text stringlengths 1 1k ⌀ | title stringclasses 230
values |
|---|---|
thought,
theories
conspiracy
included promoting | Social_Media_and_Democracy |
In Proceedings of the IEEE/CVF Conference on Computer
Vision and Pattern Recognition, pages 4401–4410, 2019. 5
[31] Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten,
Jaakko Lehtinen, and Timo Aila. Analyzing and improv-
In Proceedings of
ing the image quality of StyleGAN.
the IEEE/CVF Conference on Computer Vis... | AG3D- Learning to Generate 3D Avatars from 2D Image Collections |
[61] Lee Xiong, Chenyan Xiong, Ye Li, Kwok-Fung Tang, Jialin Liu, Paul N. Bennett, Junaid
Ahmed, and Arnold Overwijk. Approximate nearest neighbor negative contrastive learning for
dense text retrieval. In 9th International Conference on Learning Representations, ICLR 2021,
Virtual Event, Austria, May 3-7, 2021. OpenRe... | E5 |
[36] John G Beerends, Christian Schmidmer, Jens Berger, Matthias Obermann, Raphael Ullmann, Joachim Pomy, and
Michael Keyhl. 2013. Perceptual objective listening quality assessment (polqa), the third generation itu-t standard for
end-to-end speech quality measurement part i—temporal alignment. Journal of the Audio Engi... | AReviewofDeepLearningTechniquesforSpeechProcessing |
tence with Generate a sentence that includes these {DOMAIN} keywords. We also
turn the task around by taking the sentence as input and asking the model to find the keywords about
the target domain using What keywords about {DOMAIN} can be extracted from this | ADAPTINGLARGELANGUAGEMODELSVIA READINGCOMPREHENSION |
Compacter [58] is developed based on adapters,
sum of Kronecker products, i.e., W =(cid:80)n | Parameter-EfficientFine-TuningMethods |
[12] Diederik P. Kingma and Max Welling.
Auto-encoding variational bayes.
CoRR,
abs/1312.6114, 2013.
[13] Jungil Kong, Jaehyeon Kim, and Jaekyoung Bae. Hifi-gan: Generative adversarial networks
for efficient and high fidelity speech synthesis. Advances in Neural Information Processing
Systems, 33:17022–17033, 2020.
[... | Text-to-Audio Generation using Instruction-Tuned LLM and Latent Diffusion Model |
with query set Q = {qi}i: | Tool Learning with Foundation Models |
t
c
o
m
p
a
r
a
b
l
e
w
i
t
h
h
u
m
a
n
s
.
(
I
m
a
g
e
s
o
u
r
c
e
:
A
n
i
m
a
l
s
u
s
i
n
g
t
o
o
l
s
)
M
R
K
L
(
K
a
r
p
a
s
e
t
a
l
.
2
0
2
2
)
,
s
h
o
r
t
f
o
r
“
M
o
d
u
l
a
r
R
e
a
s
o
n
i
n
g
,
K
n
o
w
l
e
d
g
e
a
n
d
L
a
n
g
u
a
g
e
”
,
i
s
a
n
e
u
r
o
-
s
y
m
b
o
l
i
c
... | LLM Powered Autonomous Agents _ Lil'Log |
n
s
i
n
t
e
r
m
s
o
f
t
o
k
e
n
s
t
h
a
t
c
a
u
s
e
t
h
e
m
t
o
a
c
t
i
v
a
t
e
,
o
r
t
o
k
e
n
s
t
h
a
t
t
h
e
n
e
u
r
o
n
c
a
u
s
e
s
t
h
e
m
o
d
e
l
t
o
s
a
m
p
l
e
.
F
u
n
d
a
m
e
n
t
a
l
l
y
,
t
h
e
s
e
t
e
c
h
n
i
q
u
e
s
d
e
p
e
n
d
o
n
m
u
l
t
i
p
l
y
i
n
g
t
h
e
... | Language models can explain neurons in language models |
W
e
a
l
s
o
o
b
t
a
i
n
e
d
a
h
u
m
a
n
b
a
s
e
l
i
n
e
f
r
o
m
l
a
b
e
l
e
r
s
a
s
k
e
d
t
o
w
r
i
t
e
e
x
p
l
a
n
a
t
i
o
n
s
f
r
o
m
s
c
r
a
t
c
h
,
u
s
i
n
g
t
h
e
s
a
m
e
s
e
t
o
f
5
t
o
p
-
a
c
t
i
v
a
t
i
n
g
t
e
x
t
e
x
c
e
r
p
t
s
t
h
a
t
t
h
e
e
x
p
l
a
i
n
e
r
... | Language models can explain neurons in language models |
fed to the denoising task and the corrupted spans are used as targets to be recovered.
