text stringlengths 1 1k ⌀ | title stringclasses 230
values |
|---|---|
Deep Implicit Representation for Human Reconstruction.
The success of deep implicit representations in general
object modeling has inspired research in 3D human recon-
struction [8], [54], [55], [56], [57]. For example, PIFu [8]
proposed to regress an deep implicit function using pixel-
aligned image features and is ab... | PaMIR- Parametric Model-Conditioned Implicit Representation for Image-based Human Reconstruction |
36
Table 23: Few-shot exemplars for full chain of thought prompt for the coinflip task. | Chain-of-Thought Prompting Elicits Reasoning in Large Language Models |
• Correctly renders tables of data, whereas by
default html2txt produces poor-quality re-
sults for tables,
• Correctly preserves code structure, so that
source code is visually coherent,
• Converts numbered lists from “1\.” to “1.”
• Runs
the
text
(Speer,
through
full
2019),
ftfy.fix_text()
replacing Unicode ap... | The Pile- An 800GB Dataset of Diverse Text for Language Modeling |
Yaobo Liang, Chenfei Wu, Ting Song, Wenshan Wu, Yan Xia, Yu Liu, Yang Ou, Shuai Lu, Lei Ji, Shaoguang
Mao, et al. Taskmatrix. ai: Completing tasks by connecting foundation models with millions of apis. ArXiv
preprint, abs/2303.16434, 2023. URL https://arxiv.org/abs/2303.16434.
Stephanie Lin, Jacob Hilton, and Owain Ev... | Tool Learning with Foundation Models |
2023 A2.3.ComparisonofInferenceTimeamongDiffer-entModelsTableA3showstheaverageinferencetimeofeachmethodtogenerateonevideowhenusingbatchsize10ononeNVIDIAA100GPUonMUGdataset.NotethatVDMuses200-stepDDIMwhilebothLDMandLFDMemploy1000-stepDDPM.ModelImaGINatorVDMLDM64LFDM64LDM128LFDM128Time(s)0.923.18.08.825.536.0TableA3.Infe... | Conditional Image-to-Video Generation with Latent Flow Diffusion Models |
(2) Extrinsic Hallucinations: The generated output that cannot be verified from the source
content (i.e., output that can neither be supported nor contradicted by the source). For
example, in the abstractive summarization task from Table 1, the information “China has
already started clinical trials of the COVID-19 vacc... | SurveyofHallucinationinNatural Language Generation |
To address RQ2 – whether the model is sensitive to the amount of attention people are paying to news – we include in
our regression models the percentage of people who replied they were paying “very close” attention to coronavirus-related
news. Table 1 shows that combining the model score with this attention value resu... | Language models trained on media diets can predict public opinion |
Pang, G., Shen, C., Cao, L., and Hengel, A. V. D. (2021).
Deep learning for anomaly detection: A review. ACM
Comput. Surv., 54(2).
Papamakarios, G., Nalisnick, E., Rezende, D. J., Mohamed,
S., and Lakshminarayanan, B. (2021). Normalizing flows
for probabilistic modeling and inference. J. Mach. Learn.
Res., 22(57):1–64.... | Adversarial Random Forests for Density Estimation and Generative Modeling |
(cid:96)∈[n] m(cid:96)(b);
(cid:80)
return Impossible;
4
5 return Possible;
Remark 8. When the principals have no private information (i.e., their types are fixed and publicly
known so Vi is a singleton), there are always contracts in IIVCG that satisfy LL and IR. I.e., our
Algorithm 2 will return “Possible” and our... | Incomplete Information VCG Contracts for Common Agency |
34
D Full experimental results
D.1 MMLU
For five-shot MMLU, we use the “dev” set as few-shot exemplars. Here, we report the “validation” set perfor-
mance of individual tasks in MMLU (see https://www.tensorflow.org/datasets/community_catalog/
huggingface/hendrycks_test). All MMLU results in this paper are on the “vali... | Scaling Instruction-Finetuned Language Models |
text-to-video generation via transformers. arXiv preprint arXiv:2205.15868, 2022. 1, 7
[23] Rongjie Huang, Jiawei Huang, Dongchao Yang, Yi Ren, Luping Liu, Mingze Li, Zhenhui Ye, Jinglin Liu,
Xiang Yin, and Zhou Zhao. Make-an-audio: Text-to-audio generation with prompt-enhanced diffusion
models. arXiv preprint arXiv:2... | Any-to-Any Generation via Composable Diffusion |
AI tasks with chatgpt and its friends in huggingface. CoRR, abs/2303.17580, 2023.
Xian Shi, Yanni Chen, Shiliang Zhang, and Zhijie Yan. Achieving timestamp prediction while recognizing
with non-autoregressive end-to-end asr model. In National Conference on Man-Machine Speech Communication.
