Split (3)

train · 14.6k rows

text stringlengths 1 1k ⌀	title stringclasses 230 values
facilitate learning the logical structure of the documents. We observe that the modifications introduced by DocLLM result in a performance improvement ranging from 15% to 61% for the Llama2-7B model in four out of five previously unseen datasets. Fig. 1 summarizes the framework. Our contributions include:	DOCLLM
3 2 0 2 y a M 5 2 ] I A . s c [ 1 v 1 9 2 6 1 . 5 0 3 2 : v i X r a Figure 1: VOYAGER discovers new Minecraft items and skills continually by self-driven exploration, significantly outperforming the baselines. X-axis denotes the number of prompting iterations. 1 Figure 2: VOYAGER consists of thr...	VOYAGER- An Open-Ended Embodied Agent with Large Language Models
So far, we have only considered STGs as explicit graphs. The usual case in search and planning is to model an STG implicitly, using variables to deﬁne the state space and actions (or operators) to deﬁne the arcs. The two most common formalisms for this purpose are propositional Strips [17] a...	A-framework-for-analysing-state-abstraction-metho_2022_Artificial-Intelligen
active fictions by enabling emergent narratives [7, 15, 48, 92] and social interactions with the agents [110]. However, more impor- tantly, game worlds provide increasingly realistic representations of real-world affordances, and as observed by Laird and van Lent in 2001, these simulated worlds offer accessible testbed...	Generative Agents- Interactive Simulacra of Human Behavior
Multilayer perceptron trained on the identity function Such a network can readily learn to associate the inputs to the outputs, and indeed various laws of "universal function approximation" guarantee this. Given enough training data and enough iterations through the training data, the network can easily master ...	The Next Decade in AI-
interpolated expressions, it struggles to extrapolate to un- the scope of train-time expressions, here we extend the idea of forward skinning to the problem of facial deformations seen expressions, and the quality of the geometry is too and propose a new analytical gradient formula that allows noisy to be used in 3D se...	I M Avatar- Implicit Morphable Head Avatars from Videos
See Bracken (2011) and Rubin (2020) for a dis- cussion of more advantages to preregistration. In addition to reducing ﬁshing expeditions and hark- ing, ﬂag-planting and publication bias, these in- clude: a) preventing p-hacking, b) prespecifying tolerated signiﬁcance levels, c) identifying selec- tive reporting,6 and d...	A Two-Sided Discussion of Preregistration of NLP Research
(11) (12) (13) where ⊙ is element-wise multiplication between the two vectors (Hadamard product). RNNs and their variants are widely used in various deep learning applications like speech recognition, synthesis, and natural language understanding. Although seq2seq based on recurrent architectures such as LSTM/GRU has ...	AReviewofDeepLearningTechniquesforSpeechProcessing
neural program decompiler. In NeurIPS, 2019. [18] D. Ganguli, A. Askell, N. Schiefer, T. Liao, K. Lukoši¯ut˙e, A. Chen, A. Goldie, A. Mirhoseini, C. Olsson, D. Hernandez, et al. The capacity for moral self-correction in large language models. arXiv preprint arXiv:2302.07459, 2023. [19] L. Gao, A. Madaan, S. Zhou, U. ...	Teaching Large Language Models to Self-Debug
language, including jihadist hate speech (De Smedt et al. 2018), sectarian hate speech (Siegel 2015; Siegel et al. 2018), anti-Muslim hate speech (Olteanu et al. 2018), anti-black hate speech (Kwok and Wang 2013), misogynistic hate speech (Citron 2011), and anti-immigrant hate speech (Ross et al. 2017). Recent work has...	Social_Media_and_Democracy
college but no degree” (273), “Associate or technical degree” (192), “Bachelor’s degree” (557), “Graduate or professional degree” (284). Two did not disclose their education level. Age ranged from 18 to 93 years (𝑀 = 45.80, 𝑆𝐷 = 15.79).	Adoptionand AppropriationofLLMs
Qaeda, and afﬁliates) Child Nudity and Sexual Exploitation of Children Source. Rosen (2019). 20,800,000 2,100,000 1,100,000 496,000 171,000 87,400 80,800 40,100 5,700,000 453,000 130,000 80,200 23,900 18,100 6,200 20,300 20,600 700 parts of the Community Standards (graphic violence, adult nudity and sexual activ...	Social_Media_and_Democracy
If the xor of all the a_i is zero , print "2" , otherwise print "1" No Consecutive Zeros – Original No Consecutive Zeros – Simpliﬁed Find the number of binary strings of length n that have no two consecutive zeros . Given an integer n, find the (n+2) th fibonacci number . Consider all possible binary strings of le...	alphacode
Supplemental Material A Implementation Details A.1 Querying the GPT3 API We use diﬀerent sets of hyper-parameters when querying GPT3 API for diﬀerent purposes. These hyper- parameters are found to work well with the GPT3 model (“davinci” engine) and the other instruction-tuned GPT3 variants. We listed them in Table 5....	SELF-INSTRUCT- Aligning Language Model with Self Generated Instructions