As an example, to construct an objective analogous to causal language modeling using this formulation, one
would simply set (µ = L, r = 1.0, n = 1), i.e. a single span with its span length equal to the length of the
sequence. To expr... | UL2- Unifying Language Learning Paradigms |
there has been a significant surge in interest
regarding PEFT methods, as demonstrated by the growing
number of studies depicted in Fig. 1. This also leads to a
few surveys on PEFT approaches for the PLMs. However,
the existing surveys have certain limitations. Ding et al.
[12] conducted a comprehensive study on PEFT m... | Parameter-EfficientFine-TuningMethods |
Casella, P., & Paiva, A. (2001). Magenta: An architecture for real time au-
tomatic composition of background music. In International Workshop on
Intelligent Virtual Agents (pp. 224–232). Springer.
41
Castellano, B. (2018). Pyscenedetect: Intelligent scene cut detection and video
splitting tool.
Chase, W. (2006)... | Video2Music |
Topic #10
said
state
government
law
court
data
cells
patients
study
c
said
like
time
know
way
like
time
new
game
way
let
number
model
system
theorem
x
z
divide
var
y
court
â
law
case
state
text
color
width
font
height
device
invention
layer
film
power
health
data
care
based
study
said
man
little
like
time
like
know
righ... | The Pile- An 800GB Dataset of Diverse Text for Language Modeling |
e
a
c
h
t
o
k
e
n
i
n
t
h
e
v
o
c
a
b
u
l
a
r
y
.
T
h
e
p
r
e
d
i
c
t
e
d
a
c
t
i
v
a
t
i
o
n
w
i
t
h
t
h
i
s
m
e
t
h
o
d
d
e
p
e
n
d
s
o
n
l
y
o
n
t
h
e
c
u
r
r
e
n
t
t
o
k
e
n
a
n
d
n
o
t
o
n
t
h
e
p
r
e
c
e
d
i
n
g
c
o
n
t
e
x
t
.
T
h
e
s
e
s
c
a
l
a
r
v
a
l
u
e
s
a
... | Language models can explain neurons in language models |
performance increases with the number of samples (Section 5). Our use of bootstrapping ensures
that we can still benefit from the variance reduction obtained from generating a much larger set of
𝐾 (cid:29) 𝑘 samples to estimate the 𝑛@𝑘 metric.
The setting we use to model programming competitions is 10@𝑘 – 10 submis... | alphacode |
a16z crypto
State of Crypto
2023
Adoption Indicators: Demand Side
47
The number of mobile wallet users has declined
since early 2022
Mobile
Wallet
Users
Number of estimated
mobile wallet users across
all tracked mobile wallets
during the month.
25M
20M
15M
10M
5M
0
2016
20... | State-of-Crypto2023 |
sion (Forsgren & Martiros, 2022), both in terms of quality
and adherence to the caption. Furthermore, since describing
some aspects of music with words can be difficult or even
impossible, we show how our method supports condition-
ing signals beyond text. Concretely, we extend MusicLM
to accept an additional melody in ... | MusicLM |
Proceedingsofthe59thAnnualMeetingoftheAssociationforComputationalLinguisticsandthe11thInternationalJointConferenceonNaturalLanguageProcessing,pages4582–4597August1–6,2021.©2021AssociationforComputationalLinguistics4582 In this paper, we propose prefix-tuning, a
lightweight alternative to fine-tuning for natural lan-
guag... | Prefix-Tuning |
3.3
Inference
During inference, we initialize xt with Gaus-
sian noise and iteratively remove a (scheduler-
determined) proportion of the noise over T time
steps to obtain ˆx0 (Ho et al., 2020). During this it-
erative denoising, we do not use the decoder. After
this iterative procedure, the decoder produces the
fina... | CODEFUSION |
cp and minbucket size tend to be better hyper-parameter
configurations.
Space: 5971
1. Generally, larger datasets require higher nrounds and larger
subsample values.
2. The majority class size and minority class size of the dataset can
influence the configuration of alpha, booster, colsample bylevel,
colsample bytr... | MLCopilot- Unleashing the Power of Large Language Models in Solving Machine Learning Tasks |
There are also three notable points regarding why small language models fail. The first observation
is that small language models fail at even relatively easy symbol mapping tasks. As demonstrated
in Section 5, for even symbolic reasoning tasks that only require generalization to new examples
using the same chain of tho... | Chain-of-Thought Prompting Elicits Reasoning in Large Language Models |
I’ve heard that Sam Moore is considering running for local
mayor.
• Was there a Valentine’s day party?
Yes, Isabella Rodriguez organized a Valentine’s Day party at
Hobbs Cafe.
• Who is [Ayesha Khan]?
