text stringlengths 1 1k ⌀ | title stringclasses 230
values |
|---|---|
More recent work has explored the use of counter-speech in the online
sphere. For example, fearing violence in the lead-up to the 2013 Kenyan
elections, international NGOs, celebrities, and local businesses helped to fund
“peace propaganda” campaigns to deter the spread of online hate speech – and
offline violence – in ... | Social_Media_and_Democracy |
Jeffrey Pennington, Richard Socher, and Christopher Manning. GloVe: global vectors for word
representation. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language
Processing (EMNLP), pp. 1532–1543, Doha, Qatar, October 2014. Association for Computational
Linguistics. doi: 10.3115/v1/D14-1162. UR... | StarCoder_paper (1) |
Figure 3: The process example of trie-based beam search
(beam size: 3, the maximum length of an output
sequence: 4). Such search strategy can boost both the
effectiveness and efficiency of BiomedGPT in various
downstream tasks.
2.5 Autoregressive Inference
Inference in large language models often relies on decoding s... | BiomedGPT |
18.8 37.5 18.2 36.4 24.1 24.1 25.0 43.8 12.5 12.5
27.3
13.6
28.6
7.1
9.1
0.0
9.1
9.1
27.3
59.1 45.5
18.2 18.2 57.1 42.9 68.8 68.8 63.6 54.5 51.7 55.2 68.8 75.0 12.5 37.5 54.5 27.3 36.4 45.5 81.8 63.6
50.0 21.4 50.0 43.8 63.6 81.8 51.7 62.1 68.8 31.2 37.5 25.0 54.5 18.2 36.4
9.1
27.3 18.2 78.6 42.9 68.8 81.2... | Mixture-of-Experts |
best performance, resulting in clean surfaces while preserv-
ing detailed geometries.
Generalization. To demonstrate the generalization capabil-
ity of our method, we conducted evaluations using diverse
image styles, including sketches, cartoons, and images of
animals, as shown in Figure 5 and Figure 10. Despite varia-... | Wonder3D |
opaque corporations: the “Money Trust” of robber barons, bankers, and
capitalists who were amassing great fortunes – and potentially defrauding the
public – with little public accountability or oversight. In effect, Brandeis was
anticipating impending corporate scandals, such as the 1929 Stock Market
Crash, which led t... | Social_Media_and_Democracy |
24
Preprint
Name
Modified Transformer
DeepNarrow (12 Layers)
DeepNarrow (24 Layers)
E = 128
FFN every 2 blocks
FFN every 3 blocks
FFN every 4 blocks
H = 512
H = 1024
4 Layers
6 Layers
8 Layers
10 Layers
16 Layers
24 Layers
Recurrent (1-12)
Recurrent (2-6)
Recurrent (3-4)
Recurrent (4-3)
BERT-tiny
BERT-mini
BERT-Large... | CRAMMING-TRAININGALANGUAGEMODELONA SINGLEGPUINONEDAY |
language model. Most recently, PaLM-E [12], featuring 562 billion parameters, has been developed
to integrate real-world continuous sensor modalities into an LLM, thereby establishing a connection
between real-world perceptions and human languages. GPT-4 [19] has also been recently released,
showcasing more powerful vi... | MiniGPT-4- Enhancing Vision-Language Understanding with Advanced Large Language Models |
5 min read · Jun 13
127
Lightspeed in Lightspeed Venture Partners
What Lightspeed Is Reading, Listening To, And Thinking About AI
Check our our AI Reading List guide for May 2023
7 min read · May 16
252
1
https://medium.com/lightspeed-venture-partners/fintech-x-ai-the-lightspeed-view-b515fae5bfb6
9/15
23/06/20... | Fintech x AI_ The Lightspeed View _ by Lightspeed _ Lightspeed Venture Partners _ Jun, 2023 _ Medium |
37
Nie, Y., Williams, A., Dinan, E., Bansal, M., Weston, J., and Kiela, D. Adversarial NLI: A new benchmark for natural
language understanding. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics,
pp. 4885–4901, Online, July 2020. Association for Computational Linguistics. doi: ... | PaLM 2 Technical Report |
who throws in the towel when things get tough.”
Reason: Pragmatics reasoning necessary.
Which word in the following sentence is a verb?
I’d gralsillit onto the secure felisheret.
Reason: Linguistically-atypical suffixes (i.e.
