Our LabelBench work has been accepted to the DMLR journal🎉 Super smooth experience and highly recommended for anyone working on data centric ML work.

Check out labeltrain.ai: our broader set of label efficient learning work
More on LabelBench: x.com/jifan_zhang/st…

(1/4) Announcing FlashInfer, a kernel library that provides state-of-the-art kernel implementations for LLM Inference/Serving.

FlashInfer's unique features include:
- Comprehensive Attention Kernels: covering prefill/decode/append attention for various KV-Cache formats (Page

Excite to announce that our work on posterior sampling for pure exploration linear bandits will be at #AISTATS2024 !

arxiv.org/pdf/2310.06069…

Our method extends top-two sampling for MAB and hopefully opens up new promising directions for exploration in RL.

New paper on label-efficient supervised finetuning of LLMs.

We address the expensive prompt annotation cost by humans/proprietary LLMs, saving as much as 50% on FLAN V2.

Paper: arxiv.org/abs/2401.06692
Work led by: Jifan Zhang Yifang Chen Gantavya Bhatt Arnav Das
1/

In NOLA for NeurIPS Conference! Present two papers and one workshop paper. Excited to chat about RL, control, and robotics. DM if you want to meet up : )
1. Optimal Exploration for Model-based RL in Nonlinear Systems (arxiv.org/abs/2306.09210). Thu 10:45-12:45pm #1507. TL;DR: Not all

I'm looking to recruit PhD students to join Georgia Tech to work with me on data aspects of core machine learning starting Fall 2024. If interested, please consider applying by the December 15 deadline to the ML PhD program or the CS PhD program (links on my webpage).

✨ I'm interviewing creative writers who use AI in their practice! ✨

Interviews will also include a writing session - if you are interested please fill out the form:

forms.gle/T9hU1eGACLGfHA…

Active Learning studies how AI best direct humans to provide feedback and annotation. Yet these algorithms have wildly different performances on new datasets, making it impossible to pick the best algo to use in practice. We now have a solution (appearing at NeurIPS 2023) 1/7

Excited to join Bourns College of Engineering at UC Riverside UCR ECE UC Riverside! I’m also actively hiring Ph.D. students for Fall 2024. Please see yinglunz.com for more details 😀

I am excited about this work and see a large potential of further extending this label efficient framework on more complicated scenarios (OOD, generative task, multimodal)

Hello world, we are Tikz Unicorn. Check out our latest tee with unicorns generated by GPT4 at tikzunicorn.com!

Wow! Maybe I should generate some poem for my research topic and put that as the main page for my personal website. 🤩

#NeurIPS2022 When can we find the Nash equilibrium and overcome the curse of multiagent in offline multi-agent RL? Come to our poster Hall J #539 today at 4pm and Hall J #740 on Thursday at 4pm for more details and Q&A!

Paper: arxiv.org/abs/2201.03522 & arxiv.org/abs/2206.00159

Reviewing 5 8-page paper within two weeks seems reasonable, but reviewing 8-page “extended abstract” with 20-30 page “actual theory paper” at appendix is too much 🤯

Me wrapped myself in a fluffy quilt refusing to wakeup

Come to our poster #1220 Hall E for more details and Q&A! #ICML2022

Multi-task representation learning is powerful, but not all source tasks are relevant to the target task. So how to select the most relevant tasks to improve efficiency?
A: use Active Learning!
Our ICML paper: arxiv.org/abs/2202.00911

Andrew is an amazing leader for our project !

It is sad that we start our new quarter (and new year) like this. But I would still proudly recommend Allen school to all my friends since I sincerely feel the inclusiveness and belongingness here ❤️

Our next talk:
09/15: Chen-Yu Wei (U Southern California)
'Learning Infinite-horizon Average-reward MDPs with Linear Function Approximation'

For details, please see the website:
sites.google.com/view/rltheorys…