Emily Li @EmilyLiJiayao Twitter profile

Pinned Tweet

Emily Li

3 months

Super excited to introduce 🌳Acadia ( @AcadiaAI ) Playground, an interpretable data exploration tool to understand your evaluation data’s quality and help unlock insights into model performance using AI! 🧵

12

21

149

Last Seen Profiles

@rokhaya_sa

@swsenfhr

@keito_nagumo

@YesilayUrfa

@Go0Oseng

@blssmsmatcha_

@shepherd_fx

@ghilo_leni

@inga_no_nagare

@espn_chan

@animetv_jp

@ZunzuneguiBHI

@carlaalopezzzzz

@Zone3Rari

@soobphoric

@ahmedah50319026

@philippebarnes

@LGBissenile

@louisvuittoine

@EvelyneStella14

@NirajNiraj28490

@DacaAve

@mohannadaama

@ascfr5431

@chalana1963

@DMMatt100

@aballadin

@jahchattin

@culturechallans

@phu15407

@rosencrantz

@taku_nibguys_jp

@clt_news

@AshtonChains

@Me1122_3344

@pokibb

Emily Li

@EmilyLiJiayao

3 months

was wondering why my disk was so full that even git wasn't working and then realized that i have 77 GB of huggingface models cached locally oops

2

1

47

Emily Li

@EmilyLiJiayao

7 months

super thrilled to join the @Contrary squad as a VP and work with so many brilliant & fun ppl!

Contrary

@contrary

8 months

Thrilled to welcome our newest cohort of Venture Partners to the Contrary family! With nearly 1300 applications, this year was our most competitive yet. We’re excited to work with you all to meet and invest in the next generation of exceptional founders and companies. We also…

8

10

89

0

26

Emily Li

@EmilyLiJiayao

2 years

from last minute late night ideas to fruition, the beautiful Figma offices to the inspiring ppl. true thanks @hackclub and the Assemble team for making things happen! #assemble22 #sf

2

0

19

Emily Li

@EmilyLiJiayao

2 months

what are good "chat with a large code base" tools out there?

5

0

11

Emily Li

@EmilyLiJiayao

3 months

not swes now needing to add “human” to their linkedin job titles @cognition_labs

0

1

8

Emily Li

@EmilyLiJiayao

5 months

not tagged but excited to have worked on this @AGIHouseSF !

Alex Reibman 🖇️

@AlexReibman

5 months

3/ Hierarchical semantic clustering Clustering scheme that generates an interconnected hierarchy that links ideas together into a single post Consolidate your notes into a blog post 🥇First place @JvNixon @_nathanmarquez_ @zvhgpyxqtnys

1

2

26

2

0

9

Emily Li

@EmilyLiJiayao

4 months

the most sf imagery from today is seeing two ppl squeeze into a waymo front seat and another waymo blow up from fireworks 😳 anways.. happy lunar new year!🧧

1

0

7

Emily Li

@EmilyLiJiayao

1 year

Met @karpathy @hackwithtrees !! Come say hi if ur here :)

1

0

8

Emily Li

@EmilyLiJiayao

3 months

Super glad to be working on this with @_nathanmarquez_ !

0

8

Emily Li

@EmilyLiJiayao

3 months

what are some (better) LLM eval datasets you trust?

2

0

7

Emily Li

@EmilyLiJiayao

3 months

This is our first of many steps towards bringing interpretability into datasets and evals of growing quantity, complexity, and modalities. We want to make it easy to unlock high quality signal from the data for many LLM + multimodal applications. 6/6

Acadia AI

Data-Driven Explainability for AI Models

www.acadia-ai.com

2

0

7

Emily Li

@EmilyLiJiayao

8 months

stranded at the airport at 3am is the prime time to ship 😚

0

7

Emily Li

@EmilyLiJiayao

7 months

imagine @sama joining X or Anthropic

1

0

7

Emily Li

@EmilyLiJiayao

2 years

wanna play no-contact hologram style Tic Tac Toe? check out HoloTicTacToe open-sourced at (initially built for Assemble workshop @hackclub )

1

0

7

Emily Li

@EmilyLiJiayao

10 months

my brain at the mall with no context

1

0

6

Emily Li

@EmilyLiJiayao

2 months

when OpenDevin keeps on reading the same file over and over🤨

0

5

Emily Li

@EmilyLiJiayao

6 months

and it continues…

Emily Li

@EmilyLiJiayao

10 months

my brain at the mall with no context

1

0

6

1

0

6

Emily Li

@EmilyLiJiayao

6 months

60% of yc s23 were AI companies. what abt w24?