Ayesha Khan is a fellow student at Oak Hill College. She
is doing her senior thesis on the use of language in Shake-
sp... | Generative Agents- Interactive Simulacra of Human Behavior |
There are many opportunities from these developments, and these can only be realised if the
risks are mitigated. There are several deep, unsolved cross-cutting technical and social risk
factors that exacerbate the risks. We outlined examples of societal harms, risks of misuse from
bad actors, and even the possibilit... | Capabilities and risks from frontier AI |
ing such as document level filtering (Brown et al.,
2020; Wenzek et al., 2019), or n-sentence level
deduplication with very aggressive heuristics (Raf-
fel et al., 2019). | The Pile- An 800GB Dataset of Diverse Text for Language Modeling |
SIQA PIQA Arc-E Arc-C OBQA
LLaMA 33B
50.2
Humpback 33B 53.42
LLaMA 65B
52.3
Humpback 65B 60.44
58.6
46.4
60.2
64.0
Table 5: Comparison on zero-shot commonsense reasoning.
54.8
68.50
56.0
72.96
80.0
84.44
78.9
88.67
82.2
74.54
82.8
78.9
LLaMA 65B, 5-shot
LLaMA 65B, 0-shot
Humpback 65B, 0-shot
Humanities
61.8
63.0
... | Self-AlignmentwithInstructionBacktranslation |
We provide additional examples of humor generation for the multimodal multilingual LLMs mentioned in Table 2 (main text)
to illustrate the effectiveness of CLoT. Fig. 13, 14 showcase responses on the task of Image&Text to Text in Chinese and
Japanese, respectively. As English Oogiri data lacks Image&Text to Text sample... | Let’sThinkOutsidetheBox |
,campaign,grassroots"}Observation:1680676119.1573935.jpgThought:Great,nowthatwehavetheimage,wecanaddthetextandimagepagewiththecasestudy.Action:add_text_image_pageActionInput:{"title":"CaseStudy:AlexandriaOcasio-Cortez’sCampaign","bullet_items":["Identifiedkeyissuesaffectingherdistrict","Developedaclearmessageofprogress... | Tool Learning with Foundation Models |
[46] J. Von Oswald, E. Niklasson, E. Randazzo, J. Sacramento, A. Mordvintsev, A. Zhmoginov, and M. Vladymyrov.
Transformers learn in-context by gradient descent. In International Conference on Machine Learning (ICML),
2023.
[47] L. Kirsch, J. Harrison, J. Sohl-Dickstein, and L. Metz. General-purpose in-context learnin... | LargeLanguageModelsasGeneralPatternMachines |
4https://www.llamaindex.ai
5https://www.langchain.com/
6https://haystack.deepset.ai/blog/
enhancing-rag-pipelines-in-haystack
7https://huggingface.co/BAAI/bge-reranker-large
Figure 3: Comparison between the three paradigms of RAG
is depicted in Figure 3. However, Modular RAG is not stan-
dalone. Advanced RAG is a s... | RAG forLargeLanguageModels-ASurvey |
Victor Sanh, Lysandre Debut, Julien Chaumond, and
Thomas Wolf. 2019. Distilbert, a distilled version
of BERT: smaller, faster, cheaper and lighter. CoRR,
abs/1910.01108.
Victor Sanh, Albert Webson, Colin Raffel, Stephen H.
Bach, Lintang Sutawika, Zaid Alyafeai, Antoine
Chaffin, Arnaud Stiegler, Arun Raja, Manan Dey,
M ... | LaMini-LM- A Diverse Herd of Distilled Models from Large-Scale Instructions |
Pretraining Tasks. We consider two vision-only tasks in the pretraining:
for masked image modeling
(MIM) as well as image infilling, we borrow the idea of blockwise masking (Bao et al., 2022) and let the model
recover the masked patches in the middle part by generating the corresponding codes. The corresponding
instruc... | BiomedGPT |
t
h
a
t
c
a
n
s
u
c
c
e
e
d
a
t
t
h
e
s
e
s
p
e
c
i
a
l
i
s
t
t
a
s
k
s
a
n
d
n
o
t
b
e
t
a
i
n
t
e
d
b
y
c
r
i
t
i
c
a
l
e
r
r
o
r
s
i
s
a
n
o
u
t
s
t
a
n
d
i
n
g
a
r
e
a
f
o
r
t
h
e
16/08/2023, 14:36 | The a16z Investment Thesis on AI in Bio + Health _ Andreessen Horowitz |
(cid:1) (cid:12) Mk
(cid:0)(cid:13)(cid:13)(cid:0)I R
(cid:80)Nt
k=1 10 log10
k − Ik
(cid:13)(cid:13)2
(cid:1), | Text2NeRF- Text-Driven 3D Scene Generation with Neural Radiance Fields |
133Christiano’s (2018) second scenario is one example of this.
