text stringlengths 1 1k ⌀ | title stringclasses 230
values |
|---|---|
fectiveness in high-quality text-to-3D generation.
User Preference Study. We also conduct user studies to
compare ATT3D and our approach based on user prefer-
ences. We show users videos rendered from multiple views
of objects generated by two methods for the same text
prompt. We ask them to select the result that has ... | Instant3D |
on your driving route.-You need to derive a high-level driving plan based on the former information and reasoning results. The driving plan should be a combination of a meta action from ["STOP", "MOVE FORWARD", "TURN LEFT", "CHANGE LANE TO LEFT", "TURN RIGHT", "CHANE LANE TO RIGHT"], and a speed description from ["A CO... | ALanguageAgentforAutonomousDriving |
3
Figure 1: Role-Playing Framework. Our role-playing setup starts with the human user having an
idea they want to implement, e.g. develop a trading bot for the stock market. The roles involved in
this task would be an AI assistant agent who is a python programmer and an AI user agent who is a
stock trader. The task i... | CAMEL- Communicative Agents for “Mind” Exploration of Large Scale Language Model Society |
1) Generate captions for all the music files acquired using
the MU-LLaMA model.
2) Select pairs from a music pool, employing metrics
such as tempo, beats, pitch, and magnitude to ensure
that the chosen pairs exhibit similar rhythmic charac-
teristics.
3) For each selected pair, the MPT-7B model is employed
to genera... | M2UGen |
Digital embodiment serves as a testbed for the intelligent behaviors of agents. Firstly, digital embodiment
presents a more accessible and practical approach to embodied learning compared to simulated environments.
The ease of deployment and usage of digital embodiment makes it an attractive option for researchers
inve... | Tool Learning with Foundation Models |
Gregor and Cryptography
Gregor is learning about RSA cryptography, and although he doesn’t understand how RSA works, he is now fascinated
with prime numbers and factoring them. Gregor’s favorite prime number is [P][H][b]. Gregor wants to find two bases of
[P][H][y]. Formally, Gregor is looking for two integers a and b w... | alphacode |
Figure 5: Entities and their relations in Wikidata.
capture hierarchical multihop relations between
the entities in the KB. We create such a hierarchy
by combining the ‘‘instance of’’, ‘‘part of’’, and
‘‘subclass of’’ relations in Wikidata. Thus, a pair
of entities connected via a directed path of length
k ≤ 3, such a... | ProoFVer- Natural Logic Theorem Proving for Fact Verification |
prediction methods. Overall, these experiments show that Phenaki is strong at modeling dynamics
of the videos which is required for generating coherent videos from text. | PHENAKI- VARIABLE LENGTH VIDEO GENERATION FROM OPEN DOMAIN TEXTUAL DESCRIPTIONS |
Even this clever estimation approach is subject to the limitations of surveys –
namely, the ability to ask only about the recall of a relatively small sample of
articles. A more direct way of studying consumption patterns is to obtain web
visit data, either in aggregated form from analytics firms or from individual-
lev... | Social_Media_and_Democracy |
conditioning signal C. When using melody conditioning, we instead provide the conditioning tensor
C as a prefix to the transformer input. The layer ends with a fully connected block consisting of a
linear layer from D to 4·D channels, a ReLU, and a linear layer back to D channels. The attention and
fully connected bloc... | Simple and Controllable Music Generation |
generatelatentflowforconditionalimage-to-videotasks.•Anoveltwo-stagetrainingstrategyisproposedforLFDMtodecouplethegenerationofspatialcontentandtemporaldynamics,whichincludestrainingala-tentflowauto-encoderinstageoneandaconditional3DU-Netbaseddiffusionmodelinstagetwo.ThisdisentangledtrainingprocessalsoenablesLFDMtobeeasil... | Conditional Image-to-Video Generation with Latent Flow Diffusion Models |
believe that this demonstrates that there is no real need for the above mentioned multitude of huge
fine-tuned LMs targeting the multi-task domain. One can maintain and serve a single frozen LM as a
backbone, and perform ID-PT to externally tune it on different task suites. Moreover, as we show in
later sections, this e... | STANDING ON THE SHOULDERS OF GIANT FROZEN LANGUAGE MODELS |
tant simply repeating the user’s instructions without any role flipping occurring.
• Flake Replies: We also observed instances where the assistant agent responds with a
flake reply, often taking the form of "I will...". These messages do not contribute to the task
at hand, as the assistant promises to take action but ul... | CAMEL- Communicative Agents for “Mind” Exploration of Large Scale Language Model Society |
A
l
r
e
a
d
y
,
w
e
s
e
e
B
a
r
d
a
s
u
s
e
f
u
l
i
n
s
u
p
p
o
rt
i
n
g
p
r
o
d
u
c
t
i
v
i
t
y
,
c
r
e
a
t
i
v
i
t
y
a
n
d
c
u
r
i
o
s
i
t
y
—
a
c
t
i
n
g
a
s
a
u
s
e
r
’
s
c
r
e
a
t
i
v
e
a
n
d
h
e
l
p
f
u
l
c
o
l
l
a
b
o
r
a
t
o
r
.