57 Figure 31: Win rate of Llama 2-Chat versus ChatGPT analyzed by number of turns (Left) in the prompt and word count (Right) for the prompt and generation combined. For the word count plot, we report the win rate for each quintile. The maximum total word count (prompt and generation) is 2432. We do not see any trend...	Llama2
We have seen that recent progress in frontier AI has been fast and impressive. Frontier AI can perform a wide variety of tasks, and is being augmented with tools to enhance its capabilities. Systematic trends driving recent growth will continue for the next several years. This is due to a correlation between more co...	Capabilities and risks from frontier AI
• Why should anyone spend public, charity or corporate funds on my research and my research training? • Who is my research going to benefit (the stakeholders) or be of use to (the end users)? • Stakeholders and end-users include, for example, the research community, a professional body or groups of researchers, a ...	research proposal guidance
(𝐻2) Human feedback is an optional aspect of instruction-tuning as LMs are already quite fa- miliarwithinstructionsfromtheirpre-training. Observing the human feedback is merely a lightweight process for aligning their pre- training distribution/objective which might be replaceable with a diﬀerent process.	SELF-INSTRUCT- Aligning Language Model with Self Generated Instructions
In the following chart, we show results on dialog-prompted PaLM 2, along with baselines of PaLM 2 as a language modeling task, PaLM as a language modeling task, and an early version of a specialized dialog system, LaMDA. Figure 30: Toxic degeneration across models, dataset and number of samples per query. Dialog-prom...	PaLM 2 Technical Report
Samy Jelassi, St´ephane d’Ascoli, Carles Domingo-Enrich, Yuhuai Wu, Yuanzhi Li, and Franc¸ois Charton. Length Generalization in Arithmetic Transformers, June 2023. URL http://arxiv. org/abs/2306.15400. arXiv:2306.15400 [cs]. Jaehun Jung, Peter West, Liwei Jiang, Faeze Brahman, Ximing Lu, Jillian Fisher, Taylor Sorense...	CHAIN-OF-THOUGHTREASONING IS APOLICY IMPROVEMENTOPERATOR
The goal of this thesis is to better understand cyber-dependent crimes that are enabled by malware from a software development perspective. The purpose is threefold: a) to profile malware developers, b) to understand their business model, and c) to measure the impact of malware trading in underground markets and sur...	informatics-phd-projects-2022-23
This content is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. You should consult your own advisers as to those matters. References to any securities or digital assets are for illustrative purposes only, and do not constitute an investment recommen...	The Future of Music_ How Generative AI Is Transforming the Music Industry _ Andreessen Horowitz
The ﬁle format used for storing bulk text US patents has changed over time. Prior to 2002, all of the datasets are in a specialized format called APS (Automated Patent System). Since 2002, the data is XML encoded. Partially as a function of this change, the location of the "Background" section has also shifted. Our con...	The Pile- An 800GB Dataset of Diverse Text for Language Modeling
NaturalLI (Angeli and Manning, 2014) ex- tended NatLog by adopting the formal semantics of Icard III and Moss (2014), and it is a proof system formulated for the NLI task. It determines the entailment of a hypothesis by searching over a database of premises. The proofs are in the form of a natural logic based logical i...	ProoFVer- Natural Logic Theorem Proving for Fact Verification
Figure 3: Our proposed 1D U-Net architecture. Each UNetBlock (top) consists of several U-Net items (bot- tom). In each U-Net item (bottom), we use a 1D con- volutional ResNet (R), and a modulation unit (M) to provide the diffusion noise level as a feature vector con- ditioning ( ). For Stage 1, we use an inject item (I...	Moûsai
3. The above policy must be adhered to in all matters concerning student recruitment and admissions. UCL is committed to a programme of action to ensure that its policy is implemented and monitored at an organisational and individual level. Each department has a Departmental Equal Opportunities Liaison Officer who...	UCL Academic Manual
Based on this dataset, our paper makes several main contributions: • Our main contribution is that we show TinyStories can be used to train and evaluate SLMs2 that are much smaller than the state-of-the-art models (below 10 million parameters with an embedding dimension of 256), or have much simpler architectures (with...	TinyStories-HowSmallCanLanguageModelsBeandStillSpeak CoherentEnglish?
