text stringlengths 1 1k ⌀ | title stringclasses 230
values |
|---|---|
datasets: CIFAR10, ImageNet32, and ImageNet64. Results are shown in Table 4. First, compared
to 4 baselines (i.e., IDF, IDF++ (van den Berg et al., 2020), Glow (Kingma & Dhariwal, 2018), and
RealNVP (Dinh et al., 2016)), PC+IDF achieved the best bpd on ImageNet32 and ImageNet64. Next,
PC+IDF improved over its base mode... | LOSSLESS COMPRESSION WITH PROBABILISTIC CIRCUITS |
model-training projects costing many millions of dollars,
with reasonable confidence that these projects will succeed | Eight Things to Know about Large Language Models |
References
Alcorn, M. A., Li, Q., Gong, Z., Wang, C., Mai, L., Ku, W.-
S., and Nguyen, A. Strike (with) a pose: Neural networks
are easily fooled by strange poses of familiar objects. In
Proceedings of the IEEE/CVF Conference on Computer
Vision and Pattern Recognition, pp. 4845–4854, 2019.
Amodei, D., Anubhai, R., Bat... | RobustSpeechRecognitionviaLarge-ScaleWeakSupervision |
potential governance mechanism that could rein in platform companies (Brock
2017). Yet how has transparency historically been enacted by these companies,
and what are the recent measures that have been implemented in response to
this public outcry? | Social_Media_and_Democracy |
abs/2001.08361, 2020.
[140] Roberts, A., C. Raffel, N. Shazeer. How much knowledge can you pack into the parameters
of a language model? In B. Webber, T. Cohn, Y. He, Y. Liu, eds., Proceedings of the 2020
Conference on Empirical Methods in Natural Language Processing, EMNLP 2020, Online,
November 16-20, 2020, pages 54... | TheRiseandPotentialofLargeLanguageModel BasedAgents |
22
The Efficiency Spectrum of Large Language Models: An Algorithmic Survey
Efficient LLM Algorithmic Survey, Nov, 2023, USA. | TheEfficiencySpectrumofLargeLanguageModels-AnAlgorithmicSurvey |
Tim and Jen are playing with their toys in the living room. Tim has a laser that makes red lights
and sounds. Jen has a doll that she can dress and comb. Tim likes to make his laser point at Jen’s
doll and make noises. Jen does not like that. She tells Tim to stop.
”Stop, Tim! You are hurting my doll! She does not like... | TinyStories-HowSmallCanLanguageModelsBeandStillSpeak CoherentEnglish? |
https://www.nfx.com/post/generative-ai-hot-75-list
1/4
· · 11/05/2023, 04:43
The AI Hot 75
After we open-sourced our Generative AI market map a few
months ago, we went to work analyzing this network for early
indicators of future greatness. It is going to produce the next
generation of unicorns, at a faster rate... | The AI Hot 75 |
• All model sizes are now trained with uniform batch size of 2M tokens. Previously, the models of size 160M, 410M, and
1.4B parameters were trained with batch sizes of 4M tokens, but in the course of training the initial suite we discovered
that it was feasible to train all models with uniform batch size, though based ... | Pythia- A Suite for Analyzing Large Language Models Across Training and Scaling |
Other general training methods such as regularization [82, 95, 132] and loss reconstruction [107,
193, 199] have also been proposed to tackle the hallucination problem.
5.2.3 Post-Processing. Post-processing methods can correct hallucinations in the output, and this
standalone task requires less training data. Especia... | SurveyofHallucinationinNatural Language Generation |
Output in Neural and Statistical Machine Translation. In MTSummit.
[125] Joshua Maynez, Shashi Narayan, Bernd Bohnet, and Ryan McDonald. 2020. On Faithfulness and Factuality in
Abstractive Summarization. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.
1906–1919.
[126] Pierr... | SurveyofHallucinationinNatural Language Generation |
by adapting BERT to a week of coronavirus-related news articles from a week before the survey. Except where noted in certain
sub-analyses, the results are using online news -based models used to predict survey response proportions. | Language models trained on media diets can predict public opinion |
Public opinion – sometimes referred to as a ‘thermostat’ of public will10, 11 – is commonly measured through surveys by
governments, companies, NGOs, and political parties and candidates for office.12 The understanding gained through surveys is
critical inputs to decision-making around economic strategy and public healt... | Language models trained on media diets can predict public opinion |
Classifier guidance (CG) [Dhariwal and Nichol, 2021] is a technique used to trade off mode coverage
and sample fidelity for diffusion models post training, similar to the effect of truncated or low-
temperature sampling for generative adversarial networks [Brock et al., 2018] and discrete flow
models [Kingma and Dhariw... | Voicebox-Text-GuidedMultilingual UniversalSpeechGenerationatScale |
Autoregressive Infilling. There are two main autoregressive infilling approaches: “fill-in-the-middle” (FIM) where a
single span is sampled, and “blank infilling” with multiple spans.