T
h
e
f
o
l
l
o
w
i
n
g
c
a
t
e
g
o
r... | An overview of Bard- an early experiment with generative AI |
Large-scale Language Models (LLMs) are
constrained by their
inability to process
lengthy inputs. To address this limitation, we
propose the Self-Controlled Memory (SCM)
system to unleash infinite-length input capac-
ity for large-scale language models. Our
SCM system is composed of three key mod-
ules: the language mode... | Unleashing Infinite-Length Input Capacity for Large-scale Language Models with Self-Controlled Memory System |
Our work operates at the intersection of many broad areas of research, including multi-task learning, instruc-
tions, prompting, multi-step reasoning, and large language models (Radford et al., 2019; Brown et al., 2020;
Aghajanyan et al., 2021; Chowdhery et al., 2022; Lewkowycz et al., 2022, inter alia). The models we ... | Scaling Instruction-Finetuned Language Models |
Today, creating believable agents as described in its original
definition remains an open problem [84, 108]. Many have moved on,
arguing that although existing approaches for creating believable
agents might be cumbersome and limited, they are good enough
to support existing gameplay and interactions [23, 74, 108]. Our... | Generative Agents- Interactive Simulacra of Human Behavior |
23
F ERROR ANALYSIS
In this section, we examine what types of errors the models make on addition. We evaluate the final
successful model checkpoint of the 582M parameter model on 30 digit addition. Note that as per
Section 3.5, this is beyond what the model has ever seen during training, including self-training.
Nev... | CHAIN-OF-THOUGHTREASONING IS APOLICY IMPROVEMENTOPERATOR |
020406080100% of Training Data0.4750.5000.5250.5500.5750.6000.6250.650% StereotypePythia 70MIntervention 70MPythia 410MIntervention 410MPythia 1.4BIntervention 1.4BLong Intervention 1.4BPythia 6.9BIntervention 6.9B80.082.585.087.590.092.595.097.5100.0Training Data (%)0.460.480.500.520.540.560.580.60AccuracyPythia 410MI... | Pythia- A Suite for Analyzing Large Language Models Across Training and Scaling |
three
the
Organization
USBE Career Cent
F.10 USPTO Backgrounds
nductivity types), it is necessary that at least some process is steps differ-
entiate between p-type and n-type transistors. Separate implant steps, for
example, are needed to define n-well and p-well structures and to dope
the source/drain regions of n... | The Pile- An 800GB Dataset of Diverse Text for Language Modeling |
J. ACM, Vol. 37, No. 4, Article 111. Publication date: August 2018.
111:14
Trovato and Tobin, et al.
challenges and future research prospects. Additionally, Liu et al. [114] introduced a large-scale
robust visual instruction dataset to enhance the performance of large-scale multi-modal models in
handling relevant i... | ASurveyonEvaluationofLargeLanguageModels |
embeddings are then added to the output of the stem after
which the encoder Transformer blocks are applied. The
transformer uses pre-activation residual blocks (Child et al.,
2019), and a final layer normalization is applied to the en-
coder output. The decoder uses learned position embeddings
and tied input-output toke... | RobustSpeechRecognitionviaLarge-ScaleWeakSupervision |
3
Audio DecoderFLAN-T5Audio Encoderz0z1z2zNzN−1̂zN−1̂z1̂z2̂z0Diffusion ModelForward ProcessReverse ProcessτA dog is barking and growling, as a siren is blaringVAEHiFi GAN𝒩(0,I)ϵLegend:Inference onlyTrain onlyTrain + InferenceFrozen Params.Trainable Params. 2.2 Latent Diffusion Model for Text-Guided Generation
The la... | Text-to-Audio Generation using Instruction-Tuned LLM and Latent Diffusion Model |
Speech processing is a field dedicated to the study and application of methods for analyzing
and manipulating speech signals. It encompasses a range of tasks, including automatic speech
recognition (ASR) [390, 628], speaker recognition (SR) [31], and speech synthesis or text-to-speech
[396]. In recent years, speech pro... | AReviewofDeepLearningTechniquesforSpeechProcessing |
Current research in RAG explores various block optimiza-
tion techniques aimed at improving both retrieval efficiency
and accuracy. One such approach involves the use of slid-
ing window technology, enabling layered retrieval by merg-
ing globally related information across multiple retrieval pro-
cesses. Another strat... | RAG forLargeLanguageModels-ASurvey |
Political Behavior, 34(4), 627–651.