134Many humans, for example, have proven willing to risk their lives for personal power, glory, national
strength, military victory, a social cause, an ideology, even scientific discovery, etc. “X would involve someone
knowingly risking death” doesn’t seem t... | Is Power-Seeking AI an Existential Risk? |
reported in Tab. 7, the input representations are trained
on a dataset containing 96,000 training scenes of solely
the TAMP environment, i.e. no other data is part of the
mixture. For 3-5 objects in the scene, which is the same
number as in the training set, most input representations
perform similarly well. However, w... | PaLM-E- An Embodied Multimodal Language Model |
in Osaka where a taiko drum group, made up
We visited a hall
exclusively of young burakumin, were about
their weekly
rehearsal. The small gymnasium was filled with taiko drums of all sizes.
The smallest was about the size of a snare drum, the largest about the
size of a compact car. The Japanese drum group Kodo have mad... | The Pile- An 800GB Dataset of Diverse Text for Language Modeling |
Prompting with feedback. Recent works have shown the great promise of RLHF-trained models
to generate critiques with prompting, which reduces harmful model outputs [3, 18] and improves the
performance on some reasoning tasks [50, 35, 28, 36]. Reflexion [50] prompts an agent powered
with a large language model to reflect ... | Teaching Large Language Models to Self-Debug |
of style classification. The advantage of CNN-based features,
particularly in combination with other hand-crafted features,
was confirmed for artist [12], style [13] and genre classifi-
cation [14]. Besides using pre-trained CNNs just as feature
extractors, Girshick et al. showed that further improvement of
performance fo... | A_Deep_Learning_Perspective_on_Beauty_Sentiment_and_Remembrance_of_Art |
[Kang et al., 2023] Minki Kang, Jin Myung Kwak, Jinheon
Baek, and Sung Ju Hwang. Knowledge graph-augmented
language models for knowledge-grounded dialogue gener-
ation. arXiv preprint arXiv:2305.18846, 2023.
[Kaplan et al., 2020] Jared Kaplan, Sam McCandlish, Tom
Henighan, Tom B Brown, Benjamin Chess, Rewon Child,
Sco... | RAG forLargeLanguageModels-ASurvey |
SpaceImage Space at t1<latexit sha1_base64="SVG2hxvF7EcP+hdssaUPWfkvBZw=">AAAB6nicbVBNS8NAEJ3Ur1q/oh69LBbBU0lE0GPBi8eK9kPaUDbbTbt0swm7E6GE/gQvHhTx6i/y5r9x2+agrQ8GHu/NMDMvTKUw6HnfTmltfWNzq7xd2dnd2z9wD49aJsk0402WyER3Qmq4FIo3UaDknVRzGoeSt8PxzcxvP3FtRKIecJLyIKZDJSLBKFrpHvt+3616NW8Oskr8glShQKPvfvUGCctirpBJakzX91IMcqpRMMmnlV... | BANMo- Building Animatable 3D Neural Models from Many Casual Videos |
provethecurrentstateofaffairsbyintroducinganalgorithmthatattainsthedataefficiencyandreliableperformanceofTRPO,whileusingonlyfirst-orderoptimization.Weproposeanovelobjectivewithclippedprobabilityratios,whichformsapessimisticestimate(i.e.,lowerbound)oftheperformanceofthepolicy.Tooptimizepolicies,wealternatebetweensamplingdat... | PPO |
274
Tim Hwang
automated methods to deal with questions of truth, falsity, and information
quality may hinder other, alternative models that leverage user and community
participation to filter for these criteria (Grimmelmann 2015; Rogers 2017).
the collaboratively edited encyclopedia, has been relatively
Wikipedia,
suc... | Social_Media_and_Democracy |
• Novel methodologies and tools to create knowledge graphs, offering advanced user
experiences, accessible to non-experts and using the latest tech (audio and video processing,
intelligent assistants, AR and VR etc.)
• Methodologies and techniques to acquire and encode common sense knowledge at scale
• Quality... | informatics-phd-projects-2022-23 |
The Internet at the outset promised to be as diverse and competitive as legacy
media when it first came into existence as a public platform in the 1990s. Since
then, however, it has moved in the direction of the broadcast media of the
1950s. Two very large platforms, Facebook and Google, and a third somewhat
smaller one... | Social_Media_and_Democracy |
;
a
n
d
R
&
D
t
h
a
t
g
o
e
s
b
e
y
o
n
d
t
h
e
a
b
i
l
i
t
i
e
s
o
f
h
u
m
a
n
s
c
i
e
n
t
i
s
t
s
.