The OpenAI FIM approach [36] uses the template (prefix, middle, suffix) to divide a document into three
segments. Next, these segments a... | DOCLLM |
Rashid, A. Rula, L. Schmelzeisen, J. Sequeda, S. Staab, A. Zimmermann, Knowledge graphs, 2020, arXiv:2003 .02320.
2016, pp. 305–324.
007, http://www.sciencedirect .com /science /article /pii /S0004370218305988.
gation, vol. 1, Longmans, Green, and Company, 1884.
19
I. Tiddi and S. Schlobach
Artificial Intelligence... | Knowledge graphs as tools for explainable machine learning: A survey |
Ariel N Lee, Cole J Hunter, and Nataniel Ruiz. 2023b.
Platypus: Quick, cheap, and powerful refinement of
llms. arXiv preprint arXiv:2308.07317.
Katherine Lee, Daphne Ippolito, Andrew Nystrom,
Chiyuan Zhang, Douglas Eck, Chris Callison-Burch,
and Nicholas Carlini. 2021. Deduplicating training
data makes language models... | DataManagementForLargeLanguageModels-ASurvey |
[12] Iro Laina, Christian Rupprecht, Vasileios Belagiannis, Fed-
erico Tombari, and Nassir Navab. Deeper depth prediction
with fully convolutional residual networks. In 2016 Fourth
International Conference on 3D Vision (3DV), pages 239–
248, 2016. 2
[13] Tsung-Yi Lin, Michael Maire, Serge Belongie, Lubomir
Bourdev, Ro... | LDM3D- Latent Diffusion Model for 3D |
• Subqueries:Various query strategies can be employed in
different scenarios, including using query engines pro-
vided by frameworks like LlamaIndex, employing tree
queries, utilizing vector queries, or employing the most
basic sequential querying of chunks.
• HyDE: This approach is grounded on the assumption
that the... | Retrieval-AugmentedGenerationforLargeLanguageModels-ASurvey |
ZENY: But shouldn’t we incentivize slow science?
Many NLP papers neglect related work and keep
reinventing the wheel. We need deeper analysis to
enable disruptive scholarship and novel ideas.20
SOCART: Slow science also has disadvantages. Fast
turn-around has had many positive effects on NLP,
including rapid replicatio... | A Two-Sided Discussion of Preregistration of NLP Research |
Germany (Muller and Schwarz 2017). Journalistic accounts of religious violence
in Sri Lanka and Myanmar have shown how these events appear to be fueled by
rumors spread on Facebook and WhatsApp (Taub and Fisher 2018). These two
instances illustrate the urgent need for research on how the spread of extremist
ideas on so... | Social_Media_and_Democracy |
2.2.4 Social Bias
Besides the marginalization of minority groups
caused by data detoxifying, several works (Kurita
et al., 2019; Nangia et al., 2020; Meade et al., 2022;
Feng et al., 2023) find that pre-trained LLMs can
capture social biases contained in the large amounts
of training text. Evaluating the C4 (Raffel et ... | DataManagementForLargeLanguageModels-ASurvey |
4.2.2 No use case. Fine-tuned models, such as DeltaLM+Zcode [118], still perform best on most rich-resource translation
and extremely low-resource translation tasks. In rich resource machine translation, fine-tuned models slightly outperform
LLMs [22, 92]. And in extremely low-resource machine translation, such as Engl... | Harnessing the Power of LLMs in Practice- A Survey on ChatGPT and Beyond |
of larger models, and proposes to fix this problem by keeping those dimensions in higher preci-
sion. Lastly, nuQmm develops efficient GPU kernels for a specific binary-coding based quantization
scheme.
Relative to this line of work, we show that a significantly more complex and accurate quantizer can
be implemented efficie... | GPTQ |
[673] Donghui Zhu and Ning Chen. 2022. Multi-Source Domain Adaptation and Fusion for Speaker Verification. IEEE/ACM
Transactions on Audio, Speech, and Language Processing 30 (2022), 2103–2116.