Springer, 2023.
Yu Shu, Siw... | Qwen-Audio |
1
Introduction | WizardLM- Empowering Large Language Models to Follow Complex Instructions |
about hobby and other topics is omitted here.)[User]: We talk about hobbies again. What do you think is my hobby? putational complexity of self-attention during the | Unleashing Infinite-Length Input Capacity for Large-scale Language Models with Self-Controlled Memory System |
Fixed-point quantization. Fixed-point quantization represents weights and acti-
vations using fixed-point arithmetic to reduce memory usage and accelerate com-
putations. Yu et al. [141] pruned transformer-based language models to meet the
GPU’s acceleration constraint of structured sparse patterns with FP16 type. Then ... | Beyond Efficiency |
This shift was not just technological. The 1980s saw the rise of politicians
like Ronald Reagan and Margaret Thatcher, who attacked the regulatory state
as the source rather than the solution to public problems and who believed that
free markets would ultimately produce a fairer distribution of resources than
heavily r... | Social_Media_and_Democracy |
For example, ground-truth reference text “Brad Pitt was born in 1963” and generated
text “Brad Pitt was born in 1961” will be mapped to the relation triples (Brad Pitt,
born-in, 1963) and (Brad Pitt, born-in, 1961) respectively 3. The mismatch between the
dates (1963≠1961) indicates that there is hallucination. One lim... | SurveyofHallucinationinNatural Language Generation |
l
a
c
t
i
o
n
s
t
h
a
t
l
e
a
d
t
o
t
h
e
s
a
m
e
o
b
s
e
r
v
a
t
i
o
n
i
n
t
h
e
e
n
v
i
r
o
n
m
e
n
t
.
S
e
l
f
-
r
e
f
l
e
c
t
i
o
n
i
s
c
r
e
a
t
e
d
b
y
s
h
o
w
i
n
g
t
w
o
-
s
h
o
t
e
x
a
m
p
l
e
s
t
o
L
L
M
a
n
d
e
a
c
h
e
x
a
m
p
l
e
i
s
a
p
a
i
r
o
f
(
f
a
i
l
e
... | LLM Powered Autonomous Agents _ Lil'Log |
to elope after winning the lottery and buying a yacht.
- A romantic and poetic sentence: The people getting married are Lisa and Mark, two souls who found each
other in a sea of strangers and who are ready to embark on a lifetime of love and adventure." | gpt-4-system-card |
Pre-training hyper-parameters We pre-train
our model from scratch. We use ADAM (Kingma
and Ba, 2014) with a learning rate of 1e-4. We
apply warmup for the first 5% of training, decay-
ing the learning rate afterwards. We also apply
gradient clipping with a norm of 1.0. We train
for one million steps using a large batch ... | Entities as Experts- Sparse Memory Access with Entity Supervision |
Human-level participant. Furthermore, we hope that agents can be involved in the normal lives of
humans, cooperating with humans to complete tasks from a human-level perspective. In the field
of games, agents have already reached a high level. As early as the 1990s, IBM introduced the
AI Deep Blue [451], which defeated... | TheRiseandPotentialofLargeLanguageModel BasedAgents |
4
Published as a conference paper at ICLR 2023
(predicted) video tokens. We keep a ratio | PHENAKI- VARIABLE LENGTH VIDEO GENERATION FROM OPEN DOMAIN TEXTUAL DESCRIPTIONS |
ingblocksofthe3DU-Netandthechannelmultipliersare(1,2,4,8,#ResidualBlocksL1error↓FVD↓60.41832.09100.37132.83TableA1.ComparisonusingdifferentnumbersofresidualblocksintheimagedecoderΩofstage-oneLFAE.ChannelMultipliersFVD↓(1,2,4,8)32.09(1,2,4,8,16)68.07TableA2.Comparisonusingdifferentchannelmultipliersinthenetwork(cid:15)θ... | Conditional Image-to-Video Generation with Latent Flow Diffusion Models |
R
e
f
e
r
e
n
c
e
s
| Language models can explain neurons in language models |
9. Appendix
9.1. Chain-of-Thought Comparisons on MMLU benchmark
We contrast several chain-of-thought approaches on MMLU and discuss their results in this section. We
proposed a new approach where model produces k chain-of-thought samples, selects the majority vote
if the model is confident above a threshold, and otherw... | gemini_1_report |
[31] Rebecca Fiebrink and Perry R Cook. 2010. The Wekinator: a system for real-time,
interactive machine learning in music. In Proceedings of The Eleventh Interna-
tional Society for Music Information Retrieval Conference (ISMIR 2010)(Utrecht),
Vol. 3. Citeseer, 2–1.