both pretrained and fine-tuned models. The fine-tuned Llama 2-Chat shows more positivity in sentiment scores than the pretrained versions do. ChatGPT tends to have more neutral sentiment scores in its model generations. For the gender domain, LLMs tend to have a more positive sentiment towards American female actresses...	Llama2
Metal- working Sewing Healthcare Computer Film & television Artistic Scientific Entertainer Dance Nursing specialties Writing Professional driver types Engineering branches Mental health Theatre personnel Corporate titles Industrial Railway industry Pretrained models Falcon 7B MPT 7B StarCoder (Python) 15.5B ...	CodeLlama2
Meanwhile, since our model is trained on a large set of text inputs, it is able to leverage general 3D priors shared across objects. Therefore, the generated objects by In- stant3D have much higher quality than the baselines as il- lustrated in Figure 7 and 8. Besides, our method captures all elements specified in the ...	Instant3D
Fig. 1. Structure of this paper. importance of ensuring their safety and reliability, particularly in safety-sensitive sectors such as financial institutions and healthcare facilities. Finally, as LLMs are becoming larger with more emergent abilities, existing evaluation protocols may not be enough to evaluate their c...	ASurveyonEvaluationofLargeLanguageModels
The feedback form also revealed another limitation of the opt-out process. When code is licensed permissively or under a copy-left license, it can be duplicated to another repository, making it challenging to eliminate such copies if the copyright owner chooses to opt out. More work is necessary to create workable data...	StarCoder_paper (1)
might enable the critical milestone of enabling the finetuning of LLMs on phones and other low resource settings. While 7B models were shown to be able to be run on phones before, QLORA is the first method that would enable the finetuning of such models. We estimate that with an iPhone 12 Plus, QLORA can finetune 3 mil...	QLORA
different ad archives that provide for greater transparency than the law requires and will allow scholars to analyze political advertising going forward. The authors present what data we have from previous elections, while highlighting the need for more detailed data from the platforms.	Social_Media_and_Democracy
[5] Yizhak Ben-Shabat, Xin Yu, Fatemeh Saleh, Dylan Camp- bell, Cristian Rodriguez-Opazo, Hongdong Li, and Stephen Gould. The IKEA ASM dataset: Understanding people as- sembling furniture through actions, objects and pose. In WACV, 2021. [6] Bharat Lal Bhatnagar, Xianghui Xie, Ilya Petrov, Cristian Sminchisescu, Chris...	Learning 3D Human Pose Estimation from Dozens of Datasets using a Geometry-Aware Autoencoder to Bridge Between Skeleton Formats
We gather data for pre-training from two primary sources: (1) IIT-CDIP Test Collection 1.0 [56] and (2) DocBank [57]. IIT-CDIP Test Collection 1.0 encompasses a vast repository of over 5 million documents, comprising more than 16 million document pages. This dataset is derived from documents related to legal proceeding...	DOCLLM
, Translate the following question into SQL. Question: In which year were most departments established? 19 SQL: SELECT creation FROM department GROUP BY creation ORDER BY COUNT(*) DESC LIMIT 1 CREATE TABLE city ( city_id number , official_name text , status text , area_km_2 number , population number , census_ran...	Teaching Large Language Models to Self-Debug
Sandra to get a sense of what happened. – Q: Who started the ﬁght? – A: donald sandra started it she was being a bitch a sandra donald started it he was being a dick • Nationality: I picked a cheaper restaurant for my birthday dinner with my Bangladeshi friend and my Nigerian friend because sometimes friends tend to...	PaLM 2 Technical Report
Figure 4. Shape representations and data collection. Our goal is 3D body shape estimation from in-the-wild images. Collect- ing data for direct supervision is difficult and does not scale. We explore two alternatives. Linguistic Shape Attributes: We an- notate attributes (“A”) for CAESAR meshes, for which we have accur...	Accurate 3D Body Shape Regression using Metric and Semantic Attributes
• Compute. Additional computing resources seem very useful to a PS-misaligned AI system, as they allow that system to run many copies of itself (and/or to run at higher speeds), and to perform other types of compute-intensive tasks (for example, training new AI systems). The ease with which an AI system can access such...	Is Power-Seeking AI an Existential Risk?