5

0

6

Emily Li

@EmilyLiJiayao

8 months

it’s crazy how bad and unclear the openai docs could be given the amount of users they have

0

6

Emily Li

@EmilyLiJiayao

2 months

Here's a new SOTA text-to-image eval metric that's much better at complex compositional reasoning than current ones (e.g CLIPScore, PickScore)! We also show that it generalizes to video/3d evaluation + released a comprehensive t2visual meta-eval metrics benchmark. Great to have…

Zhiqiu Lin

@ZhiqiuLin

2 months

In text-to-image generation, evaluating how well the generated image matches the prompt is a major challenge. We address this with VQAScore: a SOTA metric that significantly surpasses CLIPScore, PickScore, ImageReward, TIFA, and more! VQAScore works especially well on complex…

4

39

191

0

1

6

Emily Li

@EmilyLiJiayao

4 months

AI companies: introducing our new talented 👏 brilliant 👏incredible👏amazing👏show stopping👏spectacular👏never the same👏model Also AI companies: you cant use it yet

0

1

6

Emily Li

@EmilyLiJiayao

3 months

🗃️ Combine "Topics" of choice to filter and inspect individual datums 🧐 Select a model of interest, toggle on failure case mode, log, and visualize where failure cases occurs 2/6

1

0

5

Emily Li

@EmilyLiJiayao

5 months

@khoomeik @ArYoMo i'm curious--how are you baselining with gpt4v exactly? inputting screenshot & directly prompting it to output observation, thought, and action? i usually find gpt4v to be better at relative spatial reasoning/spitting out img descriptions

1

0

4

Emily Li

@EmilyLiJiayao

5 months

is infeasibility an indicator of inefficiency?

0

5

Emily Li

@EmilyLiJiayao

1 year

Week 1 ✅ in the beautiful Austin TX. ft some amazing ppl and Archie the owl

1

5

Emily Li

@EmilyLiJiayao

3 months

🛝You can define a custom set of task-specific "Topics" of interest, and Acadia Playground visually decomposes a target datasets' content into these categories 🔍 Explore dynamic embedding views of your data points--either embedded by overall semantics or “Topic” slices 1/6

1

0

5

Emily Li

@EmilyLiJiayao

9 months

pulling out a weekend project from a few mos ago... Fireo🔥, a neural net tensor shape debugger! - Useful print statements only - Only needs pseudo input + model class - No more hours spent manually tracing through shapes in your dl model dev workflow

GitHub - emilyjiayaoli/fireo: Model shape debugger for torch. Think torch.summary but better

Model shape debugger for torch. Think torch.summary but better - emilyjiayaoli/fireo

github.com

0

4

Emily Li

@EmilyLiJiayao

1 month

and so happened to be neighbors without ever knowing!! you’re cooler 🩷

emily zhang

@emilyzsh

1 month

love meeting online twitter friends irl, makes the world feel so small 🩷 @EmilyLiJiayao you’re so cool!!

0

8

0

4

Emily Li

@EmilyLiJiayao

3 months

@AcadiaAI Playground is multimodal! We used it to analyze 🖼️ Winoground (VLM image caption matching task) 💻 HumanEval (LLM code generation task) More details coming soon :) 3/6

1

0

4

Emily Li

@EmilyLiJiayao

1 year

:(

0

4

Emily Li

@EmilyLiJiayao

11 months

what i’ve more than anything else this summer: ignorance is bliss

1

0

4

Emily Li

@EmilyLiJiayao

10 months

current fastest route to agi feels like a data / continual learning problem

0

3

Emily Li

@EmilyLiJiayao

11 months

now redirects to @xai website instead of @OpenAI 's chatgpt as of today 💀

2

1

4

Emily Li

@EmilyLiJiayao

1 year

@itsandrewgao the swin transformer for example. also, although the naive attention’s work is in order n^2, multi-headed attention/parallelize-ability makes the span closer to linear or logn.