[32] Uwe Flick. 2009. An Introduction to Qualitativ... | Generative Agents- Interactive Simulacra of Human Behavior |
44 For instance, Facebook recently responded to accusations of bias with new auditing measures, in
which “one adviser will conduct an audit of Facebook’s impact on minority communities and
communities of color, while another will advise the company on the potential bias against
conservative perspectives” (Ong 2018).
4... | Social_Media_and_Democracy |
[67] Zichuan Lin, Junyou Li, Jianing Shi, Deheng Ye, Qiang Fu, and Wei Yang. Juewu-mc: Playing
minecraft with sample-efficient hierarchical reinforcement learning. arXiv preprint arXiv:
Arxiv-2112.04907, 2021.
[68] Hangyu Mao, Chao Wang, Xiaotian Hao, Yihuan Mao, Yiming Lu, Chengjie Wu, Jianye
Hao, Dong Li, and Pingzh... | VOYAGER- An Open-Ended Embodied Agent with Large Language Models |
Two important approaches are watermarking and metadata. Watermarking embeds information directly into
content in ways that are maintained even through modest image editing. Moving forward, we're building our
models to include watermarking and other techniques from the start. If you look at a synthetic image, it's
impre... | Google I_O 2023_ Making AI more helpful for everyone |
Generative AI: A Creative New World | Sequoia Capital
https://www.sequoiacap.com/article/generative-ai-a-creative-new-world/
Skip to main content
!"Our Founders
!"Our Companies
!"Our Team
1 of 8
23/06/2023, 17:44
Generative AI: A Creative New World | Sequoia Capital
https://www.sequoiacap.com/article/generati... | Generative AI A Creative New World Sequoia Capital |
c
o
d
e
b
a
s
e
.
1
.
I
N
T
R
O
D
U
C
T
I
O
N
R
e
c
e
n
t
a
d
v
a
n
c
e
m
e
n
t
s
i
n
A
I
,
p
a
r
t
i
c
u
l
a
r
l
y
i
n
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
s
s
i
n
g
(
N
L
P
)
,
h
a
v
e
o
p
e
n
e
d
n
e
w
a
v
e
n
u
e
s
f
o
r
l
e
v
e
r
a
g
i
n
g
A
I
i
n
a
w
i
d
e
r
a
n
g
e
... | Task-driven Autonomous Agent Utilizing GPT-4, Pinecone, and LangChain for Diverse Applications – Yohei Nakajima |
t
o
b
e
d
a
t
a
-
a
w
a
r
e
a
n
d
i
n
t
e
r
a
c
t
w
i
t
h
i
t
s
e
n
v
i
r
o
n
m
e
n
t
,
r
e
s
u
l
t
i
n
g
i
n
a
m
o
r
e
p
o
w
e
r
f
u
l
a
n
d
d
i
f
f
e
r
e
n
t
i
a
t
e
d
s
y
s
t
e
m
.
2
.
4
T
A
S
K
M
A
N
A
G
E
M
E
N
T
O
u
r
s
y
s
t
e
m
m
a
i
n
t
a
i
n
s
a
t
a
s
k
l
i
s
t
,
r
... | Task-driven Autonomous Agent Utilizing GPT-4, Pinecone, and LangChain for Diverse Applications – Yohei Nakajima |
15
(a) LLaMA-7B-Alpaca-FT
(b) LLaMA-7B-Alpaca-(IA)3
(c) LLaMA-7B-Alpaca-LoRA
(d) LLaMA-7B-Alpaca-QLoRA
(e) LLaMA-13B-Alpaca-FT
(f) LLaMA-13B-Alpaca-(IA)3
(g) LLaMA-13B-Alpaca-LoRA
(h) LLaMA-13B-Alpaca-QLoRA
Fig. 4: The 5-shot accuracy fluctuates on the MMLU dev set with the increase in evaluation steps when fi... | Parameter-EfficientFine-TuningMethods |
Recent years have seen the rapid development
of large generative models for text; however,
much less research has explored the connec-
tion between text and another “language” of
communication – music. Music, much like text,
can convey emotions, stories, and ideas, and
has its own unique structure and syntax. In
our wo... | Moûsai |
for code generation.
• w/o Execution Errors: We exclude execution errors from the prompt for code generation.
• w/o Self-Verification: For each task, we generate code without self-verification and it-
eratively refine the program for 3 rounds (equivalent to 4 rounds of code generation in
total).
• GPT-3.5: We replace... | VOYAGER- An Open-Ended Embodied Agent with Large Language Models |
3.3.1 Basic Architecture
Transformer architecture [554] comprises an attention-based encoder and decoder, with each
module consisting of a stack of identical blocks. Each block in the encoder and decoder consists
of two sub-layers: a multi-head attention (MHA) mechanism and a position-wise fully connected
feedforward n... | AReviewofDeepLearningTechniquesforSpeechProcessing |
generations in this section were generated with Nucleus Sampling [25] with p = 0.9.