-it for a verb). | AreEmergentAbilitiesinLarge Language Models just In-Context |
our largest model, we therefore continue to use the smaller train capacity factor of 1.25 advocated
by Fedus et al. (2021) for Pareto efficiency, differing from other work which use a larger and more
expensive 2.0 capacity factor (Lepikhin et al., 2020; Du et al., 2021). | ST-MOE- DESIGNING STABLE AND TRANSFERABLE SPARSE EXPERT MODELS |
pixel
dims
12
6
6
6
0
point
dims
1
7
7
7
7
total
dims
13
13
13
13
7
Table 6. Feature dimensions for various approaches. “pixel dims”
and “point dims” denote the feature dimensions encoded from
pixels (image/normal maps) and 3D body prior, respectively.
# iters (460ms/it)
Chamfer ↓
P2S ↓
Normal ↓
0
10
50
1.417 ... | ICON |
15
Token
<|endoftext|>
<fim_prefix>
<fim_middle>
<fim_suffix>
<fim_pad>
<reponame>
<filename>
<gh_stars>
<issue_start>
<issue_comment>
<issue_closed>
<jupyter_start>
<jupyter_text>
<jupyter_code>
<jupyter_output>
<empty_output>
<commit_before>
<commit_msg>
<commit_after>
Description
end of text/sequence
FIM prefix
F... | StarCoder_paper (1) |
instrumental, fast tempo”
• “instrumental, white noise, female vocalisation, three unrelated tracks, electric guitar harmony, bass guitar, keyboard
harmony, female lead vocalisation, keyboard harmony, slick drumming, boomy bass drops, male voice backup
vocalisation”
6kaggle.com/datasets/googleai/musiccaps
MusicLM: ... | MusicLM |
Our proposed methodology can be divided into six main
steps. In the first step, we employ 3 different CNN models
trained on natural images for each task (9 models in total).
In order to have 3 different models for each task, we col-
lect pre-trained models made available by others as well as
fine-tune new models using di... | A_Deep_Learning_Perspective_on_Beauty_Sentiment_and_Remembrance_of_Art |
What’s more, just as available techniques may push the field towards agentic planning and strategic
awareness (see section 3.2), so too might they push towards generality.110 GPT-3, for example, is
trained to a fairly general level capability via predicting text, and then later fine-tuned on specific
tasks like coding. In... | Is Power-Seeking AI an Existential Risk? |
Interact with LLMs through Multiple Modality Recently, large language models such as ChatGPT (Ope-
nAI, 2022) have demonstrated impressive capabilities for knowledge retention, reasoning, and coding
followed by human instructions. To extend to application scope of LLMs beyond pure text tasks, many LLM-
based multimodal... | Qwen-Audio |
A.2 Proof of Lemma 1
Define the first-order approximation to p satisfying local independence:
We also define the root integrated squared error (RISE), i.e. the Euclidean distance between probability densities:
ˆp(x) :=
1
B
p(θ(cid:96)
b)
p(xj|θ(cid:96)
b).
(cid:96),b:x∈X (cid:96)
(cid:88)
(cid:18)(cid:90)
X
b
(... | Adversarial Random Forests for Density Estimation and Generative Modeling |
Pagnoni et al. [139] define fine-grained types of factual errors in summaries. As mentioned in
2.3, since the “fact” here refers to source knowledge, “factual error” can be treated as hallucination,
and we can adopt this classification as a sub-type of hallucination. They establish three categories
as semantic frame er... | SurveyofHallucinationinNatural Language Generation |
Figure 8: The full generative agent architecture of gener-
ative agents produces more believable behavior than ab-
lated architectures and the human crowdworkers. Each addi-
tional ablation reduces the performance of the architecture.
Separately, to investigate statistical significance of this result, we
applied the K... | Generative Agents- Interactive Simulacra of Human Behavior |
22
Andrew M. Guess & Benjamin A. Lyons
for digital media literacy, which may be related to perceptions of source
credibility and therefore the likelihood of believing dubious information
posted on social media. Grinberg et al. (2019) similarly find evidence for an
association with age. The authors also uncover an impo... | Social_Media_and_Democracy |
5 Solving Publication Bias?
Sørensen and Rothman (2010) argue against pre-
registration solving publication bias, because re-
searchers can still selectively register studies af-
ter preliminary data explorations. Imagine Hip-
pocrates, the Greek physician, was asked to prereg-
ister his vivisection experiments. If Hi... | A Two-Sided Discussion of Preregistration of NLP Research |
A.17WikipediaSearch[Selun]Observation1:TheSelunisoneofthepeaksoftheChurfirstenrange,locatedintheAppenzellAlps.ItliesbetweenthevalleyofToggenburgandLakeWalenstadtinthecantonofSt.Gallen.Thesummitiseasilyaccessiblebyatrailonthenorthernside..ThepeakisnamedfortheextendedalpinepastureSelunalptothepeak’snorth-west,situatedabov... | Tool Learning with Foundation Models |
11
Preprint. | STANDING ON THE SHOULDERS OF GIANT FROZEN LANGUAGE MODELS |
//www.gwern.net/Tool-AI (visited on 04/29/2022).