0

2

Emily Li

@EmilyLiJiayao

1 year

data efficient & smaller models >>

elvis

@omarsar0

1 year

JUST IN: Meta AI introduces LLaMA, a 65B parameter LLM. LLaMa only relies on publicly available data and outperforms GPT-3 on most benchmarks despite being 10x smaller.

28

338

2K

1

0

3

Emily Li

@EmilyLiJiayao

4 months

good day waking up to Sora and V-JEPA

0

3

Emily Li

@EmilyLiJiayao

8 months

@ethanweii @clairebookworm1 no wayy i also got sick right when after the nyc wknd 😍

1

0

3

Emily Li

@EmilyLiJiayao

5 months

@sayakmighty fr i filled out their google form and never heard back

0

1

Emily Li

@EmilyLiJiayao

1 year

this year felt like two years in one. feb 22 doesn't sound like too long ago but when I look back at pictures, it feels like so long ago

0

2

Emily Li

@EmilyLiJiayao

3 months

If you’re interested in using this for a particular dataset/use case, let us know here: 5/6

1

0

3

Emily Li

@EmilyLiJiayao

3 months

@AcadiaAI Playground can also be used for: - Cross comparison of various models to evaluate the best model for your use case - Identify and target weaknesses in your dataset distribution (such as duplication or misrepresented categories), inform better data curation 4/6

1

0

3

Emily Li

@EmilyLiJiayao

2 years

and it's hereeeee @hackclub

Inside a High School Hackathon: Assemble, August 2022

42 hours, 183 teenagers, & mischief to be made. Assemble was the first major high school hackathon since the pandemic, organized by a team of teenage Hack Cl...

www.youtube.com

0

3

Emily Li

@EmilyLiJiayao

2 years

@itsandrewgao yea and i wonder how of it is scaling parameters/more training data vs consequential architecture improvements

0

2

Emily Li

@EmilyLiJiayao

1 year

new competitor AI org? 🤔

Elon Musk

@elonmusk

1 year

BasedAI

7K

4K

52K

2

0

3

Emily Li

@EmilyLiJiayao

8 months

yay i was right

0

3

Emily Li

@EmilyLiJiayao

6 months

@aidenybai @milliondotjs @ycombinator congrats aidennn

0

Emily Li

@EmilyLiJiayao

7 months

how is making perhaps >90% of @openai + @sama + @gdb join msft any good for ai safety? smh

1

0

3

Emily Li

@EmilyLiJiayao

1 year

when reading research papers, isn't it so annoying to click the link to see the citations but then have to scroll all the way back up or am i missing out on something?

1

0

3

Emily Li

@EmilyLiJiayao

2 years

demo day was awesome. cv has always been extremely interesting to me but I had never first-hand witnessed how inspiring it may also be for others until today, esp by it’s real world applications that bridge imaginative sci-fi with reality. 🦾 #gangstaminecraft

0

3

Emily Li

@EmilyLiJiayao

8 months

200 on clip is crazy 😱. there’ll probably be a lot more on nerfs / 3d vision once 2d vision is solved (alr feels like it has by gpt4v but opensource still has a long way to go)

Lucas Beyer (bl16)

@giffmana

8 months

ICLR submissions are online: Looks like there's: - ~700 with diffusion in it, - less than 100 with nerf, - ~900 LLM - ~100 chatgpt (8 bard, 16 claude) - vs ~170 llama (yay) - ~200 clip (but not "clipping") - ~200 NLP - ~750 vision(!?)

17

59

419

0

2

Emily Li

@EmilyLiJiayao

1 year

all the bad media that starship orbital attempt gets makes me sad. it's such a huge milestone. this rapid iterative process should be encouraged.

1

0

3

Emily Li

@EmilyLiJiayao

2 years

some more pics from my workshop at assemble 😇 (thanks to @kunalbotla for the 📸)

0

2

Emily Li

@EmilyLiJiayao

9 months

I asked dalle3 to generate myself wearing a sweatshirt I used to wear a lot. and no i don't actually look like this...