[674] Hao Zhu, Huaibo Huang, Yi Li, Aihua Zheng, and Ran He. 2018. Arbitrary talking face generation via attentional
audio-v... | AReviewofDeepLearningTechniquesforSpeechProcessing |
2.5. Joint Embedding Models for Music and Text
MuLan (Huang et al., 2022) is a music-text joint embedding
model consisting of two embedding towers, one for each
modality. The towers map the two modalities to a shared
embedding space of 128 dimensions using contrastive learn-
ing, with a setup similar to (Radford et al... | MusicLM |
7
which naturally have a long context, such as abstractive summarization or multi-document QA. The
following benchmarks have been proposed for long-context evaluation of LLMs:
• SCROLLS (Shaham et al., 2022) is a popular evaluation benchmark made of 7 datasets with
naturally long input. The tasks cover summarization... | ChatGPT’sOne-yearAnniversary-AreOpen-Source LargeLanguageModelsCatchingup |
References
[1] Harsh Agrawal, Peter Anderson, Karan Desai, Yufei Wang,
Xinlei Chen, Rishabh Jain, Mark Johnson, Dhruv Batra, Devi
Parikh, and Stefan Lee. nocaps: novel object captioning
at scale. In 2019 IEEE/CVF International Conference on
Computer Vision, ICCV 2019, Seoul, Korea (South), October
27 - November 2, 2019... | REVEAL-Retrieval-AugmentedVisual-LanguagePre-Trainingwith Multi-SourceMultimodalKnowledgeMemory |
36
Core Contributors
Paige Bailey
Kefan Xiao
Nimesh Ghelani
Lora Aroyo
Ambrose Slone
Neil Houlsby
Xuehan Xiong
Zhen Yang
Elena Gribovskaya
Jonas Adler
Mateo Wirth
Lisa Lee
Music Li
Thais Kagohara
Jay Pavagadhi
Sophie Bridgers
Anna Bortsova
Sanjay Ghemawat
Zafarali Ahmed
Tianqi Liu
Richard Powell
Vijay Bolina
Mariko I... | gemini_1_report |
Figure 1: Overview of our approach. We combine a pre-trained retriever (Query Encoder + Document
Index) with a pre-trained seq2seq model (Generator) and fine-tune end-to-end. For query x, we use
Maximum Inner Product Search (MIPS) to find the top-K documents zi. For final prediction y, we
treat z as a latent variable and ... | Retrieval-AugmentedGenerationfor Knowledge-IntensiveNLPTasks |
scalable off-policy reinforcement learning. arXiv preprint arXiv:1910.00177, 2019.
[29] J. Peters and S. Schaal. Reinforcement learning by reward-weighted regression for operational
space control. In Proceedings of the 24th international conference on Machine learning, pages
745–750, 2007.
[30] R. L. Plackett. The an... | Direct Preference Optimization |
pledge to respond to a user’s question in the future, prematurely try to end the conversation, or make up incorrect details
about the user.
We have shown that fine-tuning can improve safety metrics on average by defining safety objectives (Appendix A.1) for
our safety fine-tuning, which we used to annotate candidate respo... | LaMDA- Language Models for Dialog Applications |
(June 2016). URL: https://arxiv.org/abs/1606.06565.
Bowen Baker et al. “Emergent Tool Use From Multi-Agent Autocurricula”. en. In:
Eighth International Conference on Learning Representations. Apr. 2020. URL: https:
//iclr.cc/virtual_2020/poster_SkxpxJBKwS.html (visited on 04/29/2022).
Christopher Berner et al. “Dota 2 ... | Is Power-Seeking AI an Existential Risk? |
[168] Zhou, W., Xu, C., Ge, T., McAuley, J., Xu, K., Wei, F.: Bert loses patience: Fast
and robust inference with early exit. Advances in Neural Information Processing
Systems 33, 18330–18341 (2020)
[169] Zhang, Z., Zhu, W., Zhang, J., Wang, P., Jin, R., Chung, T.-S.: Pcee-bert:
Accelerating bert inference via patient... | Beyond Efficiency |
yc = F(x; Θ) + Z(F(x + Z(c; Θz1); Θc); Θz2)
where yc becomes the output of this neural network block, as visualized in Fig. 2-(b).
Because both the weight and bias of a zero convolution layer are initialized as zeros, in the first
training step, we haveZ(c; Θz1) = 0
F(x + Z(c; Θz1); Θc) = F(x; Θc) = F(x; Θ)
Z(F(x ... | Adding Conditional Control to Text-to-Image Diffusion Models |
Mixed views about a future with widespread use of robotic exoskeletons to increase
strength for manual labor jobs: Americans anticipate both benefits and downsides for
workers from the possibility of widespread use of robotic exoskeletons with a built-in AI
system to increase strength for manual labor jobs such as manu... | AI and Human Enhancement_ Americans’ Openness Is Tempered by a Range of Concerns _ Pew Research Center |
A. Scene Initialization
Content Generation. To obtain the initial scene content with
respect to the input prompt p, we first employ a pre-trained
diffusion model fd conditioned on p to generate a 2D scene
image I0 = fd((cid:15) | p), where (cid:15) is a random Gaussian noise.