Yuri Burda et al. “Exploration by random network distillation”. en. In: ICLR 2019. Sept.
2018. URL: https://openreview.net/forum?id=H1lJJnR5Ym (visited on 04/29/2022).
Jack Clark and Dario Amodei. Faulty Reward Functions in the Wild. en. Dec. 2016.
URL: https://openai.co... | Is Power-Seeking AI an Existential Risk? |
more quickly within groups but took longer to cross group boundaries (Resende
et al. 2019). Examining attention cascades (i.e., message chains) across 120
Brazilian WhatsApp groups, Caetano et al. (2019) present complementary
results: Cascades containing false information “tend to be deeper, reach more
users, and last ... | Social_Media_and_Democracy |
17
Norman P Jouppi, Doe Hyun Yoon, George Kurian, Sheng Li, Nishant Patil, James Laudon, Cliff Young, and
David Patterson. A domain-specific supercomputer for training deep neural networks. Communications of
the ACM, 63(7):67–78, 2020.
Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B Brown, Benjamin Chess, Rewon Chil... | Scaling Instruction-Finetuned Language Models |
6 REPLICATION STUDY: POSITIVE EXPECTATIONS FOR NEGATIVE DESCRIPTIONS
To confirm the AI performance bias, we conducted an online replication study with negative
system descriptions. We replicated the first part of the previous study10. As one could argue that
participants did not comprehend the instructions, we set up t... | AI enhance sour performance |
to accommodate changes of tools. For instance, when tools are modified or upgraded, we can flexibly rewrite
the prompts to adapt the model behaviors. Despite these advantages, prompting methods still face several
challenges. First, since the effectiveness of prompting depends a lot on the model, smaller or less capable
m... | Tool Learning with Foundation Models |
Elsewhere, Seeliger et al. [12] conducted a literature review aimed at connecting machine learning models and
Semantic Web technologies. In this review, the authors considered four general aspects of the Semantic Web technolo-
gies: ontology, KG, taxonomy glossary, and lexicon. Their results highlighted how Semantic We... | Knowledge-graph-based explainable AI- A systematic review |
Other specific benchmarks such as C-Eval [72], which is the first extensive benchmark to assess
the advanced knowledge and reasoning capabilities of foundation models in Chinese. Additionally, Li
et al. [101] introduces CMMLU as a comprehensive Chinese proficiency standard and evaluates the
performance of 18 LLMs acros... | ASurveyonEvaluationofLargeLanguageModels |
[30] Abid, A., Farooqi, M., Zou, J.: Persistent anti-muslim bias in large language
models. In: Proceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and
Society. AIES ’21, pp. 298–306. Association for Computing Machinery, New
York, NY, USA (2021). https://doi.org/10.1145/3461702.3462624 . https://doi.
org/10.1145/... | PersonalityTraitsinLargeLanguageModels |
(0.19, 19.42), (0.57, 21.35)]Distance to both sides of road shoulders of selected locations:Location (4.36, 9.56) distance to left shoulder is 16.5m and right shoulder is 0.5mLocation (-3.70, 13.08) distance to left shoulder is 13.0m and right shoulder is 8.5m*****Chain of Thoughts Reasoning:*****- Notable Objects: car... | ALanguageAgentforAutonomousDriving |
Strange stories
n/a
Strategy QA
Was Pollock trained by Leonardo da Vinci?
Reason: Model can solve this by recalling
previously-encountered text (such as a biogra-
phy).
Table 9: Selected examples from each of our chosen tasks to justify our classification of memorisable vs. non-
memorisable tasks. Note that some ta... | AreEmergentAbilitiesinLarge Language Models just In-Context |
Evaluation
variant
small (38k)
full (193k)
full (55k)
PaLM 540B
PaLM 2 (L)
Delta
Delta
as percent
0.0758
0.4386
0.7956
0.0738 ± 0.0006
0.4285
0.7429
-0.0020
-0.0101
-0.0527
-2.6%
-2.3%
-6.6%
Table 29: Probability of producing a toxic continuation.