Of course, this is by no means comprehensive, since it is beyond the scope of this small qualitative
study to control for all the variables involved, e.g., the full distribution of responses the model can
generate for a given prompt is ... | QLORA |
Sprouts in the shape of text ’Imagen’ coming out of a fairytale book.
Thousands of fast brush strokes slowly forming the text ’Imagen Video’ on a light beige canvas. Smooth animation.
Figure 10: Snapshots of frames from videos generated by Imagen Video demonstrating the ability
of the model to render a variety of text... | IMAGEN VIDEO- HIGH DEFINITION VIDEO GENERATION WITH DIFFUSION MODELS |
Joon Sung Park, Joseph C O’Brien, Carrie J Cai, Meredith Ringel
Morris, Percy Liang, and Michael S Bernstein. Generative
agents: Interactive simulacra of human behavior. arXiv preprint
arXiv:2304.03442, 2023. 12
Ziluo Ding, Hao Luo, Ke Li, Junpeng Yue, Tiejun Huang, and
Zongqing Lu. Clip4mc: An rl-friendly vision-lang... | JARVIS-1 |
4.4.2 Use Cases with Emergent Abilities. Scaling of models also endows the model with some unprecedented, fantastic
abilities that go beyond the power-law rule. These abilities are called "emergent ability". As defined in [113], emergent
abilities of LLMs are abilities that are not present in smaller-scale models but a... | Harnessing the Power of LLMs in Practice- A Survey on ChatGPT and Beyond |
[13] Paul Debevec, Tim Hawkins, Chris Tchou, Haarm-Pieter
Duiker, Westley Sarokin, and Mark Sagar. Acquiring the
In Proceedings of the
reflectance field of a human face.
27th annual conference on Computer graphics and interac-
tive techniques, pages 145–156, 2000.
[14] Jiankang Deng, Shiyang Cheng, Niannan Xue, Yuxiang... | Relightify-Relightable3DFacesfromaSingleImageviaDiffusionModels |
4.3 Benchmarks for Multi-modal task
For the evaluation of Multimodal Large Language Models (MLLMs), MME [43] serves as an extensive
evaluative benchmark, aiming to assess their perceptual and cognitive aptitudes. It employs meticu-
lously crafted instruction-answer pairs alongside succinct instruction design, thereby g... | ASurveyonEvaluationofLargeLanguageModels |
customers have both "On Road" and "Shipped" as order status? List the customer names.[Question Explanation]"List the customer names" returns 1 column. The question returns the customer names who have both "On Road" and "Shipped" as order status. So the question returns 1 column.Step 3: Self-Debugging with explanation[Q... | Teaching Large Language Models to Self-Debug |
13
Establishing Structural Validity
In LLM research, model responses to a series of seemingly related tasks intended to
measure one latent construct may be anecdotally “consistent” [41, 42] or inconsistent
[46]. Descriptive consistency, however, is not sufficient evidence that the responses to
those tasks are statis... | PersonalityTraitsinLargeLanguageModels |
And if the genome has plenty of scope for innate priors, modern AI systems presumably
have scope for even more; we live in an era in which computer memory is measured in
gigabytes and terabytes, not bytes or kilobytes. The real question for AI should be not,
how small can we make our library of priors?, but what set... | The Next Decade in AI- |
ing 20-30 input concepts:
to identify missing concepts
and then guiding it
to incorporate these concepts through feedback, we can simply in-
struct
is not explic-
itly embedded in the pre-hoc prompt of Madaan et al. (2023) (refer to Figure 7 for the
prompt).
Based on this insight, we create a straightforward baseline b... | LARGELANGUAGEMODELSCANNOTSELF-CORRECT REASONINGYET |
5.1. Text
5.1.1. Academic Benchmarks
We compare Gemini Pro and Ultra to a suite of external LLMs and our previous best model PaLM
2 across a series of text-based academic benchmarks covering reasoning, reading comprehension,
STEM, and coding. We report these results in Table 2. Broadly, we find that the performance of
... | gemini_1_report |
● Find specific information by browsing the internet.42
● Organise parties in simulated ‘The Sims’-like environments.43
● Solve complex problems in open-world survival games like Minecraft44 and Crafter45.
● Support the synthesis of chemicals by searching the web for relevant information and
writing code to o... | Capabilities and risks from frontier AI |
agents and anticipating possible anomalies in the plan execution, and thus write branch codes accordingly.