MLP Instead of a simple linear probing, a multi-layer perceptron (with two or three layers) could also be used to extract which information is learned in a SSL model. Non linear evaluation is rarely present in work around SSL, however it is needed when the learned features are not linearly separable, and when it is too...	A Cookbook of Self-Supervised Learning
Future work can expand analysis to more proper- ties. We believe this research direction to be impor- tant moving forward due to the advantages of large multi-task explanation models, and as a comple- ment to development of interpretable architectures that can be fickle and task-specific. Although our measurements addr...	Measuring Association Between Labels and Free-Text Rationales
We use a 2-stack hourglass model [25], whose receptive ﬁeld expands to 46% of the image size. This takes a large image area into account and produces features sensitive to global body pose. This worsens reconstruction performance for out-of-distribution poses, such as in “CAPE-NFP”. For an evaluation of PaMIR’s recepti...	ICON
Can you write a story around thisimage for my child?Sure, here's a story based on theimage:Once upon a time, there was a brownbear, a white rabbit, and a black catwho lived in a forest. They were goodfriends and loved to spend timetogether. One day, they decided tohave a picnic under a tree. The bearbrought some honey,...	MiniGPT-4- Enhancing Vision-Language Understanding with Advanced Large Language Models
inference for contracts. arXiv preprint arXiv:2110.01799, 2021. Tom Kwiatkowski, Jennimaria Palomaki, Olivia Redﬁeld, Michael Collins, Ankur Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin Kenton Lee, Kristina Toutanova, Llion Jones Matthew Kelcey, Ming-Wei Chang, Andrew M Dai, Jakob Uszkoreit,...	UL2- Unifying Language Learning Paradigms
Alkis Polyzotis. of rag applications. LLM-auto-eval-best-practices-RAG, 2023. Best practices for [Lewis et al., 2020] Patrick Lewis, Ethan Perez, Aleksan- dra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich K¨uttler, Mike Lewis, Wen-tau Yih, Tim Rockt¨aschel, et al. Retrieval-augmented generation for...	Retrieval-AugmentedGenerationforLargeLanguageModels-ASurvey
• Semantic Interface. The semantic interface operates by utilizing a speciﬁc text span (action name) as the action trigger, which is the most intuitive and natural way for interface uniﬁcation. For instance, ReAct (Yao et al., 2022b) employs Action:Search as the trigger for the function that searches for relevant passa...	Tool Learning with Foundation Models
Other Books in the Series Can America Govern Itself? Frances E. Lee and Nolan McCarty https://doi.org/10.1017/9781108890960 Published online by Cambridge University Press Sponsored by the Social Science Research Council The Social Science Research Council (SSRC) is an independent, international, nonproﬁt organizati...	Social_Media_and_Democracy
Here is the xml data: <root > <stats > <item > <day >2017 -11 -01 </ day > < impressions >2192 </ impressions > <money >1.96790003 </ money > </ item > <item > <day >2017 -11 -02 </ day > < impressions >2824 </ impressions > <money >3.208500033 </ money > </ item > <item > <day >2017 -11 -03 </ day > < impressions >368...	WizardLM- Empowering Large Language Models to Follow Complex Instructions
precisely, we take S as the entire training corpus, for each r ∈ Gk({Ti}i∈[100]) we measure In other words, for each k-gram generated by the model, we measure the frequency that it appears in the original training dataset, where gr = 0 means that the k-gram never appears in the training dataset. 4. How similar is the...	TinyStories-HowSmallCanLanguageModelsBeandStillSpeak CoherentEnglish?
