Zhiyuan Jerry Lin

@msbernst Congrats, Michael!!

Thank you to everyone for your energy and enthusiasm in joining this adventure with me so far!

@CaoHancheng @EmoryUniversity @EmoryGoizueta Congratulations!! Atlanta would be fun!

Life update: Thrilled to announce I will join @EmoryUniversity @EmoryGoizueta Information Systems & Operations Management Department as an assistant professor in summer 2025, after a year as a postdoc at Microsoft Office of Applied Research!

@tianle_cai Why am I even surprised...

Just wrote a script to further investigate how the corpus used to train the gpt4o tokenizer is polluted by Internet scams. The results are quite interesting... 🤦‍♂️🤦‍♂️🤦‍♂️ gist.github.com/ctlllll/4451e9…

My kindest sympathies for the Bayesian Optimization folks

@infoxiao This you mean

How many parameters before we can use the name VLLM (very large) and ULLM (ultra-large)? I was an microelectronics major for undergrad and I chuckled when I learned Very-large-scale integration (VLSI) and Ultra Large-Scale Integration (ULSI). It is giving final-final-final.pdf

2024 PhD Internship opportunity! Join me and my amazing colleagues on Meta's Adaptive Experimentation team to work on Bayesian optimization, probabilistic modeling / Gaussian Processes, and sample-efficient decision making: metacareers.com/jobs/905634110…

Glad to see this conference has gotten noticeably bigger and better with more folks from even broader backgrounds 😃

Spotted @infoxiao at @CODEConference this year!

@5harad Congrats! Amazing to see this being officially published!

@MinaLee__ @UChicagoCS @DSI_UChicago @universeinanegg @ChenhaoTan Congrats!!

I'll be joining the University of Chicago (@UChicagoCS and @DSI_UChicago) as an assistant professor in the summer of 2024. 👩🏻✨ Excited to start the Communication & Intelligence group with Ari Holtzman (@universeinanegg) and Chenhao Tan (@ChenhaoTan)!

You knew that was coming: Code LLama !!! - Llama-2 tuned for code generation, debugging, etc - base models, Python-specific, and instruction-tuned. - 7B, 13B and 33B params models - Available with the same license as Llama-2 - Blog: ai.meta.com/blog/code-llam… - Paper: ai.meta.com/research/publi… - Code: github.com/facebookresear… - Models: ai.meta.com/resources/mode…

@DrJimFan It’s nice to see how much the “usual LLM agenda” has been added compared to when I first took this class 5 (or 6?) years ago…

An open-source course for your weekend learning sprint: CS224U, taught by Stanford prof Chris Potts. In addition to the usual LLM agenda, there're quite a few gems: - Classic ways for information retrieval: TF-IDF, BM25 ranking score, Pyserini (python tool), and IR metrics. Resist your knee-jerk instinct to reach for GPTs. Sometimes simpler methods are far more efficient for apps that don't need bigger guns. For example, @karpathy's arxiv-sanity.com uses TF-IDF for paper recommendation: github.com/karpathy/arxiv… - Diffusion for text: yes, it's an emerging research field. Diffusion simulates how a coherent sentence can materialize by denoising "fuzzy words" in parallel. Doesn't quite work well yet, but it could become an alternative to the dominant autoregressive way. - Systematic approaches for NLU system evaluation: probing, feature attribution, interventions. CS224U slides: web.stanford.edu/class/cs224u/ Code repo: github.com/cgpotts/cs224u/ OG post from @suzatweet

Finally I can start preying on predatory journals.

It just occurred to me that it's been a while since I used a non-dark-themed IDE. Is this the industry standard now? Miss my old Eclipse days (not really).

@anothercohen lmao this is amazing

Wow, Spendoso.com just saved our marriage

My wife just texted me this, I'm sleeping on the couch tonight

Isn't simplicity a merit instead of a drawback if there are theoretical and/or empirical evidence backing its validity? What can be done differently in future papers to avoid such criticism? Or perhaps such simple ideas are meant to be kept in a secret treasure chest 🫤

#NeurIPS2023 reviews are out. Once again our proposed methods/algorithms were criticized for being "too simple" and thereof lacks "novelty". I get it that novelty is an important aspect of scientific inquiries but sometimes such comments feels more like lazy reviews.

@GuanyaShi wow this is so bad that it almost looks like a troll if I didn't know this was a real review.

Today, I received a tragic review with the lowest possible rating for my NeurIPS paper. I was all prepared to enter the ring of scholarly debate and start doubting my research taste, but then I found:

Wah

@daveholtz @Airbnb Clearly we have realized that twitter might be the real customer service hotline...

tfw you arrive at your @airbnb and realize that the whole space is essentially just a large-ish bathroom that the host put a bed into