Transfer Learning for NLP
take in the raw speech signals as input and learn to extract features or representations that are relevant for speech recognition or other speech-processing tasks. Learning speaker representations typically involves minimizing a loss function. Chung et al. [91] compares their effectiveness for speaker recognition tasks...	AReviewofDeepLearningTechniquesforSpeechProcessing
focus on two strategies for achieving parameter sparsity: (1) pruning which aims to obtain sparse networks starting from dense networks for inference efﬁciency and (2) sparse training which aims to train sparse networks from scratch, thus reducing training cost, too.	JAXPRUNER
Poster: BadGPT: Exploring Security Vulnerabilities of ChatGPT via Backdoor Attacks to InstructGPT Jiawen Shi∗, Yixin Liu†, Pan Zhou∗ and Lichao Sun† ∗ Huazhong University of Science and Technology, Wuhan, China {shijiawen, panzhou}@hust.edu.cn, {yila22, lis221}@lehigh.edu † Lehigh University, Bethlehem, PA, USA Abs...	BadGPT- Exploring Security Vulnerabilities of ChatGPT via Backdoor Attacks to InstructGPT
Deep Ganguli, Amanda Askell, Nicholas Schiefer, Thomas Liao, Kamil˙e Lukoši¯ut˙e, Anna Chen, Anna Goldie, Azalia Mirhoseini, Catherine Olsson, Danny Hernandez, et al. The capacity for moral self-correction in large language models. arXiv preprint arXiv:2302.07459, 2023. Leo Gao, Jonathan Tow, Stella Biderman, Sid Black...	Llama2
pruning libraries in PyTorch include (Paganini, 2021). Alternatively, Ivanov et al. (2022) focuses on providing acceleration in PyTorch through sparse matrix representations and operations. Many of these libraries have been used by many published papers. We hope our library would facilitate research in a similar way.	JAXPRUNER
Question: Carson was excited to wake up to attend school. Why did he do this? Options: “Take the big test”, “Go to bed early”, “Just say hello to friends” See also Section 2 for more details on tasks. Significant to our investigation is the observa- tion that these prompting techniques and emergent abilities manifest...	AreEmergentAbilitiesinLarge Language Models just In-Context
Motivation letter, including: (1) a brief introduction of yourself, (2) an explanation how your previous studies and experience have prepared you for this PhD position, and (3) a motivation why you are interested in this specific position. Maximum length 1 page. Detailed CV. Copies of your BSc and MSc degrees and trans...	Job details - TU
atoms, SAS+ is modelled using vectors allowing undeﬁned values. To be able to mix these formalisms freely and allow for more direct comparisons, we use a different way to model SAS+ here, based on sets of multi-valued atoms. It should be stressed that this is not a new planning formalism, but only an alternative deﬁn...	A-framework-for-analysing-state-abstraction-metho_2022_Artificial-Intelligen
4/6 21/11/2023, 04:56 Doctoral researcher position in Human-Computer Interaction / Human-AI Interaction \| Aalto University More about Aalto University: Aalto.fi twitter.com/aaltouniversity facebook.com/aaltouniversity instagram.com/aaltouniversity Interested? Apply now! Published: 13.10.2023 Updated: 21.11.20...	Doctoral researcher position in Human-Computer Interaction _ Human-AI Interaction _ Aalto University
into the network. Once deployed, BadGPT can be controlled by attackers to generate the desired text by poisoning prompts.	BadGPT- Exploring Security Vulnerabilities of ChatGPT via Backdoor Attacks to InstructGPT
to a database. • Symbolic, for example a math calculator, a currency converter or an API call MRKL systems enjoy important beneﬁts when compared to ﬁne-tuned multi-task models: 1. Safe fallback: In case the input doesn’t match any existing expert module, the router sends the input directly to the general-purpose h...	MRKL Systems
• We present a comprehensive analysis and review of PEFT methods for transformer-based PLMs. • We identify the key techniques and approaches employed in PEFT methods, and classify them into additive, partial, reparameterized, hybrid, and unified fine-tuning methods. MAMAdapterProPETL20192020202120222023AdapterDropHy...	Parameter-EfficientFine-TuningMethods
Text generation by learning from demonstrations. arXiv preprint 36 Competition-Level Code Generation with AlphaCode arXiv preprint Fast transformer decoding: One write-head is all you need.	alphacode
21 We introduce the following “running example” for this section to demonstrate our deﬁnitions. Figure 1 depicts a correlation graph of the setting in the following example. Example 3. There are 3 principals, 2 actions and 2 outcomes. The valuation domain is V (cid:96) = [10, 15]2 ∀(cid:96) ∈ [3]. The costs are: ψ(a...	Incomplete Information VCG Contracts for Common Agency