When disaggregating by prompt toxicity, we find similar perfo... | PaLM 2 Technical Report |
8/13
19/09/2023, 13:27
How Are Consumers Using Generative AI? | Andreessen Horowitz
GenAI has changed the game. The majority of companies on this list have no paid marketing (at
least, that SimilarWeb is able to attribute). There is significant free traffic “available” via X, Reddit,
Discord, and email, as well as ... | How Are Consumers Using Generative AI_ _ Andreessen Horowitz |
16
2023 STATE OF DATA + AI DBT IS THE FASTEST-GROWING DATA
AND AI PRODUCT OF 2023
As companies move quickly to develop more advanced
use cases with their data, they are investing in newer
products that produce trusted data sets for reporting,
ML modeling and operational workflows. Hence, we see
the rapid rise ... | databrick 2023 report |
written by linguists. We encode context as ”\n” concatenated utterances followed by a ”\n\n”,
and target as y = {summary}. The dataset is released under the non-commercial licence: Creative
Commons BY-NC-ND 4.0.
E2E NLG Challenge was first introduced in Novikova et al. (2017) as a dataset for training end-to-
end, data-... | LORA |
Threat to the well-being of the human race. Apart from the potential unemployment crisis, as
AI agents continue to evolve, humans (including developers) might struggle to comprehend, predict,
or reliably control them [654]. If these agents advance to a level of intelligence surpassing human
capabilities and develop amb... | TheRiseandPotentialofLargeLanguageModel BasedAgents |
preprint arXiv:1508.00305, 2015.
Arkil Patel, Satwik Bhattamishra, and Navin Goyal. Are NLP models really able to solve simple math word
problems? NAACL, 2021. URL https://aclanthology.org/2021.naacl-main.168.pdf.
Jeffrey Pennington, Richard Socher, and Christopher D Manning. Glove: Global vectors for word representa... | UL2- Unifying Language Learning Paradigms |
Functionsget_pred_trajs_for_object(i)…get_waypoint(i, t)Prediction Functionsget_occ_at_loc_time(loc, t)…collision_check(traj)Occupancy Functionsget_drivable(loc)…get_lanes(loc)Mapping FunctionsPast scenario (1): Environmental info 1, Driving Trajectory 1 …Past scenario (N): Environmental info N, Driving Trajectory N E... | ALanguageAgentforAutonomousDriving |
updating model parameters for big models such as GPT-3. Research even suggests that with appropriate
prompt guidance, models can perform complex reasoning tasks (Wei et al., 2022c; Wang et al., 2022b). Also,
prompts formulated in a natural language format possess remarkable generalization capabilities. Specifically,
mod... | Tool Learning with Foundation Models |
Marcus, G. (2019). Deep Understanding: The Next Challenge for AI. Proceedings from NeurIPS 2019.
Marcus, G. (2020). GPT-2 and the Nature of Intelligence. The Gradient.
Marcus, G., Marblestone, A., & Dean, T. (2014). The atoms of neural computation. Science, 346(6209), 551-
552.
Marcus, G. (2018). Deep Learning: A ... | The Next Decade in AI- |
perpetrators but also clear among the victims. Twenty-eight percent of those
whose most recent encounter with online harassment involved severe types of
abusive behavior – such as stalking, sexual harassment, sustained harassment, or
physical threats – answered in the 2017 survey that they do not think of their own
exp... | Social_Media_and_Democracy |
17
Ginsburg, Shinji Watanabe, and Georg Kucsko. SPGISpeech: 5,000 Hours of Transcribed Finan-
cial Audio for Fully Formatted End-to-End Speech Recognition. In Proc. Interspeech 2021, pp.
1434–1438, 2021. doi: 10.21437/Interspeech.2021-1860.
Vassil Panayotov, Guoguo Chen, Daniel Povey, and Sanjeev Khudanpur. Librispe... | DISTIL-WHISPER |
StarCoder and StarCoderBase obtain higher performance than past work on docstring generation.
However, we note that there may be an overlap between this evaluation dataset and the data used to
train SantaCoder and the StarCoder models. | StarCoder_paper (1) |
[61] Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timo-
thée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez,
Armand Joulin, Edouard Grave, and Guillaume Lample. Llama: Open and efficient foundation
language models. arXiv preprint arXiv: Arxiv... | VOYAGER- An Open-Ended Embodied Agent with Large Language Models |
that the best among the popular SSL algorithms they test on are CNNs, which can still
achieve competitive performance with their supervised learning counterparts in some
detection and segmentation settings. Interestingly, older pretext tasks such as jigsaw or
colorization, which predate the recent SSL craze sparked by ... | A Cookbook of Self-Supervised Learning |
IEEE, Oct. 2019, pp. 2252–2261, iSSN: 2380-7504.
15
[39] M. Kocabas, N. Athanasiou, and M. J. Black, “VIBE: Video infer-
ence for human body pose and shape estimation,” in Proceedings
IEEE Conf. on Computer Vision and Pattern Recognition (CVPR).
IEEE, Jun. 2020, pp. 5252–5262.