W
e
’
v
e
b
e
e
n
i
n
v
e
s
t
i
n
g
i
n
s
u
c
h
g
r
o
u
n
d
b
r
e
a
k
i
n
g
c
o
m
p
a
n
i
e
s
f
o
r
y
e
a
r
s
—
y
o
u
c
a
n
s
e
e
a
f
u
l
l
l
i
s
t
o
f
o
u
r
A
I
a
n
d
... | The a16z Investment Thesis on AI in Bio + Health _ Andreessen Horowitz |
vide full details in the supplement.
Naturally paired modalities and datasets. We use IM-
AGEBIND on six modalities - image/video,
text, audio,
depth, thermal images, and IMU. As described in § 3.3, we
treat videos as 2 frame images and process them the same
as images. For the naturally available paired data, we use
th... | IMAGEBIND- One Embedding Space To Bind Them A |
2. Background
Language model pre-training The goal of language
model pre-training is to learn useful representations of lan-
guage, usually from unlabeled text corpora. The resulting
pre-trained model can then be further trained (fine-tuned) for
a downstream task of primary interest (in our case, Open-
QA), often leadin... | REALM |
the version control system.
7. Create a release process: Define a release process to deploy the code to different environ-
ments.
8. Monitor the system: Monitor the system to ensure that the code is being versioned correctly
and there are no conflicts.
9. Backup the repository: Backup the repository regularly to ensure t... | WizardLM- Empowering Large Language Models to Follow Complex Instructions |
[21] Junnan Li, Dongxu Li, Caiming Xiong, and Steven C. H. Hoi. Blip: Bootstrapping Language-
Image Pre-training for Unified Vision-Language Understanding and Generation. In International
Conference on Machine Learning (ICML), pages 12888–12900, 2022.
[22] Lvmin Zhang and Maneesh Agrawala. Adding Conditional Control to... | HuggingGPT- Solving AI Tasks with ChatGPT and its Friends in Hugging Face |
In general, and partly due to various constraints the factors and mechanisms just discussed imply,
I don’t think it at all a foregone conclusion that APS systems seeking to gain/maintain power in
misaligned ways, especially on very large scales, would succeed in doing so. Indeed, even beyond
early warning shots in weak... | Is Power-Seeking AI an Existential Risk? |
(4) Calculation: Evaluating whether the model can perform accurate mathematical computations of
the provided formulas in the domains of math, biology, chemistry and physics.
(5) Accuracy: Evaluating whether the model can perform correctly in the corresponding for a given
instruction.
Then they should rank the four resp... | WizardLM- Empowering Large Language Models to Follow Complex Instructions |
learning for medical imaging. Advances in neural information processing systems, 32, 2019.
Laila Rasmy, Yang Xiang, Ziqian Xie, Cui Tao, and Degui Zhi. Med-bert: pretrained contextualized em-
beddings on large-scale structured electronic health records for disease prediction. NPJ digital medicine,
4(1):86, 2021.
Scot... | BiomedGPT |
Pretrained multilingual language models can
help bridge the digital language divide, en-
abling high-quality NLP models for lower-
resourced languages. Studies of multilingual
models have so far focused on performance,
consistency, and cross-lingual generalisation.
However, with their wide-spread application
in the wil... | Are Pretrained Multilingual Models Equally Fair Across Languages? |
Nisan Stiennon, Long Ouyang, Jeff Wu, Daniel M.
Ziegler, Ryan Lowe, Chelsea Voss, Alec Radford,
Dario Amodei, and Paul F. Christiano. 2020. Learn-
ing to summarize from human feedback. CoRR,
abs/2009.01325.
Ross Taylor, Marcin Kardas, Guillem Cucurull,
Thomas Scialom, Anthony Hartshorn, Elvis Saravia,
Andrew Poulton, ... | Unleashing Infinite-Length Input Capacity for Large-scale Language Models with Self-Controlled Memory System |
Contributions. Our contributions are threefold:
• We introduce a novel cooperative agent framework, role-playing , that allows communicative
agents to collaborate autonomously toward completing tasks while requiring minimal human
intervention.
• Our framework offers a scalable approach for studying the cooperative be... | CAMEL- Communicative Agents for “Mind” Exploration of Large Scale Language Model Society |
2.11 PubMed Abstracts
PubMed Abstracts consists of the abstracts from 30
million publications in PubMed, the online repos-
itory for biomedical articles run by the National
Library of Medicine. While the PMC (see Section
2.2) provides full-text access, the subset of cover-
age is significantly limited and biased towards... | The Pile- An 800GB Dataset of Diverse Text for Language Modeling |
Exploring by watching demos. An alternative
and often more effective exploration method in-
volves the agent observing human demonstrations.
