This is kinda interesting. ChatGPT gets tricked by some of the same color illusions as we do.

✨Excited to finally drop our new paper: SSMs “look like” RNNs, but we show their statefulness is an illusion🪄🐇 Current SSMs cannot express basic state tracking, but a minimal change fixes this! 👀 w/ jackson petty, Ashish Sabharwal arxiv.org/abs/2404.08819

Announcing the alpha release of torchtune! torchtune is a PyTorch-native library for fine-tuning LLMs. It combines hackable memory-efficient fine-tuning recipes with integrations into your favorite tools. Get started fine-tuning today! Details: hubs.la/Q02t214F0

I wish @pytorch supported complex neural networks too.

It's 45TB!

Google announces Med-Gemini, a family of Gemini models fine-tuned for medical tasks! 🔬 Achieves SOTA on 10 of the 14 benchmarks, spanning text, multimodal & long-context applications. Surpasses GPT-4 on all benchmarks! This paper is super exciting, let's dive in ↓

We'll be presenting our paper (arxiv.org/abs/2401.14469) at #ICLR2024 tomorrow! Check us out at Hall B, poster #16 (10:45 a.m. – 12:45 p.m). If you are presenting tomorrow too let me know.

I am so excited that xLSTM is out. LSTM is close to my heart - for more than 30 years now. With xLSTM we close the gap to existing state-of-the-art LLMs. With NXAI we have started to build our own European LLMs. I am very proud of my team. arxiv.org/abs/2405.04517

Excited to announce the launch of our new dataset website! 🚀 Check it out Visual Graph Arena at: vga.csail.mit.edu Our paper is currently under review and will be available after publication. Stay tuned! #MachineLearning

Attended #ICML2024 poster session today. Noticed just a few papers on CNNs! it seems everyone is shifting focus to Transformers.

Just received #NeurIPS2024 reviews. They are so brutal I don't even understand why! 🫠

I was highly sceptical of the paper “Large language models are zero-shot time series forecasters.“ from NYU when it came out. Somehow the paper was published at NeurIPS, but as often the reality and the claims in transformer and LLM papers for time series are vastly different.

Trying my best to create a vintage effect for a chart in Matplotlib! 😄 I want it to clearly look like it's from an old paper, and it is not our result. I hope the reviewers wouldn't think it's unprofessional!