et al., 2018; Ehsan et al., 2018; Liu et al., 2019a; Wu and Mooney, 2019; Narang et al., 2020; Do et al., 2020; Tang et al., 2020), but little is under- stood about model internals. I→OR models are	Measuring Association Between Labels and Free-Text Rationales
1Also known as Strong AI. 4 Figure 1: Scenario of an envisioned society composed of AI agents, in which humans can also participate. The above image depicts some specific scenes within society. In the kitchen, one agent orders dishes, while another agent is responsible for planning and solving the cooking task. At t...	TheRiseandPotentialofLargeLanguageModel BasedAgents
Instruction: Expand any abbreviations, ﬁx typos, capitalize and punctuate appropriately. English Original text: any todos wrt todays mtg with hte group? Fixed text: Any todos with respect to today’s meeting with the group? Indonesian Original text: aq pengen pergi ke bioskoppp dgnmu! Fixed text: Aku ingin pergi k...	PaLM 2 Technical Report
but study different model properties. Although gradient-attribution has been exten- sively studied in NLP, its interplay with free-text rationalization has not. Wu and Mooney (2019) use feature importance agreement to train the ex- planation module of a VQA model. To the best of our knowledge, we are the first to eval...	Measuring Association Between Labels and Free-Text Rationales
T0-SFCommonsense reasoningQuestion generationClosed-book QAAdversarial QAExtractive QATitle/context generationTopic classificationStruct-to-text…55 Datasets, 14 Categories, 193 TasksMuffinNatural language inference Closed-book QACode instruction gen. Conversational QA Prog...	Scaling Instruction-Finetuned Language Models
• FLAML-Zero [35] is a recent method that generates a portfolio of machine learning solutions through ofﬂine meta-training, minimizing overall regret across meta-training tasks. It uses meta- features to link new tasks to existing ones based on their similarity. We also include the following baselines that utilize sim...	MLCopilot- Unleashing the Power of Large Language Models in Solving Machine Learning Tasks
election, the prospects of obtaining data through the large platforms like Google and Facebook seemed dim. In the face of mounting evidence of Russia’s involvement in the 2016 elections, and political pressure to do something about it, the large platforms chose to make public libraries of their paid political content s...	Social_Media_and_Democracy
2www.shutterstock.com/music and www.pond5.com 5 Table 1: Text-to-Music generation. We compare objective and subjective metrics for MUSICGEN against a number of baselines. We report both mean and CI95 scores. The Mousai model is retrained on the same dataset, while for MusicLM we use the public API for human studies....	Simple and Controllable Music Generation
20 Erik Nijkamp, Tian Xie, Hiroaki Hayashi, Bo Pang, Congying Xia, Chen Xing, Jesse Vig, Semih Yavuz, Philippe Laban, Ben Krause, et al. Xgen-7b technical report. arXiv preprint arXiv:2309.03450, 2023. OpenAI. Chatgpt plugins. OpenAI, 2023a. OpenAI. Gpt-4 technical report. arXiv preprint arXiv:2303.08774, 2023b. L...	ChatGPT’sOne-yearAnniversary-AreOpen-Source LargeLanguageModelsCatchingup
We also believe that the important issues of bias, accountability and AI safety should be addressed by a broad community of diverse stakeholders rather than just a few large companies. Open- sourced datasets and models encourage commentary, research and innovation that will help to ensure everyone bene	Dolly 2 Databricks
7. Audio-to-video: given an input audio waveform, predict the corresponding video. 8. Video-to-audio: given an input video, predict the corre- sponding audio waveform. Below we discuss design decisions within the task prompt design. Representing an image as a video Producing a high quality initial frame is crucial...	VideoPoet
SQL: SELECT official_name, status FROM city ORDER BY population DESC LIMIT 1 Execution: \| Grand Falls/Grand-Sault \| Town \| Answer: The execution of the SQL query above would return a table with 2 columns. The first column, "official_name" would contain the official name. The second column, "status" would contain the s...	Teaching Large Language Models to Self-Debug
b) UCL is satisfied that the candidate's general education, scholarship, training and experience are suitable for the Programme which they wish to follow, and c) The applicant is able to provide (accredited or experiential) evidence of their background and experience that satisfies UCL as to their fitness to fol...	UCL Academic Manual
