himanshu
@retr0sushi_
Followers
3K
Following
42K
Media
756
Statuses
6K
anhedonic diet coke addict tracing echoes through world model latent spaces @lossfunk
Joined July 2021
I'm expanding my EA team to help kickstart new projects at @lossfunk Interested? Please apply below 👇 RT for reach 🙏
4
24
242
i love my iphone 13 but it’s showing signs of aging, this boy been so good to me, i will legit hold a funeral for it and bawl my fkn eyes out
1
0
7
i had rendered this impossible yet someone did it
Today I learned that you can just walk into Pizza 4P’s during the day and still get a table!
0
0
5
bro never stopped gaming, i respect that
0
0
2
s3 is nothing but a hate crime, the ultimate fuck you to anyone who loves one punch man
0
0
3
real
0
0
2
New blog post Why all ARC-AGI solvers fail today (and how to fix them) tldr they all violate minimum description length why test time training is necessary anti-bitter lesson hacks are unneccesary falsifiable predictions Link in thread
5
15
43
i never moved on from these and seems like the world is catching up haha
0
0
3
sounds exactly like how me and @akshat_sj had our heads blown the first time we ported some code and did some jax magic and it was fast as lightning
wow jax is actually fast
0
0
6
@James_Martell_ 1- Sartre 2- Camus 3- All the great unnamed Hindu philosophers 4- William James 5- Wittgenstein 6- Quine 7- ChatGPT
7
4
156
Super delighted to share that our paper on LLMs and their forecasting ability has been accpeted at the AIR-FM workshop @RealAAAI 2026 in singapore! Research done @lossfunk by me and @paraschopra . Preprint soon:)
18
12
139
Ok, another banger result! Wee show how sequentially iterating on problems via LLM significantly boosts accuracy (over the conventional wisdom of majority voting by parallel chains). This works well not just for reasoning but for creative tasks (like generating jokes) as well.
📢 Releasing our latest paper Selected for @NeurIPSConf workshop on efficient reasoning! We show that the optimal test time scaling method is iteratively refining through sequential steps. 👉 Our method beats majority voting by parallel chains in 95% of configurations with
8
13
147
Here is our latest paper , go check it out! Worked on it the last month and some really cool results for the community to build upon during test time scaling.
📢 Releasing our latest paper Selected for @NeurIPSConf workshop on efficient reasoning! We show that the optimal test time scaling method is iteratively refining through sequential steps. 👉 Our method beats majority voting by parallel chains in 95% of configurations with
0
1
8
📢 Releasing our latest paper Selected for @NeurIPSConf workshop on efficient reasoning! We show that the optimal test time scaling method is iteratively refining through sequential steps. 👉 Our method beats majority voting by parallel chains in 95% of configurations with
3
12
68
this is so sick! 🔥🔥
introducing runable the general ai agent for every task slides, websites, reports, podcasts, images, videos... everything
0
0
2
bro takes the most breathtaking photos
0
0
1
only at a @lossfunk party folks!
I was at a party in blr and this guy pulled out a chess board while everyone else was going mad at the dance floor. what do you even call this?
0
0
9