Vy Vo

Shailee Jain @shaileeejain · Aug 7, 2023

🌟NEW PREPRINT🌟
🤖-based🧠models are all the rage!

But,
- How do we interpret 🧠if 🤖are SO un-interpretable?
- How do we _test_ if🤖are good models of 🧠? (No, your prediction score is NOT the answer)

Our new paper (co-lead @vvobot) shows you how!
tinyurl.com/vojain2023

1/9

View quoted post on X

Sebastian Michelmann @s_michelmann · Jan 26, 2023

Excited to share our new preprint bit.ly/402EYEb with @mtoneva1, @ptoncompmemlab, and @manojneuro), in which we ask if GPT-3 (a large language model) can segment narratives into meaningful events similarly to humans. We use an unconventional approach: ⬇️

View quoted post on X

Mariya Toneva @mtoneva1 · Dec 1, 2022

Come over to the MemARI workshop at #NeurIPS2022 tomorrow Dec 2 to chat about how memory mechanisms should be designed in deep learning and what this field can learn from biological memory systems.

Schedule: neurips.cc/virtual/2022/w…
Website: memari-workshop.github.io

View quoted post on X

Omri Raccah @omriraccah · Nov 29, 2022

Transformers are limited in their ability to generalize over long temporal dependencies. In our @NeurIPSConf paper (#MemARI), we propose a framework for effectively integrating findings from human memory into LMs / transformers. (1/2)

arxiv.org/abs/2210.01869

View quoted post on X