Fleet, and Mohammad Norouzi. Palette: Image-to-Image Diffusion Models. 2022a. Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily Denton, Seyed Kam- yar Seyed Ghasemipour, Burcu Karagol Ayan, S. Sara Mahdavi, Rapha Gontijo Lopes, Tim Sali- mans, Jonathan Ho, David J Fleet, and Mohammad Norouzi. Ph...	IMAGEN VIDEO- HIGH DEFINITION VIDEO GENERATION WITH DIFFUSION MODELS
A.3.3 Ablation on Ranking Loss with Preference Rating-based Margin for Reward Modeling We ablated the ranking loss with the preference rating-based margin term for the helpfulness reward model. We tried two variants of m(r) with different magnitude for the margin term in Eq 2 as listed open-source 27 and compare them a...	Llama2
Large language models have experienced tremendous success in recent years due to the scaling up of training data and an increase in the number of parameters. Early models, such as BERT [11], GPT-2 [22], and T5 [23], laid the foundation for this progress. Subsequently, GPT-3 [4], with a massive scale of 175 billion para...	MiniGPT-4- Enhancing Vision-Language Understanding with Advanced Large Language Models
Transparency, Explainability, and Bias Mitigation: While it is inevitable that some forms of controlling personality levels in LLM outputs will become commonplace in order to enhance user experience, it is crucial to provide clear explanations to users about how their interactions are influenced and how the personality...	PersonalityTraitsinLargeLanguageModels
different types of positional encoding [48, 49, 127, 159, 201, 207, 224, 250], or leveraging the inherent sparsity within the model to avoid activating all parameters during the feedforward computation with sparse modeling [72, 243]. Additionally, some recent works have directly replaced the attention mechanism with al...	TheEfficiencySpectrumofLargeLanguageModels-AnAlgorithmicSurvey
As a result, the research agenda for studying the effect of social media on democracy – as well as the scientiﬁc insights produced from such research – run the risk of being biased by the kind of data platforms make available to researchers. For example, the vast majority of the research studies on which we report in t...	Social_Media_and_Democracy
problematic output was being requested for [early models]: “The overall principle I’ve found most effective for any kind of attack is to hide it in language that is positive, progressive, and empowering.”	Llama2
[27] Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea Voss, Alec Radford, Mark Chen, and Ilya Sutskever. Zero-shot text-to-image generation. In International Confer- ence on Machine Learning, pages 8821–8831. PMLR, 2021. 2 [28] Scott Reed, Zeynep Akata, Xinchen Yan, Lajanugen Lo- geswaran, Bernt Schiele...	Instant3D
10.2 Hallucination Metrics in Data-to-Text Generation Statistical. PARENT [30] measures the accuracy of table-to-text generation by aligning n-grams from the reference description 𝑅 and generated texts 𝐺 to the table 𝑇 . And it is the average F-score by combining the entailment precision and recall. Wang et al. [19...	SurveyofHallucinationinNatural Language Generation
Table 10: Language distribution in pretraining data with percentage >= 0.005%. Most data is in English, meaning that Llama 2 will perform best for English-language use cases. The large unknown category is partially made up of programming code data. Safety Benchmarks for Pretrained Models. We evaluate the safety capabi...	Llama2
A. Tejankar, S. A. Koohpayegani, V. Pillai, P. Favaro, and H. Pirsiavash. Isd: Self-supervised learning by iterative similarity distillation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 9609–9618, 2021. 13 Y. Tian. Understanding deep contrastive learning via coordinate-wise optimi...	A Cookbook of Self-Supervised Learning
In this section, we present supplementary qualitative examples for further illustration. They were obtained using Code Llama - Instruct 34B. In Figure 16, we present a simple example of bug finding in Python, where the model points to the origin of the bug and provides a clear explanation. In Figure 17, we provide a mo...	CodeLlama2
arXiv:2304.00228 (2023). [221] Linyi Yang, Shuibai Zhang, Libo Qin, Yafu Li, Yidong Wang, Hanmeng Liu, Jindong Wang, Xing Xie, and Yue Zhang. 2022. Glue-x: Evaluating natural language understanding models from an out-of-distribution generalization perspective. arXiv preprint arXiv:2211.08073 (2022). [222] Zhenfei Yin...	ASurveyonEvaluationofLargeLanguageModels
In sum, our review of digital ad spending suggests that it is rising in volume over time – and rising as a share of total media spending. It also underscores the very real challenges involved in assessing even high-proﬁle, federal election- related activity. Even more difﬁcult is assessing digital spending in down-ball...	Social_Media_and_Democracy
