Max Chen
@maximillianc_
research scientist @googleai, phd @columbia, ba @cornell
PhD Candidate ➡️ PhDone! Excited to share that after 3.5 years of internship experience and earning 4 degrees, I have successfully defended my PhD thesis @ColumbiaCompSci and started working @GoogleAI! Huge thank you to @Zhou_Yu_AI for her unwavering support throughout my PhD!
PhD Student ➡️ PhD Candidate 🥳
Colab Pro is now available for free for verified US students and faculty for one year.
Big news for data science in higher ed! 🚀Colab now offers 1-year Pro subscriptions free of charge for verified US students/faculty, interactive Slideshow Mode for lectures, & an AI toggle per notebook. Enhance teaching & learning in the upcoming academic year! Read all about it…
Sad to be missing EMNLP this week! Definitely reach out to @qbetterk if you want to chat about dynamic LLM benchmarking: aclanthology.org/2024.findings-… We find that frontier LLMs' performance on math and language reasoning benchmarks degrades even from simple variable perturbations.
I am attending EMNLP 2024 in person! Come and talk if you are interested in our recent work in robust LLM evaluation: VarBench(arxiv.org/abs/2406.17681). #EMNLP2024
[1/7] Pre-trained LMs can do in-context learning, but this is unexpected given the distribution shift between pre-training data and ICL prompts. What structures of pre-training data yield ICL? Check out our work “Parallel Structures in Pre-training Data Yield In-Context Learning”
Attending #EMNLP2023 and interested in LLMs’ planning capabilities? @xy2437 will be giving a talk on our data-free Monte Carlo Tree Search approach to dialogue policy planning! When: December 8, 11:15am Where: Conference Room West 3
Want to improve an LLM's planning ability with *ZERO data or training*🤯? Introducing GDP-Zero, a prompt-based MCTS algorithm that can improve base LLMs decision-making/planning ability without any downstream task data or training! #EMNLP2023 #EMNLP arxiv.org/abs/2305.13660
wild that i pay a guy $45 every month to buzz the side of my head for 15 min and it feels cheap just because i’m in nyc 🫠
OpenAI really had a “major outage” 30 min before our DARPA demo 🙃
Check out our latest work on zero-data dialogue systems — improving mixed-initiative dialogue policy planning for LLMs using Monte-Carlo Tree Search! GDP-Zero will be presented at EMNLP 2023. Congrats to our collaborators, @xy2437 and @Zhou_Yu_AI !!
Want to improve an LLM's planning ability with *ZERO data or training*🤯? Introducing GDP-Zero, a prompt-based MCTS algorithm that can improve base LLMs decision-making/planning ability without any downstream task data or training! #EMNLP2023 #EMNLP arxiv.org/abs/2305.13660
was tabling for my club on campus today and fully asked a tenured professor if he was a grad student 🥲
Text Diffusion 🤝 Paraphrasing 🤝 Guidance = ✨Textual style transfer✨ Introducing ParaGuide, a diffusion-based framework for plug-and-play style transfer with off-the-shelf models! arxiv.org/abs/2308.15459🧵
Come check out our poster on mixed-initiative dialogue generation! I’ll be at poster session 3 at 11am. See you there! #ACL2023NLP
Is prompting *the end* 🤯 of fine-tuning (small) dialogue models?! ...well, maybe. We discuss this in our new @aclmeeting paper, where we directly swap out fine-tuned conditional generation dialogue models with LLMs! #ACL2023NLP #ACL #ConvAI #NLProc arxiv.org/abs/2305.04147
finally made it to toronto! all it took was a couple of canceled flights and a nine hour road trip with @MingyangKevinZh :) #ACL2023NLP