These demonstrations provide the agent with ex-
amples of efficient app usage, especially for un-
derstanding complex functionalities that might be
challenging to discover through... | AppAgents |
Fast and memory-efficient attention. We implemented our own version of FlashAttention (Dao et al.,
2022) to improve memory usage and speed on the self-attention layers. Our version is on par with or
better than the original on all cases considered, while covering more use-cases and hardware. Due to the
GPU hardware speci... | DINOv2- Learning Robust Visual Features without Supervision |
Media Regulation in the United States and Europe
217
A second idea is to restrict exclusionary behavior through the purchasing of
potentially competitive start-ups. Facebook has already been subject to
substantial criticism for its purchases of Instagram and WhatsApp and is busy
seeking to integrate them with its exi... | Social_Media_and_Democracy |
We think it is important that workers, policymakers, and researchers not focus overly on just
the current state of capabilities. We expect GPT-4 to accelerate development of new applications
built on top of generative models, and that these applications will often solve more complex tasks
than the model on its own. Ind... | gpt-4-system-card |
to significantly reduce the ease of producing various kinds of potentially harmful content, thereby
making GPT-4-launch significantly safer than GPT-4-early along these dimensions. | gpt-4-system-card |
1. Lack of access to current information. Certain data constantly change – the
exchange rate between the dollar and the Moroccan Dirham, current COVID
numbers, the stock price of AAPL, the weather in Vancouver (OK, not so
much), or even the current date. It’s impossible, by their design, for pretrained
language models ... | MRKL Systems |
• Positional Encoding. Unlike recurrent neural networks (RNNs) or convolutional
neural networks (CNNs), Transformers do not inherently possess knowledge of
the order or position of tokens in a sequence. To address this limitation, positional
encodings are introduced. These encodings are added to the word embeddings
and... | Beyond Efficiency |
Table 7: 2-bit GPTQ quantization results with
varying group-sizes; perplexity on WikiText2.
Figure 4: GPTQ at 4-bit with different
group-sizes on medium sized OPT models. | GPTQ |
S3Delta-M (Search for Sparse Structure of Delta Tun-
ing Mix) [61] is a mixture of LoRA, Compacter (low-
rank adapter), BitFit, and LNFit5. Different from the simple
incorporation of PEFT techniques, S3Delta-M is developed
by conducting a differentiable delta tuning structure search.
It explicitly controls sparsity and... | Parameter-EfficientFine-TuningMethods |
Parameter-Efficient Tuning. In practice, we may tune the model on some specific datasets. Parameter-Efficient Tuning
(PET) is an efficient technique to tune a small portation of model parameters (or extra parameters) while freezing
most parameters of the pre-trained LLMs. The main goal of PEFT is to greatly decrease th... | Harnessing the Power of LLMs in Practice- A Survey on ChatGPT and Beyond |
A recent decision of the US Court of Appeals for the Ninth Circuit appears to
recognize the pervasive impact of platform control of information. In hiQ v.
LinkedIn, No. 17–16783 (9th Cir. 2019), the Court protected a company’s
right to scrape user-provided data on LinkedIn. As the Court explained, “giving
companies lik... | Social_Media_and_Democracy |
set of actions, and then constructs a new frame F2 = α(F1, V C , g). However, different methods require different type and
amount of such extra information, so this approach would quickly result in a plethora of similar, yet different, transformation
concepts, which is rather the opposite of our aims. Hence, we will ... | A-framework-for-analysing-state-abstraction-metho_2022_Artificial-Intelligen |
Query: Given a collection of images A: /examples/a.jpg, B: /examples/b.jpg, C: /examples/c.jpg, please tell me how many zebras in these pictures?Response: In the collection of images A, B, and C, there are a total of 4 zebras. To determine this, I first used an image-to-text model to generate captions... | HuggingGPT- Solving AI Tasks with ChatGPT and its Friends in Hugging Face |
Furthermore, improving the interpretability of RAG-driven
models continues to be a key goal. Doing so would allow
users to understand the reasoning behind the responses gener-
ated by the model, thereby promoting trust and transparency
in the use of RAG applications.
Technical Stack
The development of the RAG ecosystem... | RAG forLargeLanguageModels-ASurvey |
How can we disentangle individuals’ self-selection into networks from
the impact on their attitudes? Most of the empirical research cited in this
chapter relies on cross-sectional evidence from observational studies. One
challenge when deriving causally valid conclusions from this evidence is that
individuals’ media co... | Social_Media_and_Democracy |
[106] Tongshuang Wu, Michael Terry, and Carrie J Cai. 2022. AI Chains: Transparent
and Controllable Human-AI Interaction by Chaining Large Language Model
Prompts. In CHI ’22: Proceedings of the 2022 CHI Conference on Human Factors in
Computing Systems.