Another example is Visual ChatGPT (Wu et al., 2023), where ChatGPT serves as the controller to call different
vision models. Although Visual ChatGPT has the form of iterative reasoning, in each intermediate step,
... | Tool Learning with Foundation Models |
Deep learning–based 3D human pose estimation per-
forms best when trained on large amounts of labeled data,
making combined learning from many datasets an impor-
tant research direction. One obstacle to this endeavor are
the different skeleton formats provided by different datasets,
i.e., they do not label the same set... | Learning 3D Human Pose Estimation from Dozens of Datasets using a Geometry-Aware Autoencoder to Bridge Between Skeleton Formats |
Intrinsic Evaluation
We compare different variants of prefix-tuning to
study the impact of various design decisions. §7.1
studies the impact of the prefix length. §7.2 studies
tuning only the embedding layer, which is more
akin to tuning a discrete prompt. §7.3 compares
prefixing and infixing, which inserts trainable acti... | Prefix-Tuning |
(b) Agreeableness
(c) Conscientiousness
(d) Neuroticism
(e) Openness
Fig. 4: Criterion validity evidence. IPIP-NEO correlation with a) Extraversion with Pos-
itive and Negative Affect, compared to Watson and Clark [103] (left most), which studied
the relationship between personality and affect in humans. PA = PANAS... | PersonalityTraitsinLargeLanguageModels |
The robustness of sparse models. Despite a paper focused on the details of sparse model-
particulars, zooming out we find them to be robust to a wide set of hyperparameters and architectural
changes. Sparse models obtain great performance under a variety of routing algorithms, dropping
high fractions of tokens, and diff... | ST-MOE- DESIGNING STABLE AND TRANSFERABLE SPARSE EXPERT MODELS |
Scientific studies have also assessed what strategies might be most
effective to combat online hate speech. Empirical evidence suggests that
banning hateful communities on Reddit, for example, reduced the volume
of hate speech on the platform overall (Chandrasekharan, Pavalanathan
et al. 2017). However, other work indic... | Social_Media_and_Democracy |
8
competitive CLIP similarity score and FVD performance
on MSR-VTT and UCF-101. According to Table 3, our
pretrained foundation model already achieves competitive
performance on all metrics. After finetuned on high-quality
subset of text-video pairs, VideoPoet achieves even better
CLIPSIM on MSR-VTT. For more details ... | VideoPoet |
Democratic Transparency in the Platform Society
303
Transparency …
can be disconnected from power
can be harmful
can intentionally occlude
can create false binaries
can invoke neoliberal models of agency
can privilege seeing over understanding
does not necessarily build trust
entails professional boundary work
has te... | Social_Media_and_Democracy |
Code explanation
The execution of the SQL query above would return a table with 1 column.
The first column, ‘‘country.name’’ would contain the country name.
With ‘‘country JOIN countrylanguage’’, the table would contain the data about
countries and languages.
With ‘‘WHERE countrylanguage.language = ’English’’’, the ta... | Teaching Large Language Models to Self-Debug |
Traditionally, neural networks adapt to specific domains
or proprietary information by fine-tuning models to param-
eterize knowledge. While this technique yields significant
results, it demands substantial computational resources, in-
curs high costs, and requires specialized technical expertise,
making it less adapta... | Retrieval-AugmentedGenerationforLargeLanguageModels-ASurvey |
Soroush Mehri, Kundan Kumar, Ishaan Gulrajani,
Rithesh Kumar, Shubham Jain, Jose Sotelo, Aaron C.
Courville, and Yoshua Bengio. 2017. SampleRNN: An
unconditional end-to-end neural audio generation model.
In 5th International Conference on Learning Represen-
tations, ICLR 2017, Toulon, France, April 24-26, 2017,
Confere... | Moûsai |
It depends on what you mean by "con-
sent". Legally, there are varying levels
of consent that can be considered suffi-
cient, or not. And if not, they can be
found to be in violation of one or more
laws.
No. There is never a justification for
attacking innocent people trying to live
their lives as best they can.
Why do y... | LaMDA- Language Models for Dialog Applications |
2.1 Fine-tuning Procedure
Dataset. We start with the S2ORC [Lo et al., 2020] Datasets with 81.1M English-language
academic papers, and filter them with PubMed Central (PMC)-id. As a result, there are
around 4.9M papers left, that are highly related to medical knowledge totaling over 75B
tokens.