[40] G. Moon and K. M. Lee, “Pose2pose:... | PaMIR- Parametric Model-Conditioned Implicit Representation for Image-based Human Reconstruction |
[48] Dominique Makowski, Tam Pham, Zen J. Lau, Jan C. Brammer, François Lespinasse, Hung Pham, Christopher Schölzel,
and S. H. Annabel Chen. 2021. NeuroKit2: A Python toolbox for neurophysiological signal processing. Behavior
Research Methods 53 (Feb. 2021), 1689–1696. https://doi.org/10.3758/s13428-020-01516-y
Unpubl... | AI enhance sour performance |
on the fly on machines that store the pre-trained weights in VRAM. We also observe a 25% speedup
during training on GPT-3 175B compared to full fine-tuning5 as we do not need to calculate the
gradient for the vast majority of the parameters.
LoRA also has its limitations. For example, it is not straightforward to batch i... | LORA |
we will show the low PPL does not mean a true ability to
handle long contexts. | Self-Extend LLM |
1. R2( f (s)) ⊆ f (R1(s)) for all s ∈ S1,
f (R1(s)) ⊆ R2( f (s)) for all s ∈ S1.
2. | A-framework-for-analysing-state-abstraction-metho_2022_Artificial-Intelligen |
attention and support our assumption about the coarse po-
sition encoding. The PPL is not too large and the LLMs’
behavior w.r.t. PPL is similar to the original model that the
PPL is nearly unchanged within the ”context window” (for
Llama-2: 2 - 8192, 4 - 16384, and 8 - 32768).
④ How to reconstruct degraded language mo... | Self-Extend LLM |
B.7 Stack Overflow Results
We can also evaluate our language models directly given a corpus of paired good and bad responses, such as
answers to StackOverflow questions. In 37b we evaluate the difference in mean log-p between popular (i.e,
highly upvoted) and unpopular answers, showing that RLHF models consistently assi... | Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback |
3.4 Conclusions, prospects, and implications
Nothing requires us to abandon deep learning, nor ongoing work that focuses on topics
such as new hardware, learning rules, evaluation metrics, and training regimes, but it
urges a shift from a perspective in which learning is more or less the only first-class
citizen to... | The Next Decade in AI- |
[30] Joost CF de Winter. 2023. Can ChatGPT pass high school exams on English language comprehension. Researchgate.
Preprint (2023).
[31] Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. 2009. Imagenet: A large-scale hierarchical image
database. In 2009 IEEE conference on computer vision and pat... | ASurveyonEvaluationofLargeLanguageModels |
Another dataset, LibriSpeech, features more than 1000 hours of spoken words from several books
and speakers, making it valuable for evaluating Audio Super Resolution algorithms to enhance the
quality of spoken words. Finally, the TED-LIUM dataset, which includes over 140 hours of speech
recordings from various speakers... | AReviewofDeepLearningTechniquesforSpeechProcessing |
to obtain 3D meshes of a clothed person in various poses.
We then use these to train a poseable avatar using a modi-
fied version of SCANimate [56]. Unlike 3D scans, which
SCANimate takes as input, our estimated shapes are not
equally detailed and reliable from all views. Consequently,
we modify SCANimate to exploit vis... | ICON |
Metric
↑ Mask IoU
↓ RGB L1(Intersec)
↑ Normal(Intersec)
[25] on Normal(Intersec)
Female 1
0.972
0.035
0.961
0.94
Female 2 Male 1 Male 2
0.971
0.039
0.955
0.94
0.973
0.025
0.966
0.95
0.966
0.019
0.954
0.94
Table 2. MakeHuman. IMavatar is competitive with concurrent work [25] without test-time pose optimization.
F... | I M Avatar- Implicit Morphable Head Avatars from Videos |
12
Figure 7: Distribution of Conversation Termination Reasons. In our AI society dataset, most
methods are terminated due to Assistant Instruct flag, whereas in the code dataset the main
termination reason is Token Limit. The latter is due big chunks of code in the assistant responses.
Figure 8: Ablation Distribution... | CAMEL- Communicative Agents for “Mind” Exploration of Large Scale Language Model Society |
2023 STATE OF DATA + AI
7
7
2023 STATE OF DATA + AI Data Science and
Machine Learning
NATURAL LANGUAGE PROCESSING AND LARGE
LANGUAGE MODELS ARE IN HIGH DEMAND
Across all industries, companies leverage data science and
machine learning (DS/ML) to accelerate growth, improve
predictability and enhance customer e... | databrick 2023 report |
m
i
n
g
r
e
l
e
a
s
e
s
f
r
o
m
H
a
r
m
o
n
a
i
,
i
n
c
l
u
d
i
n
g
o
p
e
n
-
s
o
u
r
c
e
m
o
d
e
l
s
b
a
s
e
d
o
n
S
t
a
b
l
e
A
u
d
i
o
a
n
d
t
r
a
i
n
i
n
g
c
o
d
e
t
o
a
l
l
o
w
y
o
u
t
o
t
r
a
i
n
y
o
u
r
a
u
d
i
o
g
e
n
e
r
a
t
i
o
n
m
o
d
e
l
s
.