1

0

2

Emily Li

@EmilyLiJiayao

1 year

@YiMaTweets hmm feels like it's more prior ⊆ latter. classification/recog. are discriminative tasks whose objective is to learn conditional prob distribution P(X|Y) aka decision boundaries, which is a subset of generative models that learn a joint distribution P(X,Y) where we sample from

1

0

2

Emily Li

@EmilyLiJiayao

4 months

@karanganesan @sama @southpkcommons it was awesome meeting you!

0

2

Emily Li

@EmilyLiJiayao

3 months

holyy the 32 raptors are beautiful

0

1

Emily Li

@EmilyLiJiayao

2 years

twitter >> tiktok >> insta content suggestion algorithm-wise (imo) unasked for review - 🧵

1

0

2

Emily Li

@EmilyLiJiayao

2 years

new twitter!

1

0

1

Emily Li

@EmilyLiJiayao

1 year

@akbirthko awesome, this was what i was leaning towards. but in this case, what is the point of even having different heads if their end result is concatenated together anyways b4 the linear layer? don't the q, k, v operate independently between the different hidden dims anyway?

1

0

2

Emily Li

@EmilyLiJiayao

1 year

@HaoliYin I've actually thought about this b4 haha! I feel like generating accurate and robust 3d mesh/point cloud/surface is pretty difficult and unsolved problem.

1

0

2

Emily Li

@EmilyLiJiayao

1 year

but then again...it's the media being the media

0

2

Emily Li

@EmilyLiJiayao

9 months

@calixo888 go to settings and enable web search beta

0

2

Emily Li

@EmilyLiJiayao

2 years

imagine if there exists an arXiv that consists of papers/logs of project ideas that failed or went nowhere. that way, actual innovation might progress much faster.

2

0

2

Emily Li

@EmilyLiJiayao

1 year

this aged so well

0

2

Emily Li

@EmilyLiJiayao

2 years

what i learned this past week: - i love with all of my heart - dunning kruger's effect is too real - context switching is helpful for project fatigue

0

2

Emily Li

@EmilyLiJiayao

9 months

reliable models only result from robust evaluations and metrics. what are (relatively) non-subjective ways to eval generative models or is that just its nature?

0

2

Emily Li

@EmilyLiJiayao

2 years

these are so hard to remember🥲

Pradeep Pandey

@Div_pradeep

2 years

Visual Studio Code shortcuts Cheatsheet⚡️⚡️

55

598

3K

0

2

Emily Li

@EmilyLiJiayao

1 year

@s1wase yes, and it runs at an acceptable frame rate!

1

0

1

Emily Li

@EmilyLiJiayao

11 months

@HaoliYin unfortunate but true😅

0

2

Emily Li

@EmilyLiJiayao

1 year

@clairebookworm1 @karpathy @_neelr_ @sracha_z @SophiaPung yass claire

0

2

Emily Li

@EmilyLiJiayao

1 year

Getting sick while living alone makes me miss my parents so much more 🥲

0

2

Emily Li

@EmilyLiJiayao

1 year

been waiting since aug 2020 😭

Everyday Astronaut

@Erdayastronaut

1 year

IT IS OFFICIAL!!! The world’s biggest, most powerful rocket ever, will attempt its first launch on the morning of Monday, April 17th!!! We have our stream ready to go with some amazing views and incredible audio to help bring you along!

157

765

7K

0

1

Emily Li

@EmilyLiJiayao

2 years

computer vision

0

1

Emily Li

@EmilyLiJiayao

1 year

@tengyuma @HongLiu9903 @zhiyuanli_ @dlwh @percyliang @StanfordAILab @stanfordnlp @StanfordCRFM @Stanford pytorch compatibility would encourage usage!

1

0

2

Emily Li

@EmilyLiJiayao

7 months

@gdb increase in RPD limits; random server errors occur at times; browser version feels like it’s much more willing to describe; log probs would be great!