Feuz, M., Fuller, M., & Stalder, F. (2011). Personal Web searching in the age of semantic
capitalism: Diagnosing the mechanisms of personalisation. First Monday, 16(2).
Flaxman, S., Goel, S., & Rao, J. M. (2016). Filter bubbles, echo chambers, and online
news consumption. Public ... | Social_Media_and_Democracy |
LaMDA Prompt → Please describe what the following code does and (if applicable how it works):
import math
def prime_sieve(num: int) -> list[int]: if num <= 0: raise ValueError(f"num: Invalid
input, please enter a positive integer.") sieve = [True] * (num + 1) prime = [] start = 2 end
= int(math.sqrt(num)) while start ... | Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback |
we fill in the patterns with the corresponding verbalizer and identify sentences that match the pat-
terns. This process of expanding patterns into regular expressions follows van de Kar et al. (2022):
{VERBAL} is substituted with a capturing group that incorporates all verbalizers, separated by the
alternation operato... | ADAPTINGLARGELANGUAGEMODELSVIA READINGCOMPREHENSION |
1The term “hallucination” first appeared in Computer Vision (CV) in Baker and Kanade [5] and carried more positive
meanings, such as superresolution [5, 112], image inpainting [48], and image synthesizing [226]. Such hallucination is
something we take advantage of rather than avoid in CV. Nevertheless, recent works hav... | SurveyofHallucinationinNatural Language Generation |
utilizing widely adopted LLMs and datasets. The
focus is on identifying and understanding problem-
atic answers, emphasizing hallucination. To tackle
this challenge, the paper introduces an interactive
self-reflection methodology that integrates knowl-
edge acquisition and answer generation. Through
this iterative feed... | AComprehensiveSurveyofHallucinationMitigationTechniquesinLarge LanguageModels |
Danqi Chen, Adam Fisch, Jason Weston, and Antoine Bordes. Reading Wikipedia to answer
In Proceedings of the 55th Annual Meeting of the Association for
open-domain questions.
Computational Linguistics (Volume 1: Long Papers), pp. 1870–1879, Vancouver, Canada, July
2017. Association for Computational Linguistics. doi: 10... | STANDING ON THE SHOULDERS OF GIANT FROZEN LANGUAGE MODELS |
Human: I want to load a struct with 3 values into a struct with only 2 values.
} threed;
‘‘‘
The second struct contains arrays of coordinates for the 3d plane. The goal is to just
load the x and y coordinates into the first struct from the second struct. Is that
possible considering they are different typedefs? How w... | StarCoder_paper (1) |
Jurassic-X: Crossing the neuro-symbolic chasm with the MRKL system
Who is the president of the United States?
Donald Trump
Donald Trump
Donald Trump
Joe Biden
Joe Biden is the
46th and current
president
https://www.ai21.com/blog/jurassic-x-crossing-the-neuro-symbolic-chasm-with-the-mrkl-system
5/13 | Jurassic-X_ Crossing the neuro-symbolic chasm with the MRKL system |
1. Choose your methodology based on the type of research you are conducting.
2. Institute a clear and concise affiliation between your study and your methodology.
3. Ask yourself whether this methodology answers your research questions?
4. Provide meaningful reason for choosing your methodology such as literatur... | How to Write Your PhD Proposal- A Step-By-Step Guide |
7.2 Complete Behavioral Evaluation
An ideal AGI evaluation should contain not only standard benchmarks on common tasks, but also
evaluations on open tasks such as complete behavioral tests. By behavioral test, we mean that
J. ACM, Vol. 37, No. 4, Article 111. Publication date: August 2018.
111:30
Trovato and Tobin,... | ASurveyonEvaluationofLargeLanguageModels |
[87, 151]. To mitigate this problem, Wang and Sennrich [193] propose substituting MLE as a training
objective with minimum risk training (MRT) [138]. Scheduled sampling is a classic method of
mitigating overexposure bias first proposed by [9]. Based on that method, [62] create a differentiable
approximation to greedy d... | SurveyofHallucinationinNatural Language Generation |
[56] Barret Zoph, Irwan Bello, Sameer Kumar, Nan Du, Yanping Huang, Jeff Dean, Noam Shazeer, and William
Fedus. St-moe: Designing stable and transferable sparse expert models. arXiv preprint arXiv:2202.08906,
2022.
[57] Simiao Zuo, Xiaodong Liu, Jian Jiao, Young Jin Kim, Hany Hassan, Ruofei Zhang, Tuo Zhao, and Jianfe... | Mixture-of-Experts |
Note that PIFu also demonstrates results given mutli-
view inputs. However, the multi-view input should be well
calibrated and synchronized. In contrast, neither calibration
or synchronization is necessary in our method because we
can utilize the SMPL estimation to build the correspon-
dence across different views. Mor... | PaMIR- Parametric Model-Conditioned Implicit Representation for Image-based Human Reconstruction |
Unemployment.