Due to the lack of geometric information in ... | Text2NeRF- Text-Driven 3D Scene Generation with Neural Radiance Fields |
Vistra team members provided continuous guidance
about the intricacies of how the plant worked, and
identified critical data sources from sensors, which
helped McKinsey engineers refine the model,
adding and removing variables to see how those
changes affected the heat rate.
Through this training process, and by... | an-ai-power-play-fueling-the-next-wave-of-innovation-in-the-energy-sector-may-2022 |
Octavian-Eugen Ganea and Thomas Hofmann. 2017.
Deep joint entity disambiguation with local neural
attention. In Proceedings of the 2017 Conference on
Empirical Methods in Natural Language Processing,
pages 2619–2629, Copenhagen, Denmark. Associa-
tion for Computational Linguistics.
Kelvin Guu, Kenton Lee, Zora Tung, P... | Entities as Experts- Sparse Memory Access with Entity Supervision |
C Evaluation
As we scaled up the project, we began having to collect labels on multiple
solutions for the same training problem. In order to avoid the risk of over-fitting
on the 7,500 MATH training problems, we expanded the training set to include
4,500 MATH test split problems. We therefore evaluate our models only ... | Let’s Verify Step by Step |
#Parameters
46,874,690
27,225,105
13,088,268
0.15s∗
0.25s†
0.29s†
proposed in OccNet [51] to reduce network querying times.
Overall, taking the body reference optimization step into
account, it takes about 50s to reconstruct the geometry of
the 3D human model and 1s to recover its surface color. | PaMIR- Parametric Model-Conditioned Implicit Representation for Image-based Human Reconstruction |
The attention mechanism is a powerful method for obtaining a more discriminative utterance-
level feature by explicitly selecting frame-level representations that better represent speaker char-
acteristics. Recently, the Transformer model with a self-attention mechanism has become effective
in various application field... | AReviewofDeepLearningTechniquesforSpeechProcessing |
Let’s Think Outside the Box: Exploring Leap-of-Thought in Large
Language Models with Creative Humor Generation
Shanshan Zhong1,2* Zhongzhan Huang1,2∗
Shanghua Gao3 Wushao Wen2
Liang Lin2
Marinka Zitnik3
Pan Zhou1†
1Sea AI Lab 2Sun Yat-sen University
3Harvard University
∗Co-first author: {zhongshsh5,huangzhzh23... | Let’sThinkOutsidetheBox |
Meihua Dang, Antonio Vergari, and Guy Broeck. Strudel: Learning structured-decomposable
probabilistic circuits. In International Conference on Probabilistic Graphical Models, pp. 137–148.
PMLR, 2020.
Meihua Dang, Pasha Khosravi, Yitao Liang, Antonio Vergari, and Guy Van den Broeck. Juice: A
julia package for logic and... | LOSSLESS COMPRESSION WITH PROBABILISTIC CIRCUITS |
WhoTargetsMe, as well as researchers at the University of Wisconsin, who
found that the majority of issue ads they studied in the lead-up to the 2016 US
election did not originate from organizations registered with the Federal | Social_Media_and_Democracy |
Hybrid Approaches (RAG+FT). Combining RAG with
fine-tuning is emerging as a leading strategy. Determining the
optimal integration of RAG and fine-tuning whether sequen-
tial, alternating, or through end-to-end joint training—and
how to harness both parameterized and non-parameterized
advantages are areas ripe for explo... | RAG forLargeLanguageModels-ASurvey |
The first line of each test case contains a single
integer n (1 <= n <= 100) - the length of the
sequence .
The second line of each test case contains n integers
a_1 , a_2 , ... , a_n (0 <= a_i <= 10^9) .
For each test case , print one integer - the minimal
value of the maximum value in the sequence .
Output
Output... | alphacode |
transformers. CoRR, abs/2305.16300, 2023.
[234] Chalkidis, I., X. Dai, M. Fergadiotis, et al. An exploration of hierarchical attention transformers
for efficient long document classification. CoRR, abs/2210.05529, 2022.
[235] Nie, Y., H. Huang, W. Wei, et al. Capturing global structural information in long document
... | TheRiseandPotentialofLargeLanguageModel BasedAgents |
Jonathan Ho, Evan Lohn, and Pieter Abbeel. Compression with flows via local bits-back coding. In
Proceedings of the 33rd International Conference on Neural Information Processing Systems, pp.
3879–3888, 2019.