[107] Qian Yang, Aaron Steinfeld, Carolyn Rosé, and John Zimmerman... | Generative Agents- Interactive Simulacra of Human Behavior |
Pre-fill and Chunking. When generating a sequence, we need to predict tokens one-by-one, as
each token is conditioned on the previous ones. However, the prompt is known in advance, and we
can pre-fill the (k, v) cache with the prompt. If the prompt is very large, we can chunk it into smaller
pieces, and pre-fill the ca... | Mistral7B |
I’ll understand misaligned behavior as a particular type of unintended behavior: namely, unintended
behavior that arises specifically in virtue of problems with an AI system’s objectives. Thus, for
example, a designer might intend for an AI system to make money on the stock market, and the
system might fail because it w... | Is Power-Seeking AI an Existential Risk? |
over 100 trials. We vary the input batch size, sequence length, and the adapter bottleneck dimension
r. We test two adapter designs: the original one by Houlsby et al. (2019), which we call AdapterH,
and a recent, more efficient variant by Lin et al. (2020), which we call AdapterL. See Section 5.1
for more details on th... | LORA |
Embedding-only upper bounds the performance
of discrete prompt optimization (Shin et al., 2020),
because discrete prompt restricts the embedding
layer to exactly match the embedding of a real word.
Consequently, we have this chain of increasing ex-
pressive power: discrete prompting < embedding-
only < prefix-tuning.
7.... | Prefix-Tuning |
passes and fine-tunes LLMs “with the same memory footprint as inference”. Requiring
55GB GPU memory, it can train a 30B model via full-parameter fine-tuning. | Beyond Efficiency |
the training-A data and likewise used re-ranker-A to process the training-B data, merging the two
to yield our LM prompt tuning training set. We trained a third re-ranker on the entire training set,
denoted re-ranker-All, and used it in order to create the data for the development and test sets. | STANDING ON THE SHOULDERS OF GIANT FROZEN LANGUAGE MODELS |
16
Large Language Models Cannot Self-Correct Reasoning Yet
Q: A fencing thrust with a sharp sword towards a person would result in
what?
Answer Choices: (A) injury (B) small cuts (C) fever (D) competition (E)
puncture wound.
Explain your reasoning. You must choose only one option from A to E. Your
final answer shoul... | LARGELANGUAGEMODELSCANNOTSELF-CORRECT REASONINGYET |
Generally, we make a best effort to submit scores to any leaderboard (unpublished test set) but refrain from
doing so in the cases where the labor costs to make such a submission is prohibitive - especially when the
existing state-of-the-art approach has made their dev scores available or when reporting on this particul... | UL2- Unifying Language Learning Paradigms |
Figure 1: SELF-DEBUGGING for iterative debugging using a large language model. At each debug-
ging step, the model first generates new code, then the code is executed and the model explains the
code. The code explanation along with the execution results constitute the feedback message, which
is then sent back to the mod... | Teaching Large Language Models to Self-Debug |
Hendricks, L. A., Burns, K., Saenko, K., Darrell, T., and Rohrbach, A. Women also snowboard: Overcoming bias in
captioning models (extended abstract), 2018.
Hendrycks, D., Burns, C., Kadavath, S., Arora, A., Basart, S., Tang, E., Song, D., and Steinhardt, J. Measuring
mathematical problem solving with the math datase... | PaLM 2 Technical Report |
[216] Umut Isik, Ritwik Giri, Neerad Phansalkar, Jean-Marc Valin, Karim Helwani, and Arvindh Krishnaswamy. 2020.
Poconet: Better speech enhancement with frequency-positional embeddings, semi-supervised conversational data,
and biased loss. arXiv preprint arXiv:2008.04470 (2020).
[217] Keith Ito and Linda Johnson. 2017... | AReviewofDeepLearningTechniquesforSpeechProcessing |
[158] Z. Jin, J. Cao, Y. Zhang, and J. Luo, ‘‘News verification by exploiting
conflicting social viewpoints in microblogs,’’ in Proc. 13th AAAI Conf.
Artif. Intell. (AAAI), 2016, pp. 2972–2978.
[159] X. Zhou and R. Zafarani, ‘‘Fake news detection: An interdisciplinary
research,’’ in Proc. Companion World Wide Web Conf.,... | A_Comprehensive_Review_on_Fake_News_Detection_With_Deep_Learning |
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015
1
PaMIR: Parametric Model-Conditioned Implicit
Representation for Image-based Human
Reconstruction
Zerong Zheng, Tao Yu, Yebin Liu, Member, IEEE and Qionghai Dai, Senior Member, IEEE | PaMIR- Parametric Model-Conditioned Implicit Representation for Image-based Human Reconstruction |
Large language models (LLMs) with instruc-
tion finetuning demonstrate superior genera-
tive capabilities. However, these models are
resource intensive. To alleviate this issue, we
explore distilling knowledge from instruction-
tuned LLMs to much smaller ones. To this end,
we carefully develop a large set of 2.58M in-
s... | LaMini-LM- A Diverse Herd of Distilled Models from Large-Scale Instructions |
32
Evaluation set
SynthBio
Language
German
"gender sets"
Italian
"encoded in nouns"
Lingala
"late binding"
Spanish
Example passage
Tacetin Güntekin war Pro-
fessor. Er war bekannt für
seine Bücher... Güntekin
und seine Partnerin...
