Tyler Zhu

@tyleryzhu

PhD in seeing things @Princeton | prev @Meta Superintelligence Labs @GoogleDeepMind, @berkeley_ai | @SFGiants @warriors guy

Berkeley, CA

tylerzhu.com

Replies Reposts Media

Pinned
Tyler Zhu
@tyleryzhu
Nov 18, 2025
Today seems to be a fitting day for @GoogleDeepMind news, so I'm excited to announce our new preprint! Prior work suggests that text & img repr's are converging, albeit weakly. We found these same models actually have strong alignment; the inputs were too impoverished to see it!
11
Tyler Zhu
@tyleryzhu
11h
“The incentives may now be pointing in the direction of no longer sharing any promising research directions with the broader community, which would reverse centuries of traditions of open science and do serious long-term damage to the future of the field” - Terence Tao
Tyler Zhu
@tyleryzhu
20h
The irony of this drama would be that math becomes dead not by any AI model per se, but by researchers corrupting the very _ethos_ of academic collaboration that has kept math alive in spirit.
2
Tyler Zhu
@tyleryzhu
20h
The irony of this drama would be that math becomes dead not by any AI model per se, but by researchers corrupting the very _ethos_ of academic collaboration that has kept math alive in spirit.
Konsti Wohlwend
@konstiwohlwend
Sep 8
drama summary for those confused: - Aug 15th: Tristan Buckmaster & Levent Alpöge make progress on a few important math problems - they do NOT have a proof for the $1,000,000 Millenium Prize problem. BUT, they do claim to have a proof for a similar (non-Millenium) Navier Stokes
2
Tyler Zhu
@tyleryzhu
Apr 20
We'll be presenting this work on video-text alignment this week at #ICLR2026! Come check out our poster in Pavillion 4 #3405 in the Thursday AM session from 10:30am to 1pm. Happy to chat about PRH, repr learning, video understanding, or how vision plays into text models!
Tyler Zhu
@tyleryzhu
Nov 18, 2025
Today seems to be a fitting day for @GoogleDeepMind news, so I'm excited to announce our new preprint! Prior work suggests that text & img repr's are converging, albeit weakly. We found these same models actually have strong alignment; the inputs were too impoverished to see it!
2
Tyler Zhu
@tyleryzhu
Apr 7
We peeled back the curtains on how video models _really_ reason, and found that only the first few denoising steps matter! This lets us sample more efficiently from them if we only care about trajectory diversity, purely at test-time. Same budget, much stronger reasoning.
Kaleb Newman
@kalebnewman8
Apr 7
Video models surprisingly can solve mazes, but inconsistently. We understand little about how they reason, making it hard to use such abilities. We investigate the denoising process and find models commit to a plan early, letting us screen far more candidates for better perf. 🧵
00:00