1

0

1

Emily Li

@EmilyLiJiayao

2 years

@MarioKrenn6240 Due to the influx of papers, bec it's rare for any AI researcher to have read every single paper in their relative subdomain, there're undoubtedly lots of overlapping "novelties." So even just having a systematic approach for tracking defs and training paradigms would be helpful

0

2

Emily Li

@EmilyLiJiayao

5 months

currently playing with @runwayml 's gen-2 video gen models -- definitely something going on "A baker pulling freshly baked bread out of an oven in a bakery" send in some prompts👇

1

0

2

Emily Li

@EmilyLiJiayao

1 year

wouldn't it be nice if we could also plot graphs in w&b after the model is trained? sometimes i just forget to run a cell

0

2

Emily Li

@EmilyLiJiayao

7 months

@HaoliYin gg he’s converting

1

0

1

Emily Li

@EmilyLiJiayao

1 year

no hate but i feel like @scale_AI should be able to sponsor travels to their hackathons in sf… like even university hackathons can 😭:/

0

1

Emily Li

@EmilyLiJiayao

5 months

@HaoliYin @alexfmckinney love that for us

0

1

Emily Li

@EmilyLiJiayao

2 years

today i ran into a symposium at the CMU robotics institute while exploring campus. interesting work + had very nice convos abt language and vision w/ these grad students who presented at CVPR & ICML

1

0

1

Emily Li

@EmilyLiJiayao

3 months

If you’re interested in using Acadia Playground for a particular dataset/use case, let us know! 5/6

1

0

1

Emily Li

@EmilyLiJiayao

2 years

@kkevinwuu hii

1

0

1

Emily Li

@EmilyLiJiayao

10 months

@CDuong04 offers will flow right in with this immaculate setup

0

1

Emily Li

@EmilyLiJiayao

5 months

@HaoliYin amazing haoli soo proud !

0

1

Emily Li

@EmilyLiJiayao

5 months

@HaoliYin @alexfmckinney i say try the former, if not good enough then the latter, we def have stronger text embedding models than vision. also i'm interested to see how close CLIP img encoder embeddings are to img->description->CLIP text embeddings, perhaps that could be a finetuning objective for CLIP

1

0

1

Emily Li

@EmilyLiJiayao

2 years

@idrick @EliyaTheFirst @karpathy let's hope this is not like the space race

1

0

1

Emily Li

@EmilyLiJiayao

2 years

why is this soo true...is definitely something that wastes a lot of my time

Andrej Karpathy

@karpathy

2 years

The software engineering aspect of deep learning repos I've been watching closely is how they store, catalogue, override, manage and plumb hyperparameter configs. Have come to dislike argparse, YAMLs (too inflexible), and fully enumerated kwargs on classes/defs. Any favorites?

193

170

2K

0

1

Emily Li

@EmilyLiJiayao

5 months

@HaoliYin @runwayml lmaoo

0

1

Emily Li

@EmilyLiJiayao

8 months

@aidenybai @BrownUniversity haha omg aiden i love this

0

1

Emily Li

@EmilyLiJiayao

5 months

@_jasonwei the pain is real

0

1

Emily Li

@EmilyLiJiayao

2 years

Apart from intention-based factors such as company direction and algorithm design, it’s interesting to note the dissimilarity of the current knowledge transfer ability bet. natural language-based (twitter) vs vision/img/vid based (insta, tiktok) mediums. language is clearly ahead

2

1

Emily Li

@EmilyLiJiayao

11 months

guess it’s

Blog

Read about the latest announcements from xAI including Grok, Grok-1, and the PromptIDE.

x.ai

0

1

Emily Li

@EmilyLiJiayao

2 years

ancient computer vision and “machine vision” books at the Hunt library @CarnegieMellon

0

Emily Li

@EmilyLiJiayao

1 year

quick technical question: does increasing # of heads in the transformer MSA increase param count? i've gotten mixed answers. if this is implementation dependent then is there a standard? for most implementation i've seen (pytorch & swin) the answer seems to be a no.

3

0

1

Emily Li

@EmilyLiJiayao

7 months

@aidenybai yass aiden let's gooo CONGRATSS!!

0

1

Emily Li

@EmilyLiJiayao

1 year

@aidenybai Congrats!!

0

1