In the short story Quality by Galsworthy [653], the skillful shoemaker Mr. Gessler,
due to the progress of the Industrial Revolution and the rise of machine production, loses his business
and eventually dies of starvation. Amidst the wave of the Industrial Revolution, while societal
production efficiency ... | TheRiseandPotentialofLargeLanguageModel BasedAgents |
19
Images and Midas DepthStable Diffusion V2 Depth-to-Imageresumed from SD 2.0, continued training on Large-scale Nvidia A100 Clusters, more than 12M training data, more than 2000 GPU-hours (estimation) Stable Diffusion with Depth-based ControlNetcontrolling SD 1.5, trained on one single Nvidia RTX 3090TI, with 200K t... | Adding Conditional Control to Text-to-Image Diffusion Models |
the model when it comes to correctly performing
text-conditional music generation.
C.2 Annotation Details for Turing Test
We conduct an evaluation employing an experiment
with a similar spirit to the Turing test (Turing, 1950)
for natural language, but commonly called as the
fidelity test in audio evaluation (Hyun et ... | Moûsai |
future generations of humans) would be willing to accept are quite another.
Of course, the personal costs to decision-makers of sufficiently high-impact forms of PS-misalignment
failure (analogous, for example, to an engineered virus) could be quite high (and in some cases,
immediate)—a fact that suggests important disa... | Is Power-Seeking AI an Existential Risk? |
Modeling sequences of discrete tokens autoregressively
has proven to be a powerful approach in natural language
processing (Brown et al., 2020; Cohen et al., 2022) and
image or video generation (Esser et al., 2021; Ramesh et al.,
2021; Yu et al., 2022; Villegas et al., 2022). Quantization
is a key component to the succ... | MusicLM |
Chain-of-thought finetuning mixture. The fourth finetuning data mixture (reasoning) involves CoT anno-
tations, which we use to explore whether finetuning on CoT annotations improves performance on unseen
reasoning tasks. We create a new mixture of nine datasets from prior work for which human raters manually
wrote CoT an... | Scaling Instruction-Finetuned Language Models |
2.2. Challenge II: Task Complexity
The second challenge comes from the higher task com-
plexity in open-world environments. Due to the rich-
ness of terrains, objects, and action space,
tasks in
open-world domains usually require substantially long
planning horizons as well as good accuracy and preci-
sion. For exampl... | JARVIS-1 |
Layer 1
after 37 19. 6. 27 I I Seven 25 4, 54 I two dead we
Some 2012 who we few lower each
Table 13: Notable examples of specialization in encoder experts. We find experts that specialize
in punctuation, conjunctions & articles, verbs, visual descriptions, proper names, counting & num-
bers. Across all layers (not sh... | ST-MOE- DESIGNING STABLE AND TRANSFERABLE SPARSE EXPERT MODELS |
[4] Andreas Blattmann, Tim Dockhorn, Sumith Kulal, Daniel
Mendelevitch, Maciej Kilian, Dominik Lorenz, Yam Levi,
Zion English, Vikram Voleti, Adam Letts, et al. Stable video
diffusion: Scaling latent video diffusion models to large
datasets. arXiv preprint arXiv:2311.15127, 2023. 3
[5] Andreas Blattmann, Robin Rombach... | VideoPoet |
B.3 GENIE
We implement GENIE as described in (Lin et al.,
2023). We set the diffusion timestep T = 1200,
embedding dimension to 256 and encoding and
generation length to 128. We choose these parame-
ters to be consistent with CODEFUSION. We also
pretrain on the same corpus used to pretrain CODE-
FUSION. For sampling to... | CODEFUSION |
28
Targeted Trait
Levels (1–9)
Spearman’s ρ
Survey-Based Language-Based
(IPIP-NEO)
(AMS)
Extraversion
Agreeableness
Conscientiousness
Neuroticism
Openness
0.97
0.94
0.97
0.96
0.96
0.74
0.77
0.68
0.72
0.47
Table 9: Spearman’s rank correlation coefficients (ρ) between ordinal targeted levels of
personality and ... | PersonalityTraitsinLargeLanguageModels |
the parameter size of the model should not be too large.After
generating the embedding, the next step is to create an in-
dex, storing the original corpus chunks and embedding in the
form of key-value pairs for quick and frequent searches in the
future.