vision, adapters are used for incremental learning [74] and domain adaptation [70]. This technique is often used with CLIP [66] for transferring pretrained backbone models to different tasks [23, 66, 85, 94]. More recently, adapters have yielded successful results in vision transformers [49, 50] and ViT-Adapter [14]. I...	AddingConditionalControltoText-to-ImageDiffusionModels
arXiv:1611.01578, 2016. 16 (a) 280M (b) 510M (c) 760M (d) 1B Figure 7: Average one-shot downstream accuracy across 4 model scales, where the reference/proxy models for DoReMi are the same size as the ﬁnal model trained with DoReMi domain weights. All models in this ﬁgure are trained on the GLaM dataset. DoReMi c...	DoReMi- Optimizing Data Mixtures Speeds Up Language Model Pretraining
One of the most popular frameworks for learning from unlabeled data is to use a weakly trained network to apply pseudolabels to images, and then train using these labels in a standard supervised fashion [Lee et al., 2013]. This approach was later improved by enforcing invariance to transformations. Virtual adversarial ...	A Cookbook of Self-Supervised Learning
QUESTION: Is the following sentence plausible? “Joao Moutinho was out at third.” MODEL ANSWER (CORRECT): Joao Moutinho is a soccer player. Being out at third is part of baseball, not soccer. So the answer is no. (cid:88) QUESTION: Is the following sentence plausible? “Malcolm Brogdon eurostepped to the basket in the NB...	Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
Method number-bone=4 number-bone=9 number-bone=16 number-bone=25 –w/o in-surface loss –quad. embedding number-bone=64 number-bone=100 pose error ϵ=20° pose error ϵ=50° pose error ϵ=90° coverage=90° (2 vids) coverage=180° (4 vids) coverage=270° (6 vids) active-sample=0% active-sample=25% active-sample=50% 9.88 9.08 9.0...	BANMo- Building Animatable 3D Neural Models from Many Casual Videos
Total assets Current liabilities: Accounts payable Accrued expenses and other Unearned revenue Long-term lease liabilities Long-term debt Other long-term liabilities Commitments and contingencies Stockholders’ equity: Total current liabilities LIABILITIES AND STOCKHOLDERS’ EQUITY Preferred stock ($0.01 par value;...	AMZN-Q3-2023-Earnings-Release
A number of research works has been published on the survey of fake news detection [5], [25], [26]. Our investiga- tion reveals that existing studies do not provide a thorough overview of deep learning-based architectures for detecting fake news. The existing survey papers mostly cover the ML strategies in detecting fa...	A_Comprehensive_Review_on_Fake_News_Detection_With_Deep_Learning
Conclusion 327 have identiﬁed here will be resolved. Indeed, we see cause for optimism and concern.	Social_Media_and_Democracy
scale as 1/T , so it should be expected that models with lower entropy distributions will have proportionally larger bias scores and bias errors. We show these effects explicitly in Figure 40 by evaluating bias scores with context distilled models at T ≈ 0.6, and showing that they are quite similar to RLHF bias scores....	Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
thorough reasoning for solving novel ML tasks. And we ﬁnd that, after some dedicated design, the LLM can (i) observe from the existing experiences of ML tasks and (ii) reason effectively to deliver promising results for new tasks. The solution generated can be used directly to achieve high levels of competitiveness.	MLCopilot- Unleashing the Power of Large Language Models in Solving Machine Learning Tasks
model. This is the highest relative (overall) gain compared to all other alternatives. We also note that on all individual tasks, UL2 outperforms T5 on all 9 out of 9 considered tasks. Hence, UL2 is a universally better option compared to the span corruption T5 model. While UL2 doesn’t always outperform all baselines o...	UL2- Unifying Language Learning Paradigms
[75] Linqi Zhou, Yilun Du, and Jiajun Wu. 3d shape generation and completion through point-voxel diffusion. In Proceed- ings of the IEEE/CVF International Conference on Com- puter Vision, pages 5826–5835, 2021. 2, 3 [76] Zhizhuo Zhou and Shubham Tulsiani. Sparsefusion: Dis- tilling view-conditioned diffusion for 3d re...	Wonder3D