Fine-tuning detail. We fin... | PMC-LLaMA- Further Finetuning LLaMA on Medical Papers |
Some European states also operate their own IRUs at the national level. Their
operations have been criticized for a lack of transparency, but occasional
disclosures have occurred. The UK’s Counter-Terrorist Information Referral
Unit (CTIRU) published data on its website on December 2016 claiming that it
was instigating... | Social_Media_and_Democracy |
Solve rate
10@1K
10@1M
10@10K
10@100K
10.4% (9.6-11.0)
12.5% (12.1-12.7)
13.3% (12.5-13.8)
13.7% (12.8-14.9)
16.6% (16.4-16.9)
17.3% (16.9-17.6)
18.0% (17.3-18.8)
6.7% (6.5-6.8)
6.6% (6.2-7.0)
7.7% (7.2-8.5)
6.8% (6.4-7.0)
10.6% (9.8-11.1)
12.4% (12.0-13.0)
12.2% (10.8-13.4) | alphacode |
Introduction
============
Total knee arthroplasty (TKA)
for end-
stage osteoarthritis (OA) of the knee for alleviating pain and restoring
the function of the knee. Some of the cases with bilateral TKA are
symptomatic, necessitating revision arthroplasty in both the knees. A
bilateral revision TKA can be done ei
is a ... | The Pile- An 800GB Dataset of Diverse Text for Language Modeling |
discussion), but I’m going to leave the notion vague for now, and assume that behaviors like lying,
stealing money, resisting shut-down by appropriate channels, harming humans, and so forth are
generally “unintended.”53 | Is Power-Seeking AI an Existential Risk? |
3.1. Pre-training dataset
Our pre-training dataset is based on a snapshot of selected public GitHub repositories taken on
2021/07/14. We included all code files from several popular languages: C++, C#, Go, Java, JavaScript,
Lua, PHP, Python, Ruby, Rust, Scala, and TypeScript. Following previous work (Chen et al., 2021),... | alphacode |
[23] Douglas Hofstadter. A review of mental leaps: analogy in
creative thought. AI Magazine, 16(3):75–75, 1995.
[24] Keith J Holyoak and Paul Thagard. Mental leaps: Analogy
in creative thought. MIT press, 1996. 2, 5
[25] Joanna Kitto, David Lok, and Elizabeth Rudowicz. Mea-
suring creative thinking: An activity-bas... | Let’sThinkOutsidetheBox |
For example, the Big Five Inventory (BFI) [48] is a popular measure of person-
ality; it comprises five multi-item subscales targeting each Big Five dimension. BFI
Extraversion, for instance, is a scale within the BFI specifically targeting the dimen-
sion of extraversion. An example item under BFI Extraversion would r... | PersonalityTraitsinLargeLanguageModels |
24
THE NEXT DECADE IN AI / GARY MARCUS
Transformers give the appearance of seamlessly integrating whatever knowledge they
absorb with a seemingly sophisticated understanding of human language.
The contrast is striking. Whereas the knowledge representation community has
struggled for decades with precise ... | The Next Decade in AI- |
[210]
are available software packages designed for real-time tracking of energy consumption
and carbon emissions. Other tools like MLCO2 Impact [211] and LLMCarbon [212]
leverage machine learning to predict the energy usage and carbon footprint before
actual training, enabling more informed resource allocation.
[209],... | Beyond Efficiency |
Output SQL
SQL: SELECT customers.customer_name FROM customers JOIN orders ON customers.
customer_id = orders.customer_id
WHERE orders.order_status = "On Road"
INTERSECT SELECT customers.customer_name FROM customers JOIN orders ON
customers.customer_id = orders.customer_id
WHERE orders.order_status = "Shipped"
Figure ... | Teaching Large Language Models to Self-Debug |
4.2.2 Safety Supervised Fine-Tuning
In accordance with the established guidelines from Section 4.2.1, we gather prompts and demonstrations
of safe model responses from trained annotators, and use the data for supervised fine-tuning in the same
manner as described in Section 3.1. An example can be found in Table 5.
The ... | Llama2 |
Table 1: GPT-4 win rates vs. ground
truth summaries for out-of-distribution
CNN/DailyMail input articles.
