MS CS at Columbia | Applied Scientist @ Microsoft, Research Intern at @ETHZurich NLP Lab | Graduate @iitroorkee. Exploring Inference time optimisations- LLMs
Lots of discussion out there about our next model(!), so I wanted to give an early look as soon as possible. Introducing Gemini 4 Argon!
It shows frontier performance in complex workflows, cyber defense and software engineering. Teams are using it extensively at Google, from coding to quantum computing, great feedback.
Here’s a look at the benchmarks:
Great session in my Agent for Work seminar with @lauraham150, @dan_cahana from @notablecap. Students really appreciated the market and fundraising insights you shared. Big thanks for joining us!
Drop by our ICLR workshop tomorrow on Building Trust in LLMs and LLM Applications! We’ll cover topics ranging from guardrails to explainability to regulation and more. We'll be in Hall 4 #6:
https://t.co/ODxQ8id7OE
2025 is the year of distillation
Knowledge Distillation (Hinton et al TEN years ago) is finally getting its moment as the entire Gemma 3 family of model was pre and post trained using distillation
What is distillation, who's doing it, and when should you use it? a 🧵 👇
If you want to do a PhD with me on Multi-Agent LLMs (see our recent #NeurIPS2024 paper on GovSim https://t.co/rLTKFbMzDB), I highly recommend you apply by *Nov 14* to Cooperative AI PhD Fellowship! Up to $40K/year w/ other benefits. Email me if you want to iterate your proposal.
If you attending ACL please reach out, I'd love to chat/learn about Question Answering, Information Retrieval, and LLMs :)
I am also looking for internship roles for summer '25!
🌟 Excited to share the latest work to be presented at ACL 2024!🌟
CausalCite, an alternative metric where we quantify paper impact by its causal effect on followup works. We ask the counterfactual question, "Had this paper not existed, what would happen to the followup papers?"
Paper link - https://t.co/i6iPGOcfXH
Zhijing will be presenting the poster in Poster Session 3 (August 13 12:15), feel free to drop by if you are attending in person!
Excited to introduce our 3 papers at #ACL2024NLP on #Causality, #NLProc & #Music:
➡️Competition of Mechanisms: Tracing How Language Models Handle Facts and Counterfactuals
➡️CausalCite: A Causal Formulation of Paper Citations
➡️Moûsai: Efficient Text-to-Music Diffusion Models
Unsatisfied with citations as a measure of paper quality? Check out CausalCite, an alternative metric where we quantify paper impact by its causal effect on followup works. We ask the counterfactual question, "Had this paper not existed, what would happen to the followup papers?"
Unsatisfied with citations as a measure of paper quality? Check out CausalCite, an alternative metric where we quantify paper impact by its causal effect on followup works. We ask the counterfactual question, "Had this paper not existed, what would happen to the followup papers?"
Excited for our paper "Beyond Good Intentions: Reporting the Research Landscape of NLP for Social Good (NLP4SG)" at #EMNLP2023 Findings. We propose 3 NLP tasks: (1) detect if a paper relates to NLP4SG , (2) check which UN SDG it addresses, and (3) extract task & methods proposed.
🚨New paper🚨
Contrastively trained Vision-Language Model’s (like CLIP) are poor at compositional reasoning
Our new paper improves this:
“Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality”
https://t.co/LccQI4ojST
#GPT4 saved my dog's life.
After my dog got diagnosed with a tick-borne disease, the vet started her on the proper treatment, and despite a serious anemia, her condition seemed to be improving relatively well.
After a few days however, things took a turn for the worse 1/
After success with Images, diffusion models are also being used in the text domain now. In this blog, I explore some of the key ideas of papers using Continuous Diffusion Models for NLP.
https://t.co/6Ku8r9wKBw
#NLP#diffusionmodels
pip install scikit-learn
It's easy to take for granted, but this single command gives you functionality I'd value at hundreds of thousands of dollars, if not more.
Not to mention amazing documentation that beautifully weaves guides and references.
Hats off to @scikit_learn
Announcing PyTorch-Ignite's v0.4.7 release 🔥
We have a new State Parameter Scheduler, improved loggers, additional features and bug fixes.
This time we had many new contributors so thank you to our community for their amazing contributions!
Read more:
https://t.co/IfTQy9SKh1