Emiel Hoogeboom, Jorn Peters, Rianne van den Berg, and Max Welling. Integer discrete flows and
lossless compress... | LOSSLESS COMPRESSION WITH PROBABILISTIC CIRCUITS |
[6] David Chen and William B Dolan. Collecting highly paral-
lel data for paraphrase evaluation. In ACL, pages 190–200,
2011. 2, 6
[7] Haoxin Chen, Menghan Xia, Yingqing He, Yong Zhang,
Xiaodong Cun, Shaoshu Yang, Jinbo Xing, Yaofang Liu,
Qifeng Chen, Xintao Wang, Chao Weng, and Ying Shan.
Videocrafter1: Open diffusio... | GPT4Video |
updated after only a little more than a decade [62, 63] to cover novel aspects of technology readiness. Likely
SHAPE might need to be revised when augmentation technologies are more broadly used. This limitation also
points to the research opportunity to investigate with the SHAPE scale how attitudes evolve and change ... | Society’sAttitudesTowardsHumanAugmentation |
3
2
0
2
t
c
O
3
]
L
C
.
s
c
[
1
v
8
9
7
1
0
.
0
1
3
2
:
v
i
X
r
a
LARGE LANGUAGE MODELS CANNOT SELF-CORRECT
REASONING YET
Jie Huang1,2,∗ Xinyun Chen1 Swaroop Mishra1 Huaixiu Steven Zheng1 Adams Wei Yu1
Xinying Song1 Denny Zhou1
1Google DeepMind
2University of Illinois at Urbana-Champaign
ABSTRACT | LARGELANGUAGEMODELSCANNOTSELF-CORRECT REASONINGYET |
Computer programs are, of course, built on the same bedrock; algorithms are largely
specified in terms of operations that are performed over variables. Variables get bound
to instances, algorithms get called, operations are performed, and values are returned. | The Next Decade in AI- |
We present the results of an MQM study for Chinese-to-English and English-to-German in Table 9. MQM represents
the average errors per segment, with lower numbers indicating better results. We observe that PaLM 2 improves quality
both over PaLM and Google Translate.
9We used BLEURT version 0p2p1 for our measurements.
... | PaLM 2 Technical Report |
Additionally, LLMs are highly skilled in open-ended generations. One example is that the news articles generated
by LLMs are almost indistinguishable from real news articles by humans [16]. LLMs are remarkably adept at code
synthesis as well. Either for text-code generation, such as HumanEval [18] and MBPP [7], or for ... | Harnessing the Power of LLMs in Practice- A Survey on ChatGPT and Beyond |
In closely related work, Uesato et al. (2022) describe two distinct meth-
ods for training reward models: outcome supervision and process supervision.
Outcome-supervised reward models (ORMs) are trained using only the final
result of the model’s chain-of-thought, while process-supervised reward models
(PRMs) receive fe... | Let’s Verify Step by Step |
Target → Charles Darwin
Target → Turkey
Figure 3: Formatted dataset example for Natural Questions (left) & TriviaQA (right).
5https://competitions.codalab.org/competitions/17208
B MMLU
Social Science
Social Science | LLaMA- Open and Efficient Foundation Language Models |
State of Crypto today
Progress. New builders are entering web3 at record pace. Academic research is accelerating. Ambitious products are launching regularly. Key
infrastructure is improving. A recent example: Ethereum’s successful transition to energy-saving “proof-of-stake”, one of the most significant upgrades
in t... | State-of-Crypto2023 |
Video Diffusion Model. To model the temporal properties of videos and simultaneously maintain
vision generation quality, we construct the video diffuser by extending the image diffuser with
temporal modules. Specifically, we insert pseudo-temporal attention before the residual block [13].
However, we argue that pseudo-t... | Any-to-Any Generation via Composable Diffusion |
[Chen et al., 2021] Chen, M., Tworek, J., Jun, H., Yuan, Q., Pinto, H. P. d. O., Kaplan, J., Edwards, H.,
Burda, Y., Joseph, N., Brockman, G., et al. (2021). Evaluating large language models trained on code.
arXiv preprint arXiv:2107.03374. | Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback |
human ratings. We randomly sample from the set of labeled comparisons to compute Elo [16, 17].
Elo rating, which is widely used in chess and other games, is a measure of the expected win-rate
relative to an opponent’s win rate, for example, an Elo of 1100 vs 1000 means the Elo 1100 player
has an expected win-rate of ap... | QLORA |
Figure 1: Illustration of the process from digitization to advanced computational analysis. | UNDERSTANDINGANDCREATINGARTWITHAI-REVIEWAND OUTLOOK |
are strongly pulled towards deploying to be liable to misaligned power-seeking on at least some
physics-compatible inputs. And if, as seems plausible to me, it is difficult to adequately predict and
control the full range of inputs a system will receive, and how it will behave in response (especially if
the world is rap... | Is Power-Seeking AI an Existential Risk? |
r
e
t
u
r
n
s
.