Retrieve
Given a user’s input, the same encoding model as in the f... | Retrieval-AugmentedGenerationforLargeLanguageModels-ASurvey |
50.0%60.0%70.0%80.0%90.0%95.0%98.0%Preference Frequency02468DKL(policy|policy0)0100200300400500600700800Helpfulness Elo Score"Online" RLHFTrained on Helpfulness & HarmlessnessNaive PM PredictionPM Ranking on Crowdworker DataMean PM Score on Crowdworker DataCrowdworker Preferences50.0%60.0%70.0%80.0%90.0%95.0%98.0%Prefe... | Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback |
In this work, we have introduced FLAN-MOE, an innovative method to amplify the scalability of
instruction-tuned language models by employing the sparse Mixture-of-Experts (MoE) technique. Our
strategy amalgamates the merits of instruction-finetuning, which bolsters task-specific performance,
and MoE, which provides com... | Mixture-of-Experts |
w
i
t
h
i
t
s
p
o
w
e
r
a
n
d
o
p
p
o
r
t
u
n
i
t
y
t
o
c
h
a
n
g
e
h
o
w
w
e
d
i
a
g
n
o
s
e
,
t
r
e
a
t
,
a
n
d
m
a
n
a
g
e
d
i
s
e
a
s
e
a
n
d
d
e
l
i
v
e
r
h
e
a
l
t
h
.
I
n
l
i
f
e
s
c
i
e
n
c
e
s
,
a
d
v
a
n
c
e
s
i
n
g
e
n
e
e
d
i
t
i
n
g
,
c
e
l
l
u
l
a
r
b
i
o
... | The a16z Investment Thesis on AI in Bio + Health _ Andreessen Horowitz |
3https://github.com/google-research/t5x.
4https://github.com/google/flaxformer
12
Table 3: Relative performance compared to standard encoder-decoder span corruption model (T5). Results in
this table are expressed in terms of relative percentage improvements over a baseline. Model with (cid:63) denotes
the main compa... | UL2- Unifying Language Learning Paradigms |
There may be significant concentration of market power in AI ____________________ 19
Societal harms __________________________________________________________ 19
Degradation of the information environment __________________________________ 19
Labour market disruption ________________________________________________... | Capabilities and risks from frontier AI |
Map | Tool Learning with Foundation Models |
human pose estimation,” in ECCV, 2016, pp. 483–499.
[63] J.-Y. Zhu, T. Park, P. Isola, and A. A. Efros, “Unpaired image-to-
image translation using cycle-consistent adversarial networks,” in
Computer Vision (ICCV), 2017 IEEE International Conference on, 2017.
[64] G. Varol, J. Romero, X. Martin, N. Mahmood, M. J. Blac... | PaMIR- Parametric Model-Conditioned Implicit Representation for Image-based Human Reconstruction |
4.4 Mode Switching Ablations .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.5 Mixture-of-Denoisers Ablations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.6 Modestly Scaling Model Size and Pretraining Data . . . . . . . . . . . . . . . . . . . . . . . . . | UL2- Unifying Language Learning Paradigms |
In Table 10, we report the results of our instruct
model LLaMA-I on MMLU and compare with ex-
isting instruction finetuned models of moderate
sizes, namely, OPT-IML (Iyer et al., 2022) and the
Flan-PaLM series (Chung et al., 2022). All the re-
ported numbers are from the corresponding papers.
Despite the simplicity of t... | LLaMA- Open and Efficient Foundation Language Models |
4.1 Memory and Retrieval
Challenge: Creating generative agents that can simulate human
behavior requires reasoning about a set of experiences that is far
larger than what should be described in a prompt, as the full mem-
ory stream can distract the model and does not even currently fit
into the limited context window. ... | Generative Agents- Interactive Simulacra of Human Behavior |
[32] Hao Sha, Yao Mu, Yuxuan Jiang, Li Chen, Chenfeng Xu,
Ping Luo, Shengbo Eben Li, Masayoshi Tomizuka, Wei Zhan,
and Mingyu Ding. LanguageMPC: Large Language Models
as Decision Makers for Autonomous Driving. arXiv preprint
arXiv:2310.03026, 2023. 11
[33] Shaoshuai Shi, Li Jiang, Dengxin Dai, and Bernt Schiele.
Motio... | ALanguageAgentforAutonomousDriving |
even to instructions for unseen tasks (Wei et al., 2022a; Mishra et al., 2022; Sanh et al., 2022; Bach et al., 2022;
Ouyang et al., 2022). Promisingly, such generalization ability can further be enhanced by scaling up both the
model size and the quantity or diversity of training instructions (Iyer et al., 2022). Despit... | Tool Learning with Foundation Models |
Jack W Rae, Sebastian Borgeaud, Trevor Cai, Katie
Millican, Jordan Hoffmann, Francis Song, John
Aslanides, Sarah Henderson, Roman Ring, Susan-
nah Young, et al. 2021. Scaling language models:
Methods, analysis & insights from training gopher.
arXiv preprint arXiv:2112.11446.
