events

Rob and Cosmin go over "Full Bandwidth Transformer"

The Deep Learning Study Group (San Francisco) meets online to discuss the paper "Full Bandwidth Transformer" from Microsoft Research. Rob and Cosmin lead the session, working through the paper and the author's related Twitter thread, which

Starts

2026-09-23T01:30:00+00:00

Organizer

Deep Learning Study Group (San Francisco)

Location

Online event

The Deep Learning Study Group (San Francisco) meets online to discuss the paper "Full Bandwidth Transformer" from Microsoft Research. Rob and Cosmin lead the session, working through the paper and the author's related Twitter thread, which speculates about a rumored architecture from OpenAI. A set of companion readings is listed: NextLat and its hierarchical follow-up, Almost Free State Prediction Separation, Google's Recirculation paper, and MIT's work on pretraining recurrent nets without recurrence. The gathering is aimed at people who want to read and pick apart current machine learning research together.