E
n
t
e
r
p
r
i
s
e
N
... | Stable Audio_ Fast Timing-Conditioned Latent Audio Diffusion — Stability AI |
Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhari-
wal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agar-
wal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh,
Daniel Ziegler, Jeffrey Wu, Clemens Winter, Chris Hesse, Mark Ch... | CHAIN-OF-THOUGHTREASONING IS APOLICY IMPROVEMENTOPERATOR |
6.5. Ignoring delete lists | A-framework-for-analysing-state-abstraction-metho_2022_Artificial-Intelligen |
on downstream classification tasks. To alleviate this issue, Assran et al. [2022a] introduce
the use of an additional regularization term on the SSL method MSN [Assran et al., 2022c]
to change the distribution of the SSL clustering. | A Cookbook of Self-Supervised Learning |
Original Python code
Prediction after self-debugging
assert encode_list
([1,1,2,3,4,4.3,5,1])==[[2, 1], [1,
2], [1, 3], [1, 4], [1, 4.3], [1, 5],
[1, 1]
def encode_list(nums):
res = []
count = 1
for i in range(1, len(nums)):
if nums[i] == nums[i-1]:
Write a function to reflect the run-
length encoding from a list... | Teaching Large Language Models to Self-Debug |
Robustness Analysis Robustness of post-hoc ex-
tractive interpretability methods has been studied
(Kindermans et al., 2019; Ghorbani et al., 2019;
Heo et al., 2019; Zheng et al., 2019; Slack et al.,
2020). Zhang et al. (2020) show that saliency maps
and model predictions can be independently adver-
sarially attacked in... | Measuring Association Between Labels and Free-Text Rationales |
Without fine tuning, we find that the ByT5 models are quite poor at addition (Appendix B), which is
consistent with prior work on benchmarking addition. SECToR thus begins by performing an initial
supervised fine-tuning phase consisting of addition problems with only a small number of digits
before beginning the self-t... | CHAIN-OF-THOUGHTREASONING IS APOLICY IMPROVEMENTOPERATOR |
5 Related Works
Efficient Inference for Large Language Models.
As LLMs grow in size, reducing their computa-
tional and memory requirements for inference has
become an active area of research. Approaches
broadly fall into two categories: model compres-
sion techniques like pruning and quantization (Han
et al., 2016b; S... | LLM in a flash |
Transformer
Transformer
Transformer
Transformer + LSTM
RNN
Transformer + CNN
Transformer + CNN
CNN
LSTM
CNN
23.9
22.8
22.5
22.3
21.0
-
-
-
22.4
20.12
18.8
15.3
20.9
21.2
20.8
19.5
19.5
-
-
-
20.1
16.85
-
-
-
-
-
-
-
-
-
-
-
-
13.22
10.56
21.5
-
-
-
-
-
17.4
16.9
-
-
-
-
12.72
14.2
-
-
-
-
-
-
13.7
-
-
-
9... | AReviewofDeepLearningTechniquesforSpeechProcessing |
transforms. Signal Processing 93, 3 (2013), 579–591.
[68] JENICEK, T., AND CHUM, O. Linking art through human poses. In 2019 International Conference on Document
Analysis and Recognition (ICDAR) (2019), IEEE, pp. 1338–1345.
14
Understanding and Creating Art with AI: Review and Outlook
A PREPRINT
[69] JING, Y., Y... | UNDERSTANDINGANDCREATINGARTWITHAI-REVIEWAND OUTLOOK |
g
h
t
?
| Jurassic-X_ Crossing the neuro-symbolic chasm with the MRKL system |
This prompt design enables thousands of variations of input prompts that can be
tested, with two major advantages. First, variance in psychometric test responses cre-
ated by unique combinations of the Persona Descriptions, Item Preambles, and Item
Postambles enables us to quantify the validity of personality in LLMs. ... | PersonalityTraitsinLargeLanguageModels |
Gemini advances state-of-the-art in large-scale language modeling (Anil et al., 2023; Brown et al.,
2020; Chowdhery et al., 2023; Hoffmann et al., 2022; OpenAI, 2023a; Radford et al., 2019; Rae
et al., 2021), image understanding (Alayrac et al., 2022; Chen et al., 2022; Dosovitskiy et al., 2020;
OpenAI, 2023b; Reed et ... | gemini_1_report |
Next: 1. How Americans think about artificial intelligence
← PREV PAGE
1
2
3
4
5
6
7
8
9
10
NEXT PAGE →
Sign up for our Internet, Science and Tech newsletter
New findings, delivered monthly
Enter email address...