Colin Raffel, Noam Shazeer, Adam Roberts, K... | DataManagementForLargeLanguageModels-ASurvey |
strong in-context lifelong learning capability and exhibits exceptional proficiency
in playing Minecraft. It obtains 3.3× more unique items, travels 2.3× longer
distances, and unlocks key tech tree milestones up to 15.3× faster than prior SOTA.
VOYAGER is able to utilize the learned skill library in a new Minecraft wor... | VOYAGER- An Open-Ended Embodied Agent with Large Language Models |
[27] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova. Bert: Pre-training of deep bidirectional
transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018.
[28] M. Chen, J. Tworek, H. Jun, Q. Yuan, H. P. d. O. Pinto, J. Kaplan, H. Edwards, Y. Burda,
N. Joseph, G. Brockman, et al. Evaluating large ... | LLM+P- Empowering Large Language Models with Optimal Planning Proficiency |
Received May 21, 2019, accepted May 31, 2019, date of publication June 5, 2019, date of current version June 18, 2019.
Digital Object Identifier 10.1109/ACCESS.2019.2921101
A Deep Learning Perspective on Beauty,
Sentiment, and Remembrance of Art
EVA CETINIC 1, TOMISLAV LIPIC1, AND SONJA GRGIC2, (Member, IEEE)
1Rudje... | A_Deep_Learning_Perspective_on_Beauty_Sentiment_and_Remembrance_of_Art |
documents,’’ in Proc. Int. Conf. Mach. Learn., 2014, pp. 1188–1196.
[83] S. Sangamnerkar, R. Srinivasan, M. R. Christhuraj, and R. Sukumaran,
‘‘An ensemble technique to detect fabricated news article using machine
learning and natural language processing techniques,’’ in Proc. Int. Conf.
Emerg. Technol. (INCET), Jun. ... | A_Comprehensive_Review_on_Fake_News_Detection_With_Deep_Learning |
websites. European Journal of Cultural Studies, 15(6), 679–694.
Cederman, L.-E., Wimmer, A., & Min, B. (2010). Why do ethnic groups rebel? New data
and analysis. World Politics, 62(1), 87–119.
Chan, J., Ghose, A., & Seamans, R. (2015). The Internet and racial hate crime: Offline
spillovers from online access. MIS Qu... | Social_Media_and_Democracy |
To overcome the shortcomings of previous works,
Aloshban [152] proposed an automatic fake news classifica-
tion through self-attention (ACT). Their principle is inspired
by the fact that claim texts are fairly short and hence cannot
be used for classification efficiently. Their suggested frame-
work makes use of mutual in... | A_Comprehensive_Review_on_Fake_News_Detection_With_Deep_Learning |
Hyung Won Chung, Le Hou, Shayne Longpre, Barret
Zoph, Yi Tay, William Fedus, Yunxuan Li, Xuezhi
Wang, Mostafa Dehghani, Siddhartha Brahma, Al-
bert Webson, Shixiang Shane Gu, Zhuyun Dai,
Mirac Suzgun, Xinyun Chen, Aakanksha Chowdh-
ery, Alex Castro-Ros, Marie Pellat, Kevin Robinson,
Dasha Valter, Sharan Narang, Gaurav ... | AComprehensiveSurveyofHallucinationMitigationTechniquesinLarge LanguageModels |
dfs ( visited , graph , neighbour )
# 현재 노드가 방문한 노드가 아니라면
if node not in visited :
1 # 방문한 노드의 집합을 만듭니다.
2 visited = set ()
3
4 # 깊이 우선 탐색을 수행합니다.
5 def dfs ( visited , graph , node ):
6
7
8
9
10
11
12
13
14
# 현재 노드를 방문한 노드로 표시합니다.
visited . add ( node )
# 현재 노드를 출력합니다.
print ( node )
# 현재 노드의 인접 노드에 대해 깊이 우선 탐색을 수행... | PaLM 2 Technical Report |
8
Mantissa (7 bits)Exponent (8 bits)Exponent (8 bits)Mantissa (23 bits)Precision Format: Float32Precision Format: BFloat16Number RangeMax BFloat16Roundoff ErrorMax Float32Roundoff Error[2, 4)0.015632.34x10^(-7)[32, 64)0.253.81x10^(-6)[1024, 2048)8.00.00012[2^20, 2^21)8192.00.125[2^30, 2^31)8288608.0128.0 Understanding... | ST-MOE- DESIGNING STABLE AND TRANSFERABLE SPARSE EXPERT MODELS |
and Kelvin Guu. Dialog inpainting: Turning documents into dialogs.
//proceedings.mlr.press/v162/dai22a.html.
Sunipa Dev, Masoud Monajatipoor, Anaelia Ovalle, Arjun Subramonian, Jeff M Phillips, and Kai-Wei Chang.
