rohan anil

@_arohan_

co-founder, board @coreautoai - and continuing to aspire to understand deep learning.

Joined December 2017

Pinned
rohan anil
@_arohan_
Apr 19
It turns out multi step backpropaganda is better. paper has a beautiful way of improving backpropagation. One iteration cleanly gets us backprop, multiple iterations get us a preconditioned update.
rohan anil
@_arohan_
Apr 19
Replying to @LinYorker @ryu0000000001 and @weijie444
arxiv.org/abs/2106.06199 Same update here
rohan anil
@_arohan_
1h
I thought I loved optimization but now I am falling back in love with distributed systems engineering.
rohan anil
@_arohan_
1h
No no, Levent is a baller at Ant. He is not stuck
Dan Roy
@roydanroy
2h
For those who don't know Morse code, Levent is saying the following: "Help. I'm stuck here in pretraining at Anthropic and they won't let me out until I hit my perplexity quota. I keep trying to do ridiculous things to alert my friends and family outside, like tweeting solutions
rohan anil
@_arohan_
3h
Intelligence only shines through with good model behavior.
rohan anil
@_arohan_
Aug 11
So funny reading the reasoning.
Alexander Panfilov
@kotekjedi_ml
Aug 11
Replying to @kotekjedi_ml
But we also took a chance to have a look at some in-the-wild scheming, reward seeking, etc. examples, and dumped it in appendix. 1) Summarizer unfaithfulness Reasoning summaries often omit important information from the original trace. Here, Opus 4.8 realizes it knows the