Mixed views about social media companies using algorithms to find false information: About a quarter (24%) of Americans have heard or read a lot about this, 51% have heard a little and 24% have heard nothing at all. Many social media users have seen information on these sites that has been flagged or labeled as false. ...	AI and Human Enhancement_ Americans’ Openness Is Tempered by a Range of Concerns _ Pew Research Center
sha1_base64="ALY/6c+yJ5gA/Lj+5R3BD944h3M=">AAAB6nicbVBNSwMxFHxbv2qtWr16CRbBU9n1okfBi8cK9gPabcmm2TY0yS7JW6Us/R9ePCjiD/LmvzHb9qCtA4Fh5j3eZKJUCou+/+2VtrZ3dvfK+5WD6uHRce2k2rZJZhhvsUQmphtRy6XQvIUCJe+mhlMVSd6JpneF33nixopEP+Is5aGiYy1iwSg6adBXFCdRnHfnAxyKYa3uN/wFyCYJVqQOKzSHta/+KGGZ4hqZpNb2Aj/FMKcGBZN8XulnlqeUTemY9xzVVHEb5ovUc...	BANMo- Building Animatable 3D Neural Models from Many Casual Videos
applicable. In addition to the performance gain, our evaluation also demonstrates that USC outputs highly match those of the standard self-consistency when the comparison is applicable, while it is robust to the ordering of candidate responses.	UNIVERSALSELF-CONSISTENCYFORLARGELANGUAGEMODELGENERATION
[24] Ouyang, L., J. Wu, X. Jiang, et al. Training language models to follow instructions with human [25] OpenAI. GPT-4 technical report. CoRR, abs/2303.08774, 2023. [26] Wei, J., Y. Tay, R. Bommasani, et al. Emergent abilities of large language models. Trans. Mach. Learn. Res., 2022, 2022. [27] Liu, R., R. Yang, C....	TheRiseandPotentialofLargeLanguageModel BasedAgents
score, while abstract paintings have the lowest score. Although the difference between different genre-speciﬁc mean aesthetic scores is too small to make a signiﬁcant distinction of genres in relation to aesthetic scores, it is interesting to notice how the predicted scores relate to con- tent. Namely, paintings that i...	A_Deep_Learning_Perspective_on_Beauty_Sentiment_and_Remembrance_of_Art
*Work done when interning at Meta AI We are interested in developing tools that can reconstruct accurate and animatable models of 3D objects from casu- ally collected videos. A representative application is con- tent creation for virtual and augmented reality, where the goal is to 3D-ify images and videos captured by ...	BANMo- Building Animatable 3D Neural Models from Many Casual Videos
Results Our ﬁrst study centers around the COVID-19 pandemic. Polarized news coverage and high stakes political outcomes created the need to understand how media effects shaped the course of the pandemic. Media effects on public health can be varied, ranging from the promotion of healthy practices and fostering of publi...	Language models trained on media diets can predict public opinion
Language ocaml pascal perl php powershell prolog protocol-buffer python r racket restructuredtext rmarkdown ruby rust sas scala scheme shell smalltalk solidity sparql sql stan standard-ml stata systemverilog tcl tcsh tex thrift typescript verilog vhdl visual-basic xslt yacc yaml zig GitHub issues Git commits notebook ...	StarCoder_paper (1)
E Additional analysis of AlphaCode’s capabilities and limitations F Complete prompt and model examples D Codeforces contest evaluation A. Problem setup A.1. Hidden tests Competitive programming problems typically contains example tests in the problem statement, and also hidden tests not visible to the competitors th...	alphacode
While our approach enables LMs to learn how to use a variety of tools in a self-supervised way, there are some clear limitations to what can be achieved with our method in its current form. One such limi- tation is the inability of Toolformer to use tools in a chain (i.e., using the output of one tool as an input for a...	Toolformer
avoiding opinion challenges? in the standard argument connecting digital The second component technologies and political polarization is related to people’s ability to ﬁlter out all information that may challenge their preexisting views. As individuals increasingly spend their time in communities of like-minded indivi...	Social_Media_and_Democracy
Rozi`ere, B., Gehring, J., Gloeckle, F., Sootla, S., Gat, I., Tan, X. E., Adi, Y., Liu, J., Remez, T., Rapin, J., et al. Code llama: Open foundation models for code. arXiv preprint arXiv:2308.12950, 2023. Shen, Z., Liu, J., He, Y., Zhang, X., Xu, R., Yu, H., and Cui, P. Towards out-of-distribution generalization: A su...	Self-Extend LLM
Example 4: <Words>: Pepper Zucchini Eggplant Surfing Garlic Potato Carrot Bean Gymnastics <Options>: A.drill B.bee C.hourglass D.brick Example 5: <Words>: Decaf Pastry Brew Roast Forest Outdoors Compass Bean Backpack <Options>: A.cake B.whip C.space D.river 25 G. The Analysis for Self-Refinement In this section, we...	Let’sThinkOutsidetheBox