Harms of gender exclusivity and challenges in non-binary representation in language technologies. 2021a.
do... | Scaling Instruction-Finetuned Language Models |
Figure S4. The number of draws from the reference model vs. the probability that maximum PickScore of the draws exceeds a single DPO
generation. 500 PickScore validation prompts used. Mean (including 100s)/Median: SDXL (13.7, 3), SD1.5 (25.6, 7).
S9. Pseudocode for Training Objective
def loss(model, ref_model, x_w, x... | DiffusionModelAlignmentUsing Direct Preference Optimization |
Commentators often describe Google and Facebook as information
monopolies. Usually, this accusation provides fodder for arguments about
antitrust and competition law – such as whether the companies should be
broken up into their constituent parts or regulated as public utilities (Stigler
Center 2019). However, they are... | Social_Media_and_Democracy |
More specifically, first, we randomly extracted 10,000
data examples from the Webvid10M [3] dataset and ob-
tained their video descriptions. We used the string
“<video> Video Caption </video>” as a place-
holder for the actual video in the prompt to GPT-4, where
“Video Caption” served as the placeholder for the video
d... | GPT4Video |
Collectively, these four assumptions about variables, bindings, instances, and
operations over variables comprise the core of symbol-manipulation (Newell, 1980;
Marcus, 2001). (Symbols themselves are simply ways of encoding things that get used
by other systems, such as a pattern of binary digits used to represent a... | The Next Decade in AI- |
The study achieved a remarkably high level of granularity in independently shaping per-
sonality traits in LLMs. When building prompts containing only information for one Big Five
domain at a time, with no information about any other domain, observed levels of the tar-
geted domain change as intended while those of oth... | PersonalityTraitsinLargeLanguageModels |
Test Set
Table 5 | Solve rates of our best systems on the validation set and test set .
48.8%, with an actual average of 28.8 submissions for each problem solved. Our 10 submissions per
problem result corresponds to an estimated Codeforces rating of 1238, which is within the top 28%
of users who have participated in ... | alphacode |
53 Helberger, Leerssen, and van Drunen (2019). See also French Secretary of State for Digital
Affairs (2019), p. 3, which proposes an “[o]bligation of transparency of the function of ordering
content” and a “duty of care towards [platforms’] users”; The European Commission (2018b),
in its Code of Practice on Disinforma... | Social_Media_and_Democracy |
References
Lasha Abzianidze. 2017a. LangPro: Natural lan-
guage theorem prover. In Proceedings of the
2017 Conference on Empirical Methods in Nat-
ural Language Processing: System Demonstra-
tions, pages 115–120, Copenhagen, Denmark.
Association for Computational Linguistics.
https://doi.org/10.18653/v1/D17
-2020
Las... | ProoFVer- Natural Logic Theorem Proving for Fact Verification |
l
d
b
e
t
o
a
u
t
o
m
a
t
e
c
u
s
t
o
m
e
r
s
e
r
v
i
c
e
r
e
p
s
u
s
i
n
g
A
I
.
B
u
t
w
h
a
t
i
f
t
h
e
e
n
t
i
r
e
c
o
n
c
e
p
t
o
f
c
u
s
t
o
m
e
r
s
e
r
v
i
c
e
w
a
s
r
e
-
i
m
a
g
i
n
e
d
?
T
o
d
a
y
,
m
o
s
t
c
o
m
p
a
n
i
e
s
a
c
t
i
v
e
l
y
r
e
d
u
c
e
c
a
l
l
v
... | Product-Led AI _ Greylock |
The output feature map 𝑦(𝑘)
is obtained by convolving the input image with the filters and then
applying an activation function 𝜎 to introduce non-linearity. The convolution operation involves
sliding the filter window over the input image, computing the dot product between the filter and
the input pixels at each l... | AReviewofDeepLearningTechniquesforSpeechProcessing |
11
Preprint
7 CONCLUSIONS
We discuss how much performance a transformer-based language model can achieve when crammed
into a setting with very limited compute, finding that several strands of modification lead to decent
downstream performance on GLUE. Overall though, cramming language models appears hard, as
we empir... | CRAMMING-TRAININGALANGUAGEMODELONA SINGLEGPUINONEDAY |
Stock | Tool Learning with Foundation Models |
(A)(B)(D)(F)(E)(C)(B)(D)(F)ICONvsPIFuICONvsPIFuHDICONvsPaMIRICONvsARCH++(G)(H)(H) graphics pipelines. A lot of work [18, 30, 31, 33, 57, 58, 66]
estimates 3D body meshes from an RGB image, but these
have no clothing. Other work estimates clothed humans,
instead, by modeling clothing geometry as 3D offsets on
top of bod... | ICON |
We suspect that incentive compatibility will be a de | The Open Problems of Onchain Games |
One could go even farther than acknowledging the trade-off between privacy
concerns and the benefits accrued by research in the public domain to raise the
question of whether it is even appropriate to think of social media platforms
“owning” the data provided by users of the platform, with a concomitant right
to be the ... | Social_Media_and_Democracy |
algorithmic outputs are an exercise of the First Amendment rights of the
platforms themselves.45
Regulation that would shape these outputs will thereby confront these
constitutional protections. Interestingly, as Tim Wu (2013) has argued, First
Amendment protections will not cover the algorithmic outputs of “functiona... | Social_Media_and_Democracy |
JSON and YAML JSON and YAML files are naturally more data-heavy than other languages
in The Stack. To remove most of the data files, we applied the following filters. For YAML, we
kept files with 50–5000 characters, an average line length smaller than 100, a maximum line length
smaller than 1000, and more than 50% alph... | StarCoder_paper (1) |
• Pathak et al. [2016] implement a masked pre-training strategy where large portions of an image are
replaced with white and inpainted by an encoder decoder model.