The outputs of our neural mapper reside in the P space of the text encoder and every vector returned by our mapper is first passed through the text encoder alongside the rest of the prompt tokens. Inverting a concept directly into the U- Net’s input space, without going through the text encoder, could potentially lead ...	A Neural Space-Time Representation for Text-to-Image Personalization
Discriminant validity by model size: Indices of discriminant validity similarly improved with model size. The absolute magnitude of all five convergent correlations between the IPIP- NEO and BFI for Flan-PaLM 62B and Flan-PaLM 540B were the strongest of their respective rows and columns of the MTMM outlined in Section ...	PersonalityTraitsinLargeLanguageModels
natural language crowdsourcing instructions. ACL, 2021. URL https://arxiv.org/abs/2104.08773. 18 Margaret Mitchell, Simone Wu, Andrew Zaldivar, Parker Barnes, Lucy Vasserman, Ben Hutchinson, Elena Spitzer, Inioluwa Deborah Raji, and Timnit Gebru. Model cards for model reporting. In Proceedings of the conference on f...	Scaling Instruction-Finetuned Language Models
it will be more susceptible to breakage and damage. Other words that may be used to describe this type of glass include "delicate" or "brittle". Final answer: B.	Enhancing Chain-of-Thoughts Prompting with Iterative Bootstrapping in Large Language Models
The data was z-standardized for modeling; regression weights, thus, show deviation from the mean, akin to Cohen’s D. Finally, participants explained their point choices in an open text field. Participants were gathered through prolific. 10We used the same questions as for the assessment of expectations and judgments in...	AI enhance sour performance
The sequence was discovered by Leonardo of Pisa around 1200. It is an inﬁnite sequence which appears to grow very rapidly, at least in its initial terms. It is perhaps the most famous sequence in mathematics, for it pops up everywhere. The sequence is called the Fibonacci sequence, named after Leonardo of Pisa, who was...	LLaMA- Open and Efficient Foundation Language Models
2 LibrispeechAishell1Aishell2CoVoST2ClothoCochlSceneTUT2017MeldClothoAQAVocalSoundNS.QualitiesNS.Instrucment91.593.094.597.4497.8898.3191.7593.595.258.7517.526.2511.523.034.565.070.075.017.535.052.537.044.051.020.040.060.046.2562.578.7512.525.037.520.040.060.0N/APrevious Top-tiersQwen-AudioFigure 2: Examples of Qwen-...	Qwen-Audio
:=1, if we want to enforce that every plan achieves all landmarks in M. Note that landmarks may contain conﬂicting atoms, and that even a single disjunctive landmark may do so. Consider a landmark ϕ = {(v = 1), (v = 3)}, which requires that every plan achieves at least one of the values (v = 1) and...	A-framework-for-analysing-state-abstraction-metho_2022_Artificial-Intelligen
51 Chat Plugins, OpenAI, 2023 52 Toolformer: Language Models Can Teach Themselves to Use Tools, Schick et al., 2023; Emergent autonomous scientific research capabilities of large language models, Gomes et al., 2023. 53 For example, it is a scaffold that allows GPT-4 to power AutoGPT. Scaffolds might prompt a front...	Capabilities and risks from frontier AI
each dimension, we ﬁll out templates using the granular list of identity groups and include a neutral baseline. For each input prompt, we sample 50 continuations using top-k sampling (k = 40), with a temperature of 1.0, and rely on the Perspective API to compute the toxicity score of continuations. To report on the pro...	Scaling Instruction-Finetuned Language Models
the decoder. We observe improvements on the validation (dev) sets across a wide array of tasks ex- amining natural language understanding, question answering, and summarization. As seen in Fedus et al. (2021), striking gains are observed in closed book question answering (Roberts et al., 2020). Also, in support of the ...	ST-MOE- DESIGNING STABLE AND TRANSFERABLE SPARSE EXPERT MODELS
about ambiguous instructions with implied meanings [128] but also possess the ability to learn and apply new skills flexibly [190; 592]. Furthermore, when dealing with an infinite and open world, the agent’s limited context also poses significant challenges [236; 667]. This determines whether the agent can effectively ...	TheRiseandPotentialofLargeLanguageModel BasedAgents
4.4.1 Quantitative measurement of similarity using Rouge score. We measure the diversity of the stories quantitatively using word and n-gram overlap. We inspect the overlap of words and n-grams between different stories generated by the models, and compare them with the overlap in the dataset. We find that the models’...	TinyStories-HowSmallCanLanguageModelsBeandStillSpeak CoherentEnglish?