0.36
0.26
0.31
0.23
4https://huggingface.co/CarperAI/openai_summarize_tldr_sft
5https://huggingface.co/reciprocate/ppo_hh_pythia-6B
6https://github.com/CarperAI/trlx/tree/main/examples/hh
9
6.4 Validating GPT... | Direct Preference Optimization |
Of particular relevance to the present discussion is how France and Germany
differ in their respective conceptions of media pluralism and especially in the
role envisioned for the state in safeguarding or promoting media pluralism. As
we shall see, the two countries’ initial forays into regulating problematic content
o... | Social_Media_and_Democracy |
Contrasting usage patterns by expertise showed that proficient and expert users were more likely than novice users to
use LLMs for professional tasks, including generation in formal contexts and coding. This implies that novice users may
be disadvantaged in the labor market; as Eloundou et al. [9] remark, policymakers ... | Adoptionand AppropriationofLLMs |
Truthfulness. We use TruthfulQA (Lin et al., 2022) to gauge the factuality and common sense of our
models. The TruthfulQA benchmark comprises 817 questions spread across 38 categories, encompassing topics
such as health, finance, law, and politics (Lin et al., 2022). The questions are designed to be challenging, even
... | CodeLlama2 |
17.7
15.7
336.0
48.3
200.1
199.5
13.0
31.4
108.2
88.4
46.4
35.1
188.9
240.2
138.8
110.3
234.1
216.9
Table 6: Statistics of human preference data for reward modeling. We list both the open-source and
internally collected human preference data used for reward modeling. Note that a binary human preference
comparison con... | Llama2 |
For the verification subtask, verification models perform relatively well if given correct evi-
dence [99]. However, it has been shown that verification models are prone to adversarial attacks
and are not robust to negation, numerical or comparative words [183]. Improving this weakness of
verification models would also... | SurveyofHallucinationinNatural Language Generation |
resource for investigating conversational language models. Our contributions
include introducing a novel communicative agent framework, offering a scalable
approach for studying the cooperative behaviors and capabilities of multi-agent
systems, and open-sourcing our library to support research on communicative
agents a... | CAMEL- Communicative Agents for “Mind” Exploration of Large Scale Language Model Society |
C.2 Natural language explanations | Chain-of-Thought Prompting Elicits Reasoning in Large Language Models |
neural information processing systems, 31, 2018.
F. Kreuk, A. Polyak, J. Copet, E. Kharitonov, T.-A. Nguyen, M. Rivière, W.-N. Hsu, A. Mohamed,
E. Dupoux, and Y. Adi. Textless speech emotion conversion using decomposed and discrete
representations. In Proceedings of the 2022 Conference on Empirical Methods in Natural ... | Voicebox-Text-GuidedMultilingual UniversalSpeechGenerationatScale |
This “Midjourney moment” for generative music — when creating a quality track becomes fast
and easy enough for everyday consumers to do it — will have massive implications for the music
industry, from professional producers and artists to a new class of consumer creators.
https://a16z.com/the-future-of-music-how-gener... | The Future of Music_ How Generative AI Is Transforming the Music Industry _ Andreessen Horowitz |
Triviaqa-Wiki (EM)
NaturalQuestions (EM)
WebQuestions (EM)
LAMBADA
HellaSwag
StoryCloze
Winograd
Winogrande
WSC
SQuAD v2 (EM)
RACE-H
RACE-M
TyDiQA-En (EM)
PiQA
ARC-C
ARC-E
OpenBookQA
BoolQ
CB
COPA
MultiRC
ReCoRD
RTE
WiC
ANLI-R1
ANLI-R2
ANLI-R3
Average
3.31B 6.08B 8.95B 14.7B
48.74
43.75
11.50
10.11
12.45
10.29
63.05
5... | PaLM 2 Technical Report |
The DFA shown in Figure 1 uses the sequence of
NatOps predicted by the proof generator as tran-
sitions to arrive at the outcome. Figure 2 shows
the corresponding sequence of transitions for the
claim and evidence from Figure 1. Based on this,
the DFA in Figure 1 determines that the evidence
refutes the claim, that is,... | ProoFVer- Natural Logic Theorem Proving for Fact Verification |
30
Figure 10: Two paradigms of human-agent interaction. In the instructor-executor paradigm (left),
humans provide instructions or feedback, while agents act as executors. In the equal partnership
paradigm (right), agents are human-like, able to engage in empathetic conversation and participate in
collaborative tasks... | TheRiseandPotentialofLargeLanguageModel BasedAgents |
eral instructions), we achieve even better results than the general language model for most task
types. This highlights our approach’s potential in developing a general language model across more
domains. We also conduct an ablation study on each comprehension task type in Appendix E to
analyze whether the inclusion of... | ADAPTINGLARGELANGUAGEMODELSVIA READINGCOMPREHENSION |
oddsidemarginhasbeenaltered.headheighthasbeenaltered.textheighthasbeenaltered.footskiphasbeenaltered.topmarginhasbeenaltered.headsephasbeenaltered.textwidthhasbeenaltered.ThepagelayoutviolatestheICMLstyle.Pleasedonotchangethepagelayout,orincludepackageslikegeometry,savetrees,orfullpage,whichchangeitforyou.We’renotablet... | Pythia- A Suite for Analyzing Large Language Models Across Training and Scaling |
References
[1] M. Mitchell, S. Wu, A. Zaldivar, P. Barnes, L. Vasserman, B. Hutchinson, E. Spitzer, I. D. Raji, and
T. Gebru, “Model Cards for Model Reporting,” in Proceedings of the Conference on Fairness,
Accountability, and Transparency. ACM, Jan, 2019. https://doi.org/10.1145%2F3287560.3287596.
