_arohan_ Profile Banner
rohan anil Profile
rohan anil

@_arohan_

Followers
24K
Following
25K
Media
931
Statuses
8K

all about training algorithms & efficiency. @AnthropicAI Ex: Meta (2025), Google Deepmind, Google Brain, Sibyl (2013-2024). Views are my own.

Joined December 2017
Don't wanna be here? Send us removal request.
@_arohan_
rohan anil
50 minutes
RT @miniapeur: I recommend looking at these.
Tweet media one
Tweet media two
0
48
0
@_arohan_
rohan anil
19 hours
Claude code makes me feel 10 years younger tbh.
4
0
67
@_arohan_
rohan anil
1 day
A Saturday reminder to all new followers that Shampoo stands for a preconditioner. It’s called Shampoo because thats what comes pre/before using a conditioner.
6
2
164
@_arohan_
rohan anil
2 days
With use of computer, docs, now llm. I think I am losing my ability to think with a pen and paper or at-least feel foreign to me. I wonder if I am losing access to some circuits in my brain to think better.
12
2
87
@_arohan_
rohan anil
2 days
Strange but worth a try!.
@Rainmaker1973
Massimo
3 days
Berrylicious coffee. [☕ dippacicoffeeco].
0
0
5
@_arohan_
rohan anil
3 days
As a field we probably should get rid of exploding offers.
4
0
76
@_arohan_
rohan anil
3 days
Tweet media one
0
5K
0
@_arohan_
rohan anil
3 days
Stop using adam, use SGD.
@2prime_PKU
Yiping Lu
3 days
Anyone knows adam?
Tweet media one
3
0
24
@_arohan_
rohan anil
3 days
RT @premqnair: I’ve joined Cognition to continue to work on the future of software engineering. I was employee #2 at Windsurf and have wor….
0
118
0
@_arohan_
rohan anil
3 days
Dang! This is very good!.
@lyang36
Lin Yang
4 days
Compared the first version in our paper, this code removes problem specific hints completely. It just works!.
0
1
20
@_arohan_
rohan anil
3 days
RT @lyang36: Code release! 🚀. Following up on our IMO 2025 results with the public LLM Gemini 2.5 Pro — here’s the full pipeline & general….
Tweet card summary image
github.com
Contribute to lyang36/IMO25 development by creating an account on GitHub.
0
44
0
@_arohan_
rohan anil
4 days
RT @alexalbert__: It's becoming more and more clear that Claude Code is the everything agent.
0
82
0
@_arohan_
rohan anil
4 days
RT @andrewwhite01: HLE has recently become the benchmark to beat for frontier agents. We @FutureHouseSF took a closer look at the chem and….
0
88
0
@_arohan_
rohan anil
5 days
Really enjoyed reading this work! . One way I tried to explain subliminal learning is drawing parallel to watermarking text which generally works by biasing generation at each step to a partition the partitioned token vocabulary (partitioning happens at every step using a private.
@OwainEvans_UK
Owain Evans
5 days
Paper authors: @cloud_kx @minhxle1 @jameschua_sg @BetleyJan @anna_sztyber @saprmarks & me. Arxiv pdf: Blogpost: Supported by Anthropic Fellows program and Truthful AI.
Tweet media one
0
1
14
@_arohan_
rohan anil
5 days
RT @OwainEvans_UK: Subliminal learning may be a general property of neural net learning. We prove a theorem showing it occurs in general fo….
0
30
0
@_arohan_
rohan anil
5 days
RT @OwainEvans_UK: New paper & surprising result. LLMs transmit traits to other models via hidden signals in data. Datasets consisting only….
0
1K
0
@_arohan_
rohan anil
5 days
RT @AnthropicAI: In a joint paper with @OwainEvans_UK as part of the Anthropic Fellows Program, we study a surprising phenomenon: sublimina….
0
172
0
@_arohan_
rohan anil
5 days
RT @cloneofsimo: Everyone get your top 1% quality dataset and train 100 epochs right now
Tweet media one
0
27
0
@_arohan_
rohan anil
6 days
Also what does this mean:. Problems are easy?.Intelligence overhang?.All the above?.
1
0
9