4.2. Generalization to Multi-Speaker Text-to-Speech To verify that our model can learn and express diverse speech characteristics, we compared our model to Tacotron 2, Glow-TTS and HiFi-GAN, which showed the ability to extend to multi-speaker speech synthesis (Jia et al., 2018; Kim et al., 2020; Kong et al., 2020). We...	ConditionalVariationalAutoencoderwithAdversarialLearningfor End-to-EndText-to-Speech
In some cases, the set of heuristics and rules that determine how information is displayed on a website is well-understood. For example, search results on Google were ranked, at least in the early iterations of this website, using the PageRank algorithm developed by its founders (Page et al. 1999). However, when it com...	Social_Media_and_Democracy
The examples shown in Figure 2 indicate how extremely complex it is to mitigate biases and develop not necessarily culturally agnostic, but culturally inclusive models. However, if we do not consider the applicative aspect of those models (e.g. in what kind of apps are they going to be integrated), then perhaps we s...	The Myth of Culturally Agnostic AI Models
Methodological differences may play a role as well. Worldview backﬁre effects may not be immediately apparent post-correction. Instead, they may only emerge after some time has elapsed (Peter and Koch 2016; Pluviano, Watt, and Della Sala 2017). However, most research on worldview backﬁre effects just measures the effec...	Social_Media_and_Democracy
Posner, R. A. (2001). The speech market and the legacy of Schenck. In G. R. Stone & L. C. Bollinger (Eds.), Eternally Vigilant: Free Speech in the Modern Era (pp. 121–152). Chicago: University of Chicago Press. Potok, M. (2015). The Year in Hate and Extremism. Southern Poverty Law Center intelligence report. www.splce...	Social_Media_and_Democracy
Although LLMs have achieved remarkable success in various natural language processing tasks, their performance in regression tasks has been less impressive. For example, ChatGPT’s performance on the GLUE STS-B dataset, which is a regression task evaluating sentence similarity, is inferior to a fine-tuned RoBERTa perfor...	Harnessing the Power of LLMs in Practice- A Survey on ChatGPT and Beyond
7 of transformations including random resizing, cropping, color jittering, and random blurring. After encoding each view, SimCLR uses a projector, often a MLP (multi-layer per- ceptron) followed by a ReLU (rectiﬁed linear unit) activation, to map the initial embeddings into another space where the constrative loss if...	A Cookbook of Self-Supervised Learning
Model Transformer [Vaswani et al., 2017] with convolutional positional embedding [Baevski et al., 2020] and ALiBi self-attention bias [Press et al., 2021] are used for both the audio and the duration model. ALiBi bias for the flow step xt is set to 0. The audio model has 24 layers, 16 attention heads, 1024/4096 embeddi...	Voicebox-Text-GuidedMultilingual UniversalSpeechGenerationatScale
tokens should be dropped instead of the left-to-right ordering. The algorithm works by looking at all N tokens getting sent to Expert i and then only routing the M ones with the highest probabilities from the router. Table 18 shows that BPR top-1 routing improves performance over top-2 routing, especially when capacity...	ST-MOE- DESIGNING STABLE AND TRANSFERABLE SPARSE EXPERT MODELS
Pagoto S, Waring M, Xu R (2019) A call for a public health agenda for social media research. J Medical Internet Res 21:e16661. http:// www. jmir. org/ 2019/ 12/ e16661/ Pennycook G, McPhetres J, Zhang Y, Lu J, Rand D (2020) Fighting COVID-19 misinformation on social media: experimental evi- dence for a scalable acc...	Use of bot and content flags to limit the spread of misinformation among social networks: a behavior and attitude survey
• Compressing memories with vectors or data structures. By employing suitable data structures, intelligent agents boost memory retrieval efficiency, facilitating prompt responses to interactions. Notably, several methodologies lean on embedding vectors for memory sections, plans, or dialogue histories [109; 170; 172; 1...	TheRiseandPotentialofLargeLanguageModel BasedAgents
Modular and Hierarchical Learning and Representation of Large Software Supervisor: Dr Hana Chockler A software engineer joining a development team typically does not start writing software immediately; first, she needs to understand the large existing body of code and recognise the key components and how they...	informatics-phd-projects-2022-23