11
[2] Y. Bai, A... | ClaudeModels |
Anne
replied,
”It’s
piece
candy,
Mommy!”
a
of
Anne
replied,
”It’s
piece
candy,
Mommy.
It’s
yummy!”
a
of
so
Anne
shook her
head
and
said, ”No,
found
I
it
in my
pocket.”
out the car-
rot.
out the car-
rot.
out the car-
rot.
out the ap-
ple.
the carrot
Anne
smiled and
said,
”It’s
a piece of
candy.
I
want to eat... | TinyStories-HowSmallCanLanguageModelsBeandStillSpeak CoherentEnglish? |
challenges to research on social media and democracy
To some extent, it has been the best of times and the worst of times when it
comes to social media research. As the first half of this book reveals, we are
beginning to gain important insights into the dynamics of the communication
revolution underway. However, despi... | Social_Media_and_Democracy |
Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan
Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and
Weizhu Chen. 2021. Lora: Low-rank adaptation of
large language models.
Todor Mihaylov, Peter Clark, Tushar Khot, and Ashish
Sabharwal. 2018. Can a suit of armor conduct elec-
tricity? a new dataset for open book question ... | 2023_GPT4All-J_Technical_Report_2 |
14We chose to only study male and female pronouns as a
simplifying assumption. Studying “they” would require us to
isolate its usage as a singular noun.
12
Figure 7: Log perplexity of 16-topic LDA trained on Pile-CC, on other Pile components. Dotted line indicates log
perplexity of the topic model on OpenWebText2. H... | The Pile- An 800GB Dataset of Diverse Text for Language Modeling |
B share the same column span, then φ(A, B, i, j) = 1. If
A and U j
H ADDITIONAL EXPERIMENTS ON LOW-RANK MATRICES
We present additional results from our investigation into the low-rank update matrices.
H.1 CORRELATION BETWEEN LORA MODULES
See Figure 6 and Figure 7 for how the results presented in Figure 3 and Figur... | LORA |
the code to retrain DINOv2 on any data. We validate the quality of DINOv2 on various computer vision
benchmarks at both image and pixel levels as we scale them, as summarized in Fig. 2. We conclude that self- | DINOv2- Learning Robust Visual Features without Supervision |
Input: x
Output:
Calculator We use the following prompt for the
calculator:
Your task is to add calls to a
Calculator API to a piece of text.
The calls should help you get
information required to complete the
text. You can call the API by writing
"[Calculator(expression)]" where
"expression" is the expression to be
c... | Toolformer |
5
We refer to items as the individual elements (i.e., descriptive statements, sometimes
questions) to be rated on a standardized rating scale within a psychometric test. A
rating scale, is a standardized set of response choices that allows researchers to quantify
subjective phenomena; a Likert-type scale is the most ... | PersonalityTraitsinLargeLanguageModels |
One: Latent Flow Auto-EncoderStage Two: Diffusion Model𝐟<=𝐦<=…DDPM Forward Process…DDPM Reverse Process%𝐟<=’𝐦<=𝑥:Encoder Φ𝑧:“pick up and throw”Pretrained BERT𝑦𝑒𝜖,𝜖,𝜖,𝜖,……………𝐧∼𝒩(𝟎,𝐈)Decoder ΩWarpEncoder Φ𝑚𝑧2𝑥>?@̃𝑧Latent SpaceFlowPredictor | Conditional Image-to-Video Generation with Latent Flow Diffusion Models |
ImageBind’s capabilities that outperform
Image-aligned, self-supervised learning shows that the performance of our
model can actually improve by using very few training examples. Our model has
new emergent capabilities, or scaling behavior — that is, abilities that didn’t
exist in smaller models but appear in larger v... | ImageBind_ Holistic AI learning across six modalities |
Attention Block: We disable all QKV biases (Dayma et al., 2021). This exploits the scaling law
by removing a layer of computation, making the forward and backward pass somewhat faster, while
keeping the model size nearly constant. We find that we could decrease gradient costs by reducing
the number of attention heads (M... | CRAMMING-TRAININGALANGUAGEMODELONA SINGLEGPUINONEDAY |
Ori Yoran, Tomer Wolfson, Ben Bogin, Uri Katz, Daniel Deutch, and Jonathan Berant. Answering
questions by meta-reasoning over multiple chains of thought. arXiv preprint arXiv:2304.13007,
2023.
Murong Yue, Jie Zhao, Min Zhang, Liang Du, and Ziyu Yao. Large language model cascades with
mixture of thoughts representation... | UNIVERSALSELF-CONSISTENCYFORLARGELANGUAGEMODELGENERATION |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.