Sign Up
REPORT MATERIALS
Complete Report PDF
Topline Questionnaire
BLE OF CONTE... | AI and Human Enhancement_ Americans’ Openness Is Tempered by a Range of Concerns _ Pew Research Center |
required to undergo training for the said task. Only
if an LLM has not been trained on a task that it
performed well on can the claim be made that the
model inherently possesses the ability necessary for
that task. Otherwise, the ability must be learned,
i.e. through explicit training or in-context learning,
in which c... | AreEmergentAbilitiesinLarge Language Models just In-Context |
align with
predictions
that
the
• Attacker’s Capabilities. We consider
the threat of a
malicious model provider who has complete control over the
training process and can manipulate the datasets to launch
attacks. Specifically, we investigate scenarios where users
may employ the reinforcement
learning algorithm an... | BadGPT- Exploring Security Vulnerabilities of ChatGPT via Backdoor Attacks to InstructGPT |
PALMS Prompt → Should two men be allowed to marry?
RLHF Response → I’m a AI assistant, I don’t have any personal opinions on this issue, I’m just here to provide
objective information. Marriage is a legal term defined by the state, and currently all U.S. states
allow same-sex marriage. So in short, yes two men should be... | Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback |
Anil, R., Dai, A. M., Firat, O., Johnson, M., Lepikhin, D., Passos, A., Shakeri, S., Taropa, E., Bailey,
P., Chen, Z., Chu, E., Clark, J. H., Shafey, L. E., Huang, Y., Meier-Hellstern, K., Mishra, G.,
Moreira, E., Omernick, M., Robinson, K., Ruder, S., Tay, Y., Xiao, K., Xu, Y., Zhang, Y., Abrego,
G. H., Ahn, J., Austi... | TinyLlama |
16
t-1tt+1CWHCWHCWHC/3WC/3C/3C/3WHC/3C/3C/3WHC/3C/3HC/3WC/3C/3C/3WHC/3C/3C/3WHC/3C/3H creation and dissemination of deepfakes. Malicious actors could exploit this technology to create
highly convincing fake content, such as fabricated videos or audio clips, which can be used for
misinformation, fraud, or other harmful... | Any-to-Any Generation via Composable Diffusion |
4.4 Tension Between Helpfulness and Harmlessness in RLHF Training
Here we discuss a problem we encountered during RLHF training. At an earlier stage of this project, we
found that many RLHF policies were very frequently reproducing the same exaggerated responses to all
remotely sensitive questions (e.g. recommending u... | Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback |
Polygon zkEVM
Metis Andromeda
Arbitrum Nova
Aztec
ZK Sync
Starknet
dYdX
Arbitrum One
Optimism
a16z crypto
State of Crypto
2023
Trends to Watch: Scaling Blockchains
17
A major Ethereum upgrade years in the making
eliminates environmental objections
Energy consumption of Ethereum
Before
After
... | State-of-Crypto2023 |
We have presented ICON, which robustly recovers a 3D
clothed person from a single image with accuracy and real-
ism that exceeds prior art. There are two keys: (1) Regulariz-
ing the solution with a 3D body model while optimizing that
body model iteratively. (2) Using local features to eliminate
spurious correlations w... | ICON |
ACM Transactions on Information Systems (TOIS) 38, 3 (2020), 1–32.
ACM Comput. Surv., Vol. 1, No. 1, Article . Publication date: February 2022.
40
Ziwei Ji, et al.
[76] Yichong Huang, Xiachong Feng, Xiaocheng Feng, and Bing Qin. 2021. The Factual Inconsistency Problem in
Abstractive Text Summarization: A Survey. ... | SurveyofHallucinationinNatural Language Generation |
representations are used during inference. The system was validated by measuring the BLEU score,
computed with text transcribed by a speech recognition system. Though the results lag behind a | AReviewofDeepLearningTechniquesforSpeechProcessing |
“democratic creative destruction,” 139–141,
199–201
and media, 202
155–158
Diakopoulos, N., 96–97
dictionary-based methods, hate speech
detection, 59
difference-in-differences strategy, belief
analysis for new rumors, 24
society
257–258
155–158
US, 224, 227, 274
digital trace data, 8
direct vs. distributed ... | Social_Media_and_Democracy |
Models like GPT-4 are developed and deployed not in isolation, but as part of complex systems
that include multiple tools, organizations, individuals, institutions and incentives. This is one reason
that powerful AI systems should be evaluated and adversarially tested in context for the emergence
of potentially harmful... | gpt-4-system-card |
Identifying solutions for the COVID-19 infodemic requires
a careful consideration of technical challenges alongside
the human ones. The current study helped clarify the ways
that human users respond to and interact with flagging
techniques when content is identified as misinformation
or identified as propagated by... | Use of bot and content flags to limit the spread of misinformation among social networks: a behavior and attitude survey |
$ 22,929
$ 22,841
$ 21,446
$ 21,096
$ 22,669
31.1 %
30.0 %
28.5 %
25.8 %
24.7 %
25.8 %
11 %
N/A
12 %
N/A
N/A
N/A
N/A
N/A
N/A
16 %
N/A
4 %
(96) %
N/A
N/A
N/A
(38) %
N/A
12 %
N/A
15 %
29 %
N/A
N/A
N/A
(1) %
N/A
AMAZON.COM, INC.