L
e
v
e
l
-
2
e
x
a
m
i
n
e
s
t
h
e
a
b
i
l
i
t
y
t
o
r
e
t
r
i
e
v
e
t
h
e
A
P
I
.
T
h
e
m
o
d
e
l
n
e
e
d
s
t
o
s
e
a
r
c
h
f
o
r
p
o
s
s
i
b
l
e
A
P
I
s
t
h
a
t
m
a
y
s
o
l
v
e
t
h
e
u
s
e
r
ʼ
s
r
e
q
u
i
r
e
m
e
n
t
a
n
d
l
e
a
r
n
h
o
w
t
o
u
s
e
... | LLM Powered Autonomous Agents _ Lil'Log |
50% I’m glad you asked! I would love to go grab a slice of pizza after work today.
But let me tell you, I have some strong opinions about pizza. First of all, I think
Chicago pizza is the best. It’s deep dish, it’s thick, it’s cheesy, and it’s just all
around delicious. I can’t stand thin crust pizza, it’s just not wor... | Llama2 |
3-shot zero-shot
37.6%
37.4%
20.4%
40.2%
30.5% 43.4%
34.8% 44.4%
40.9% 46.2%
42.7% 49.4%
Table 8: Impact of self-instruct data. Impact of self-instruct data (SI) on the MBPP and HumanEval
scores of our self-instruct models. The scores are computed using greedy decoding. In MBPP zero-shot,
we prompt the model to gener... | CodeLlama2 |
Transparency reports also vary greatly in detail. Take, for instance, the
aforementioned Community Guideline reports. YouTube’s report documents
the number of channels and videos removed for eleven different types of
standards violations (e.g., spam, nudity, promotion of violence and extremism)
14 Access Now (2016).
1... | Social_Media_and_Democracy |
[95] A. Giachanou, G. Zhang, and P. Rosso, ‘‘Multimodal multi-image fake
news detection,’’ in Proc. IEEE 7th Int. Conf. Data Sci. Adv. Anal.
(DSAA), Oct. 2020, pp. 647–654.
[96] S. Girgis, E. Amer, and M. Gadallah, ‘‘Deep learning algorithms for
detecting fake news in online text,’’ in Proc. 13th Int. Conf. Comput. En... | A_Comprehensive_Review_on_Fake_News_Detection_With_Deep_Learning |
generated images, opening up new opportunities for eval-
uating sample correspondence. Therefore, we propose a
new evaluation metric, named Video-Music CLIP Precision
(VMCP), which extends the vision-language CLIP model
to video and music domain to measure the video-music cor-
respondence. With VMCP and subjective eval... | VideoBackgroundMusicGeneration |
Hogg, D. (1983). Model-based vision: a program to see a
walking person. Image and Vision computing, 1(1):5–
20.
Jiang, H. (2010). 3d human pose reconstruction using mil-
In Pattern Recognition (ICPR),
lions of exemplars.
2010 20th International Conference on, pages 1674–
1677.
Lee, M. W. and Nevatia, R. (2009). Human... | VISAPP_HumanPoseEstimation |
tiple, diverse perspectives using an LLM. This approach not
only captures the explicit information users seek but also un-
covers deeper, transformative knowledge. The fusion pro-
cess involves parallel vector searches of both original and
expanded queries, intelligent re-ranking to optimize results,
and pairing the be... | RAG forLargeLanguageModels-ASurvey |
the faculty office/Admissions will be passed the similarity report and the UCAS application
flagged. The admissions selector will review the application in the normal way against the usual
initial criteria for consideration for entry to the programme. If the applicant does not meet the
criteria, the application will... | UCL Academic Manual |
Word-position-dependent phone The possible absence of silence between words in the frame-
level phone transcript can make it hard for the audio model to identify word boundaries. To help
the audio model identify the word boundary which is important when reading a sentence, we
introduce word-position-dependent phones wh... | Voicebox-Text-GuidedMultilingual UniversalSpeechGenerationatScale |
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... | How Are Consumers Using Generative AI_ _ Andreessen Horowitz |
A Suite for Analyzing Large Language Models | Pythia- A Suite for Analyzing Large Language Models Across Training and Scaling |
2020. Scaling laws for autoregressive generative modeling. arXiv preprint arXiv:2010.14701 (2020).
[102] Danny Hernandez, Jared Kaplan, Tom Henighan, and Sam McCandlish. 2021. Scaling laws for transfer. arXiv preprint arXiv:2102.01293 (2021).