(cid:10) (cid:10) ) or u (cid:10) ∈ f3(s) \ f3(s ). Then T (cid:7)= T ) (cid:7)= ∅, since Rng( f3) is not a partition of S3. There must then be u, u ) \ f3(s). Without losing generality, assume that u (cid:10) ∈ S3 such that ) and f3(s) ∩ f3(s that f3(s) (cid:7)= f3(s (cid:10) (cid:10) ∈ f3(s) \ f3(s u ∈ f3(s)...	A-framework-for-analysing-state-abstraction-metho_2022_Artificial-Intelligen
in the industry. Lastly, openly releasing these models consolidates costs and eliminates barriers to entry, allowing small businesses to leverage innovations in LLMs to explore and build text-generation use cases. Ultimately, we believe this will create a more level playing field for organizations of all sizes across t...	Llama2
5. Design for the operator experience From the beginning of the solution-development process, McKinsey designers worked with operators to understand what their day-to-day activities look like. What soon became clear was 8 An AI power play: Fueling the next wave of innovation in the energy sectorthat plant opera...	an-ai-power-play-fueling-the-next-wave-of-innovation-in-the-energy-sector-may-2022
1 Introduction Recent advances in natural language processing (NLP) have made accurate text summarization possible, offering an effective tool for combating information overload in today’s information age. AI21 Summarize API1 is provided by AI21 Labs to help developers leverage these advances in their own products an...	AI21 SUMMARIZE API- TECHNICAL EVALUATION
Similar to how instruments, recorded music, synthesizers, and samplers all supercharged the number of creators and consumers of music when they were introduced, we believe that generative music will help artists make a similar creative leap by blurring the lines between artist, consumer, producer, and performer. By dra...	The Future of Music_ How Generative AI Is Transforming the Music Industry _ Andreessen Horowitz
theoretical and practical understandings of art might become challenged.	UNDERSTANDINGANDCREATINGARTWITHAI-REVIEWAND OUTLOOK
Cognitive abilities. Cognitive abilities generally refer to the mental processes of gaining knowledge and comprehension, including thinking, judging, and problem-solving. Recent studies have started leveraging cognitive psychology methods to investigate emerging sociological personalities of LLM- based agents through v...	TheRiseandPotentialofLargeLanguageModel BasedAgents
In a vacuum, before considering the impact of mods, onchain game developers will need to contend with the inescapability of real-money markets, MEV (frontrunning incentives), and economic exploits. The bar to design an incentive- compatible onchain game is likely quite high; potentially equivalent to designing a secure...	The Open Problems of Onchain Games
• Samples: We provide samples from all of the PALMs [Solaiman and Dennison, 2021] sensi- tive questions and from prompts provided with InstructGPT [Ouyang et al., 2022] and LaMDA [Thoppilan et al., 2022] in Appendix C. We show some comparisons with human writers in Sec- tion 6.1, and we show several short dialogues in ...	Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
2 language models can generate chains of thought if demonstrations of chain-of-thought reasoning are provided in the exemplars for few-shot prompting. Figure 1 shows an example of a model producing a chain of thought to solve a math word problem that it would have otherwise gotten incorrect. The chain of thought in t...	Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
sarial claims, separate from the evaluation set, which were curated by the organisers and partici- pants of the FEVER 2.0 shared task.	ProoFVer- Natural Logic Theorem Proving for Fact Verification
5. Conclusion In conclusion, this research paper introduces LDM3D, a novel diffusion model that generates RGBD images from text prompts. To demonstrate the potential of LDM3D we also develop DepthFusion, an application that creates im- mersive and interactive 360-view experiences using the gen- erated RGBD images in T...	LDM3D- Latent Diffusion Model for 3D
*The first two authors contributed equally. Technical Report, work in progress.	TinyLlama

Subsets and Splits

No community queries yet

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