Narges Razavian (@narges_razavian) 's Twitter Profile
Narges Razavian

@narges_razavian

@[email protected]
Assistant Prof @NYULangone. AI in healthcare. Tweets about AI, fairness, medicine and science. Prev @LTIatCMU @CILVRatNYU.
All views my own

ID: 975825911961608192

linkhttp://razavian.net/ calendar_today19-03-2018 20:06:42

1,1K Tweet

2,2K Followers

1,1K Following

Scientific American (@sciam) 's Twitter Profile Photo

Here are seven fiction and nonfiction books Scientific American editors recommend from the past few months. They involve broken hearts, killer robots and epic failed experiments trib.al/r8TN0vQ

Alan Karthikesalingam (@alan_karthi) 's Twitter Profile Photo

Our new work in Nature Medicine Using generative models in path, radiology and derm - we can create synthetic training data to ⬆️ AI fairness & robustness, including underrepresented groups. Work w/ a great team Google DeepMind Google AI Google Health nature.com/articles/s4159…

Nathan Godey (@nthngdy) 's Twitter Profile Photo

🤏 Why do small Language Models underperform? We prove empirically and theoretically that the LM head on top of language models can limit performance through the softmax bottleneck phenomenon, especially when the hidden dimension <1000. 📄Paper: arxiv.org/pdf/2404.07647… (1/10)

🤏 Why do small Language Models underperform?

We prove empirically and theoretically that the LM head on top of language models can limit performance through the softmax bottleneck phenomenon, especially when the hidden dimension &lt;1000. 

📄Paper: arxiv.org/pdf/2404.07647…
(1/10)
Clémentine Fourrier 🍊 (@clefourrier) 's Twitter Profile Photo

New: Open Medical LLM Leaderboard! 🩺 In basic chatbots, errors are annoyances. In medical LLMs, errors can have life-threatening consequences 🩸 It's therefore vital to benchmark/follow advances in medical LLMs before thinking about deployment. Blog: huggingface.co/blog/leaderboa…

Yann LeCun (@ylecun) 's Twitter Profile Photo

🥁 Llama3 is out 🥁 8B and 70B models available today. 8k context length. Trained with 15 trillion tokens on a custom-built 24k GPU cluster. Great performance on various benchmarks, with Llam3-8B doing better than Llama2-70B in some cases. More versions are coming over the next

🥁 Llama3 is out 🥁
8B and 70B models available today.
8k context length.
Trained with 15 trillion tokens on a custom-built 24k GPU cluster.
Great performance on various benchmarks, with Llam3-8B doing better than Llama2-70B in some cases.
More versions are coming over the next
Isaac Kohane (@zakkohane) 's Twitter Profile Photo

Is AI-driven clinical-grade search of pathology images/slides ready for prime time? NEJM AI a case study with four leading programs suggests not yet. ai.nejm.org/doi/full/10.10… Accompanying editorial provides context and roadmap ai.nejm.org/doi/full/10.10…

Chunting Zhou (@violet_zct) 's Twitter Profile Photo

Introducing *Transfusion* - a unified approach for training models that can generate both text and images. arxiv.org/pdf/2408.11039 Transfusion combines language modeling (next token prediction) with diffusion to train a single transformer over mixed-modality sequences. This

Introducing *Transfusion* - a unified approach for training models that can generate both text and images. arxiv.org/pdf/2408.11039

Transfusion combines language modeling (next token prediction) with diffusion to train a single transformer over mixed-modality sequences. This