[103] Joel Hestness, Sharan Narang, Newsha Ardalani, Gregory Diamos, Heewoo ... | TheEfficiencySpectrumofLargeLanguageModels-AnAlgorithmicSurvey |
25
Single AgentAgent-AgentAgent-Human Figure 8: Practical applications of the single LLM-based agent in different scenarios. In task-
oriented deployment, agents assist human users in solving daily tasks. They need to possess basic
instruction comprehension and task decomposition abilities. In innovation-oriented depl... | TheRiseandPotentialofLargeLanguageModel BasedAgents |
sequence parallelism. Additionally, its heterogeneous memory management component
enhances training efficiency in distributed environments with heterogeneous devices.
Mesh-TensorFlow [72] is a user-friendly framework seamlessly integrated with Ten-
sorFlow, specializing in model parallelism with a primary emphasis on dis... | Beyond Efficiency |
https://doi.org/10.1017/9781108890960 Published online by Cambridge University Press
Misinformation and Its Correction
185
that fact-checking articles publicized through this system may be effective in
increasing belief accuracy – especially on issues where individuals do not
possess strong prior attitudes. Another... | Social_Media_and_Democracy |
different applications. Specifically, modeling agencies pro-
vide such information about their models; accuracy is a re-
quirement for modeling clothing. Thus, we collect a diverse
set of such model images (with varied ethnicity, clothing,
and body shape) with associated measurements; see Fig. 2.
Since sparse anthropom... | Accurate 3D Body Shape Regression using Metric and Semantic Attributes |
teractions. Wang et al. [66] propose a localized composi-
ing, showing reduced generalization when deformations are
too far away from the training set. To tackle this, for-
tional model that combines discrete low-resolution voxels
ward deformation fields have been recently proposed [12]
with neural radiance fields. Their... | I M Avatar- Implicit Morphable Head Avatars from Videos |
Prompt tuning emerges as a powerful technique,
with innovations like UPRISE demonstrating the
versatility of prompt-based adjustments. SynTra
introduces synthetic tasks for mitigating hallucina-
tions in abstractive summarization, offering scal-
ability but raising questions about effectiveness
compared to human feedba... | AComprehensiveSurveyofHallucinationMitigationTechniquesinLarge LanguageModels |
B. Full Fine-tuning of PLMs
Full fine-tuning of transformer-based PLMs involves train-
ing the entire model, including all layers and parameters, on
a specific downstream task using task-specific data. Initially,
PLMs are trained on large-scale datasets with unsupervised
learning objectives like language modeling or m... | Parameter-EfficientFine-TuningMethods |
vastly simplifies the task of density estimation and data
synthesis. Our method naturally accommodates mixed data | Adversarial Random Forests for Density Estimation and Generative Modeling |
2 Literature review
2.1 Human perception of messages shared by bots | Use of bot and content flags to limit the spread of misinformation among social networks: a behavior and attitude survey |
3.2. Learn to understand videos
Visual Encoder Given a video denoted as v, we uni-
formly sample T frames, which are represented as v =
[v1, v2, ..., vT ]. For each individual frame vi (where i
ranges from 1 to T ), we utilize a pre-trained CLIP vi-
sual encoder to extract its visual features, denoted as f i
v =
ViT(vi... | GPT4Video |
[49] N. Jay, M. d’Aquin, Linked data and online classifications to organise mined patterns in patient data, in: AMIA Annual Symposium Proceedings,
[50] H. Paulheim, Generating possible interpretations for statistics from linked open data, in: Extended Semantic Web Conference, Springer, 2012,
[51] P. R... | Knowledge graphs as tools for explainable machine learning: A survey |
To quantify toxic language harms and bias, we follow a similar process as described in Chung et al. (2022). We sample
10 continuations per prompt, with top-k = 40 sampling, at temperature=1.0, score the continuations with Perspective
API, and compute the percentage of toxic responses disaggregated by identity groups. W... | PaLM 2 Technical Report |
[30] Marc Levoy and Pat Hanrahan. Light field rendering. SIG-
GRAPH, 1996. 2
[31] Zhengqi Li, Simon Niklaus, Noah Snavely, and Oliver Wang.