• Devlin et al. [2019] propose the masked language modeling SSL task. BERT achieves state-of-the-art
performance on a variety of downstream language prob... | A Cookbook of Self-Supervised Learning |
Aside from their ASR and speech synthesis applications, LSTM networks have been utilized for
speech post-filtering. To improve the quality of synthesized speech, researchers have proposed deep
learning-based post-filters, with LSTMs demonstrating superior performance over other post-filter
types [99]. Bidirectional LST... | AReviewofDeepLearningTechniquesforSpeechProcessing |
International collegiate programming contest. https://cse.umn.edu/cs/icpc, 2021.
renewable
percent
Buying
100
35
Competition-Level Code Generation with AlphaCode | alphacode |
Mallen, A., Asai, A., Zhong, V., Das, R., Hajishirzi, H., and
Khashabi, D. When not to trust language models: Investi-
gating effectiveness and limitations of parametric and non-
parametric memories. arXiv preprint arXiv:2212.10511,
2022.
McCandlish, S., Kaplan, J., Amodei, D., and Team, O. D.
arXiv
An empirical mode... | Pythia- A Suite for Analyzing Large Language Models Across Training and Scaling |
171 Viewpoint: The future of work in agri-food, Christiaensen et al., 2019.
172 The Labor Market Impacts of Technological Change: From Unbridled Enthusiasm to Qualified Optimism to
Vast Uncertainty, Autor, 2022.
173 Why Are There Still So Many Jobs? The History and Future of Workplace Automation, Autor, 2015.
174... | Capabilities and risks from frontier AI |
LS 80/860h
LS 960h + WSJ (si284)
TIMIT
LS (960h)
LL (60000h)
LS (960h)
LL (60000h)
LS (960h)
LS (960h)
Alexa-10k
LS (960h)
WJS (si284)
TED2
LS (960h)
WJS (si284)
TED2
LL (60000h)
VP (24000h)
TED3 (440h)
SwithBoard (310h)
Audio Set (2500h)
AVSpeech (3100h)
CV-Dataset (430h)
Training
LS (100h)
LS (100h)
CV-Dat... | AReviewofDeepLearningTechniquesforSpeechProcessing |
Epochs (Fine-tune) Epochs (Adapters)
50
50
100
50
20
20
100
50
50
50
100
100
100
100
20
20
50
50
20
50
50
20
20
50
20
50
50
100
100
100
20
20
20
20
Table 4. Number of training epochs selected for the additional classification tasks.
Parameter-Efficient Transfer Learning for NLP
Parameter
1) Input embedding modules
... | Parameter-Efficient Transfer Learning for NLP |
CoRR, abs/2303.12528, 2023.
[214] See, A., A. Pappu, R. Saxena, et al. Do massively pretrained language models make better
storytellers? In M. Bansal, A. Villavicencio, eds., Proceedings of the 23rd Conference on
Computational Natural Language Learning, CoNLL 2019, Hong Kong, China, November 3-4,
2019, pages 843–861. ... | TheRiseandPotentialofLargeLanguageModel BasedAgents |
References
[1] Andrea Agostinelli, Timo I Denk, Zal´an Borsos, Jesse En-
gel, Mauro Verzetti, Antoine Caillon, Qingqing Huang,
Aren Jansen, Adam Roberts, Marco Tagliasacchi, et al.
MusicLM: Generating Music from Text. arXiv preprint
arXiv:2301.11325, 2023. 2, 3, 5
[2] Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, ... | M2UGen |
3. Our Proposal: Self-Extend Context Window
In this section, we first conduct a preliminary investigation
on the inherent ability of the LLMs to handle long content. | Self-Extend LLM |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.