Gaurav

Applied Scientist @amazon | Creator of iNLTK

[Gaurav @arggaurav]

1). I'm glad to share that my paper on iNLTK library has been accepted at #emnlp2020's #nlposs workshop. I would like to thank @jeremyphoward, @fastdotai and the community - this could not have been possible without learning from you all.

[Gaurav @arggaurav]

Our work got accepted at NAACL 2025! 🥳

TLDR: Present tasks to see if LLMs have requisite representations to reason using "equivalence" & "inheritance”; Propose compositional representations where every token is composition of equivalent tokens across languages

@naaclmeeting

[Gaurav @arggaurav]

📢📢 New preprint alert on reasoning in multilingual LLMs !!

Humans create abstract concepts to reason consistently across and within languages. Ability to create abstract concepts is crucial for efficient learning,
knowledge sharing and updation of beliefs.

1/n

[Gaurav @arggaurav]

Preprint: https://t.co/GcdnMQsAw6

6/6

[Gaurav @arggaurav]

📢📢 New preprint alert on reasoning in multilingual LLMs !!

Humans create abstract concepts to reason consistently across and within languages. Ability to create abstract concepts is crucial for efficient learning,
knowledge sharing and updation of beliefs.

1/n

[Gaurav @arggaurav]

To enhance consistency across languages, we propose novel "Compositional Representations” where tokens are represented as composition of equivalent tokens across languages, with resulting conflict reduction (up to -4.7%) indicating benefits of shared LLM representations.

5/n

[Gaurav @arggaurav]

@NirantK @HasuraHQ Thanks Nirant! Filled the form!

[Gaurav @arggaurav]

Hi @RiTUAL_Lab - Thanks for the work which you've done for code-mixing with LINCE Leaderboard, but it seems to be inaccessible. Infact https://t.co/fHCPTzinQM itself seems inaccessible. It'll be great if you can look into it. Thanks!

[Sahil Bloom @SahilBloom]

The most powerful ideas in life:

[Sahil Bloom @SahilBloom]

The most powerful paradoxes of life:

[rishi @RishiBommasani]

The deadline to apply to PhD programs is almost upon us! I recall that as an applicant around this time of year, I wondered what I could do to improve my application in this final stretch. To pay it forward, here is my advice:

[Hamel Husain @HamelHusain]

Ok data science twitter, time to put in and upvote requests to enable Jupyter Notebook diffing on GitHub 👇🙏🤞

[Sam Bowman @sleepinyourhat]

Neat negative result spotted at #ACL2021:
I've seen a number of efforts that try to use MNLI models to do other classification tasks by checking whether the input entails statements like 'this is a negative review'. (1/...)

[Seoyeon Stella Yang @codestella]

CLIP+NeRF: DietNeRF🥗 is now available! *With only 8 images*, DietNeRF is capable of neural rendering quality novel view from a 3D model. You can check out more details here:
👉𝐆𝐢𝐭𝐡𝐮𝐛 : https://t.co/TONkoiQuSd
👉𝐒𝐭𝐞𝐚𝐦𝐥𝐢𝐭 𝐒𝐩𝐚𝐜𝐞 𝐃𝐞𝐦𝐨 :
https://t.co/H6sdvJtHqY

[Mathematical Optimization Society @math_opt]

The 2021 winners are Léon Bottou, Frank E. Curtis (@frank_e_curtis), and Jorge Nocedal (@nocedal)
for their paper

"Optimization Methods for Large-Scale Machine Learning" (https://t.co/Hmapr8uQgk)

[Google DeepMind @GoogleDeepMind]

Today with @emblebi, we're launching the #AlphaFold Protein Structure Database, which offers the most complete and accurate picture of the human proteome, doubling humanity’s accumulated knowledge of high-accuracy human protein structures - for free: https://t.co/vtBGmTkKhy 1/

[Shreyas Doshi @shreyas]

Sharing this excellent operating manual for leaders of fast-growing orgs, from @shishirmehrotra. This content is easily worth millions of dollars, if you pay attention to it & adapt it to your company's context. Seriously.

https://t.co/3idFzeWqqE

[ACMI Lab (CMU) @acmi_lab]

New @ #ICML2021: When a trained model fits clean (training) data well but randomly labeled (training) data (added in) poorly, its generalization (to the population) is guaranteed!

Paper: https://t.co/zc7cPpaf9m

by ACMI PhD @saurabh_garg67, Siva B, @zicokolter, & @zacharylipton