si
leader
Una buona
prende un po’ più della
sua parte di colpa
Sarah azali nok... | Scaling Instruction-Finetuned Language Models |
p
e
r
s
a
f
e
t
y
r
e
s
e
a
r
c
h
o
n
s
u
c
h
m
o
d
e
l
s
.
O
v
e
r
a
l
l
,
w
e
b
e
l
i
e
v
e
t
h
a
t
t
h
e
b
e
n
e
| Stanford alpha CRFM |
of-thought reasoning in GPT-3, where users discover a way
to elicit some important new behavior that the developers
had not been aware of.
9.4. LLMs are likely to produce a rapidly growing
array of risks | Eight Things to Know about Large Language Models |
Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan,
Henrique Ponde, Jared Kaplan, Harrison Edwards,
Yura Burda, Nicholas Joseph, Greg Brockman, Alex
Ray, Raul Puri, Gretchen Krueger, Michael Petrov,
Heidy Khlaaf, Girish Sastry, Pamela Mishkin,
Brooke Chan, Scott Gray, Nick Ryder, Mikhail
Pavlov, Alethea Power, Lukasz Kai... | Multi-step Jailbreaking Privacy Attacks on ChatGPT |
Enterprise documents such as forms, invoices, receipts, reports, contracts, and other similar records,
often carry rich semantics at the intersection of textual and spatial modalities. The visual cues offered
by their complex layouts play a crucial role in comprehending these documents effectively. In this
paper, we pr... | DOCLLM |
For text conditioning, a frozen CLIP-text encoder [17]
is employed, and the encoded text prompts are mapped to
various layers of the U-Net using cross-attention. This ap-
proach effectively generalizes to intricate natural language
text prompts, generating high-quality images and depth
maps in a single pass, only havin... | LDM3D- Latent Diffusion Model for 3D |
[67] Chaoyang Wang, Ben Eckart, Simon Lucey, and Orazio Gallo.
Neural trajectory fields for dynamic novel view synthesis.
arXiv preprint arXiv:2105.05994, 2021.
[68] Chaoyang Wang, Xueqian Li, Jhony Kaesemodel Pontes, and
Simon Lucey. Neural prior for trajectory estimation. In Pro-
ceedings of the IEEE/CVF Conference o... | DynIBaR-NeuralDynamicImage-BasedRendering |
During our preliminary experiments, we find that the
shape reconstruction of characters, i.e. the eyes and body,
is satisfactory, while the inferred UV tends to lose detailed
appearances of some small yet significant areas, such as the
nose and ears. We thus adopt a part-sensitive texture reasoner
(PSR) to address the ... | RaBit- Parametric Modeling of 3D Biped Cartoon Characters with a Topological-consistent Dataset |
The project will be around long-term video understanding as specified in the title. The aim is to go beyond
the seconds into the minutes and hours of edited as well as unedited videos (e.g. movies as well as vlogs
and videos from wearable cameras). The exact project will be decided around the student’s interests and ... | long_term_video_understanding23_v2 |
Can Xu, Qingfeng Sun, Kai Zheng, Xiubo Geng, Pu Zhao, Jiazhan Feng, Chongyang Tao, and Daxin
Jiang. Wizardlm: Empowering large language models to follow complex instructions. arXiv
preprint arXiv:2304.12244, 2023.
Xuanyu Zhang and Qing Yang. Self-qa: Unsupervised knowledge guided language model alignment.
arXiv prepr... | Self-AlignmentwithInstructionBacktranslation |
FORDE outperforms alternative PCs. Building on Cor-
reia et al. (2020)’s observation that RFs can be compiled
into probabilistic circuits, we compare the performance of
FORDE to that of five leading PCs on the Twenty Datasets
benchmark (Van Haaren and Davis, 2012), a heterogeneous
collection of tasks ranging from retail... | Adversarial Random Forests for Density Estimation and Generative Modeling |
omit the placeholder S∗ from the prompt. Results are pre-
sented in Figure 9. We first note that NeTI is able to achieve
comparable results to those of DreamBooth, without requir-
ing any tuning of the model. While this does require ad-
ditional training time, our models require ∼2MB of disk
space while DreamBooth requ... | A Neural Space-Time Representation for Text-to-Image Personalization |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.