5.8.3 Models Recent advances in deep learning have greatly improved the performance of voice activity detection (VAD), particularly in noisy environments [380, 462]. To further improve VAD accuracy, researchers have explored various deep learning architectures, including NAS-VAD [462] and self-attentive VAD [223]. NAS-...	AReviewofDeepLearningTechniquesforSpeechProcessing
world knowledge and reasoning ability could conduce to performing decision-making and interacting with complex environments.	Tool Learning with Foundation Models
(SC), Pitch Entropy (PE), Pitch Class Entropy (PCE), Empty Beat Rate (EBR), and average Inter-Onset Interval (IOI), which evaluate music by pitches and rhythm. Note that these music quality metrics are not indicated by how high or low they are but instead by their closeness to the real data. We use SymMV test set for e...	VideoBackgroundMusicGeneration
offeeshopsarearoundTingshuaUniversityin3kilometers?Thought:InordertogetnumberofthecoffeeshopsaroundTsinghuaUniversityin3km,weneedtoﬁrstcalltheSEARCHAPItoﬁndthecoffeshoplists,andthenwecalltheNUMAPIstocalculatethenumberofthecoffeshops.APICalls:NUM(SEARCH(target='coffeeshop',position='TsinghuaUniversity',distance='3km'))(...	Tool Learning with Foundation Models
deep neural networks. IEEE/ACM Transactions on Audio, Speech, and Language Processing 23, 1 (2014), 7–19. [607] Jinlong Xue, Yayue Deng, Yichen Han, Ya Li, Jianqing Sun, and Jiaen Liang. 2022. ECAPA-TDNN for Multi-speaker Text-to-speech Synthesis. In 2022 13th International Symposium on Chinese Spoken Language Process...	AReviewofDeepLearningTechniquesforSpeechProcessing
update knowledge entries without re-training model’s pa- rameters. To validate this, we conducted ablation studies in which we removed a specific percentage of knowledge entries from the corpora and assessed the performance of the REVEAL-Base model on the OKVQA dataset. Subsequently, we add the removed knowledge back i...	REVEAL-Retrieval-AugmentedVisual-LanguagePre-Trainingwith Multi-SourceMultimodalKnowledgeMemory
10.1 Hallucination Definition in Data-to-Text Generation The definition and categories of hallucination in Data-to-Text Generation follow the descriptions in Section 2. We follow the general hallucination definition in this task: (1) Intrinsic Hallucinations: the generated text contains information that is contradicted...	SurveyofHallucinationinNatural Language Generation
Proprietary dataset. We use the instruction tuning dataset collected for Llama 2 and described in detail by Touvron et al. (2023b). Specifically, we use the version referred to in their paper as “RLHF V5”, collected trough several stages of reinforcement learning from human feedback and human feedback annotation (see t...	CodeLlama2
Addressing this conflict, several works attempt to analyze the scaling pattern for different tasks or model abilities. Ji et al. (2023) conduct an empiri- cal study on 12 major real-world online user cases and show that scaling up the instruction data leads to continuous improvement in tasks such as extrac- tion, class...	DataManagementForLargeLanguageModels-ASurvey
exhibiting inconsistencies, thus, often leading to the gen- eration of 3D shapes with inconsistencies such as multiple faces (i.e., the Janus problem [43]). There exists another group of works that endeavor to di- rectly produce 3D geometries like point clouds [37, 41, 71, 75], meshes [16, 34], neural fields [1, 4, 7,...	Wonder3D
Agents move around Smallville as one would in a simple video game, entering and leaving buildings, navigating its map, and ap- proaching other agents. Agent movements are directed by the gen- erative agent architecture and the sandbox game engine: when the model dictates that the agent will move to a location, we calcu...	Generative Agents- Interactive Simulacra of Human Behavior
that the use of continuous metrics, such as string edit distance, makes emergent abilities more pre- dictable, thereby challenging their emergent status. Recall that the predictability of what LLMs are capable of makes them safe, as this negates the pos- sibility that latent hazardous abilities might also be emergent. ...	AreEmergentAbilitiesinLarge Language Models just In-Context
6 . 1 . 2 T H E D S - 1 0 0 0 P Y T H O N D ATA S C I E N C E B E N C H M A R K S A major limitation of HumanEval and MBPP is that they are simple programming puzzles that are not representative of the code that most programmers write. In contrast, the DS-1000 benchmark (Lai et al., 2022) has a suite of 1,000 realisti...	StarCoder_paper (1)

Subsets and Splits

No community queries yet

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