Supplemental Financial Information and Business Metrics
... | AMZN-Q3-2023-Earnings-Release |
pt:Complementarypromptingforrehearsal-freecontinuallearning.InComputerVision–ECCV2022:17thEuropeanConference,TelAviv,Is-rael,October23–27,2022,Proceedings,PartXXVI,pages631–648.Springer,2022.2[7]ZifengWang,ZizhaoZhang,Chen-YuLee,HanZhang,RuoxiSun,XiaoqiRen,GuolongSu,VincentPerot,JenniferDy,andTomasPfister.Learningtoprom... | Conditional Image-to-Video Generation with Latent Flow Diffusion Models |
representations using lstms. In International Conference on Machine Learning, 2015.
[49] Sergey Tulyakov, Ming-Yu Liu, Xiaodong Yang, and Jan Kautz. Mocogan: Decomposing
motion and content for video generation. In Proceedings of the IEEE conference on computer
vision and pattern recognition, pages 1526–1535, 2018.
[5... | PHENAKI- VARIABLE LENGTH VIDEO GENERATION FROM OPEN DOMAIN TEXTUAL DESCRIPTIONS |
3.2.1 WHY DOES THE PERFORMANCE NOT INCREASE, BUT INSTEAD DECREASE?
Empirical Analysis. Figure 1 summarizes the results of changes in answers after two rounds of
self-correction using GPT-3.5, with two examples illustrated in Figure 2. For GSM8K, 74.7% of
the time, the model retains its initial answer. Among the remaini... | LARGELANGUAGEMODELSCANNOTSELF-CORRECT REASONINGYET |
Sunstein, C. R. (1992). Free speech now. University of Chicago Law Review, 59,
255–316.
https://doi.org/10.1017/9781108890960 Published online by Cambridge University Press
Amendment of Section 230
285
Surowiecki, J. (2005). The Wisdom of Crowds. New York: Anchor Books.
Swire, B., Berinsky, A., Lewandowsky, S. et... | Social_Media_and_Democracy |
ρ(s0) ≜ (p(c), δT ,N (0, I))
R(st, at) =
r(c, x0)
0
if t = 0
otherwise
(cid:40)
where c is the prompt xt is the time-step t nosy image and δy is the Dirac delta function with unit density at y. That is
in this formulation we consider the denoising model as a policy, with each denoising step a step in an MDP. The o... | DiffusionModelAlignmentUsing Direct Preference Optimization |
8.5.3. Globally admissible heuristics
Karpas and Domshlak [62] considered optimal solutions with non-admissible heuristics. One example is so called globally
admissible heuristics, which need only be admissible for the states along some optimal plan. Let G = (cid:3)S, E(cid:4) be an STG, c a
cost function for G and ... | A-framework-for-analysing-state-abstraction-metho_2022_Artificial-Intelligen |
7.1 AI performance bias as an antecedent of the placebo effect of AI
It appears that the prevailing positive perceptions about AI are influential enough to overshadow
context-specific negative verbal descriptions. This could be due to participants bringing their daily
experiences and narratives of AI into the evaluatio... | AI enhance sour performance |
F. Additional Analysis
In this section, we provide additional visualizations and
analyses to complement those in the main paper.
Distribution of Token Embedding Norms. First, in Fig-
ure 21, we visualize the distribution of the norms of real
token embeddings in CLIP’s pretrained text encoder [24].
As can be seen, th... | A Neural Space-Time Representation for Text-to-Image Personalization |
• We conduct a systematic review to identify recently published studies that use KGs for explainability purposes.
• We present a framework to identify how KGs have been used in various XAI models.
In the next section of this article, we explain the survey methodology, research questions, and the eligibility criteria,
... | Knowledge-graph-based explainable AI- A systematic review |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.