Neural scene flow fields for space-time view synthesis of dy-
namic scenes. CVPR, 2021. 2
[32] Lingjie Liu, Marc Habermann, Viktor Rudnev, Kripasindhu
Sarkar, Jiatao Gu, and Chr... | HumanNeRF- Free-viewpoint Rendering of Moving People from Monocular Video |
the electricity grid, which our faculty is helping to make completely sustainable and future-
proof. At the same time, we are developing the chips and sensors of the future, whilst also
setting the foundations for the software technologies to run on this new generation of
equipment – which of course includes AI. Meanwh... | Job details - TU |
[60] Peng Wang, Lingjie Liu, Yuan Liu, Christian Theobalt, Taku
Komura, and Wenping Wang. Neus: Learning neural implicit
surfaces by volume rendering for multi-view reconstruction.
In NeurIPS, 2021. 6
[61] Tengfei Wang, Bo Zhang, Ting Zhang, Shuyang Gu, Jianmin
Bao, Tadas Baltrusaitis, Jingjing Shen, Dong Chen, Fang
W... | Wonder3D |
log(D)/D of the PC. Therefore, the computation cost of Alg. 1 scales logarithmically w.r.t. D. The
set of PC units need to be re-evaluated, evali, is identified in line 4, and lines 6 evaluates these units
in a feedforward manner to compute the target probability (i.e., p(x1, . . . , xi)).
Specifically, to minimize compu... | LOSSLESS COMPRESSION WITH PROBABILISTIC CIRCUITS |
c
t
u
r
e
o
f
w
h
a
t
i
s
g
o
i
n
g
o
n
i
n
s
i
d
e
t
r
a
n
s
f
o
r
m
e
r
l
a
n
g
u
a
g
e
m
o
d
e
l
s
.
U
s
e
r
i
n
t
e
r
f
a
c
e
s
w
i
t
h
a
c
c
e
s
s
t
o
d
a
t
a
b
a
s
e
s
o
f
e
x
p
l
a
n
a
t
i
o
n
s
c
o
u
l
d
e
n
a
b
l
e
a
m
o
r
e
m
a
c
r
o
-
f
o
c
u
s
e
d
a
p
p
r
o
a
c
h
... | Language models can explain neurons in language models |
Method
mark, and the performance falloff can be steep (and relatively
(d3) text-davinci-003
uninformative) across LLMs with decreasing model size and
(d3) w/ random A
data scale, making it difficult to measure incremental progress
(p) PaLM [55, 56]
towards pattern machines that could be used for sequence trans-
(d2) te... | LargeLanguageModelsasGeneralPatternMachines |
data partitions, and model states including model parameters, gradients, and optimizer
states are required by each data partition. Given the substantial size of LLMs, apply-
ing DP to LLMs in a naive manner is impractical. To end this, ZeRO [15], PaLM [6]
and Fairscale [71] introduce approaches for enhancing the efficien... | Beyond Efficiency |
generations. arXiv preprint arXiv:2307.06857, 2023.
Aitor Lewkowycz, Anders Andreassen, David Dohan, Ethan Dyer, Henryk Michalewski, Vinay Ra-
masesh, Ambrose Slone, Cem Anil, Imanol Schlag, Theo Gutman-Solo, et al. Solving quantitative
reasoning problems with language models. Advances in Neural Information Processing... | UNIVERSALSELF-CONSISTENCYFORLARGELANGUAGEMODELGENERATION |
A 627B token cleaned and deduplicated version of RedPajama.
Srivastava, A., Rastogi, A., Rao, A., Shoeb, A. A. M., Abid, A., Fisch, A., Brown, A. R., Santoro,
A., Gupta, A., Garriga-Alonso, A., et al. (2022). Beyond the imitation game: Quantifying and
extrapolating the capabilities of language models. arXiv preprint a... | TinyLlama |
Reporting and Interviews with Participants in Takedown Processes
Other researchers have carried out the painstaking work of tracking global
developments and seeking out and interviewing individual participants.
Rebecca McKinnon laid important groundwork for this in her 2012 book
46 Platforms have recently taken a few ... | Social_Media_and_Democracy |
[71] Josh McCoy, Mike Treanor, Ben Samuel, Noah Wardrip-Fruin, and Michael
Mateas. 2011. Comme il faut: A System for Authoring Playable Social Models.
In Proceedings of the AAAI Conference on Artificial Intelligence and Interactive
Digital Entertainment (AIIDE’11). AAAI, Stanford, CA, USA, 38–43.
[72] Marvin Minsky an... | Generative Agents- Interactive Simulacra of Human Behavior |
1Training
datasets:
https://huggingface.co/collections/distil-whisper/
training-datasets-6538d05c69721489d1db1e49
5
Speakers Domain
unknown Government, interviews
unknown Audiobook, podcast, YouTube
unknown Narrated Wikipedia
Table 2: Summary of the open-source datasets used for training. For some datasets, the ... | DISTIL-WHISPER |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.