NSFW Detector
š¦
Detects potentially NSFW content.
amir
amirgame197
AI & ML interests
I copy others AI.
Recent Activity
new activity
2 months ago
amirgame197/Image-to-Drawing:LION DRAWING
updated
a Space
4 months ago
amirgame197/NSFW-Detector
published
a Space
4 months ago
amirgame197/NSFW-Detector
Organizations
LION DRAWING
1
#1 opened 2 months ago
by
SEMSEM303
updated
a
Space
4 months ago
published
a
Space
4 months ago
liked
a
Space
4 months ago
reacted to
m-ric's
post with š
4 months ago
Post
4937
STOP EVERYTHING NOW - we might finally have a radical architecture improvement over Transformers!!! šØ
A lone scientist just proposed Tiny Recursive Model (TRM), and it is literally the most impressive model that I've seen this year.
ā”ļø Tiny Recursive Model is 7M parameters
ā”ļø On ARC-AGI, it beats flagship models like Gemini-2.5-pro
Consider how wild this is: Gemini-2.5-pro must be over 10,000x bigger
and had 1,000 as many authors š (Alexia is alone on the paper)
What's this sorcery?
In short: it's a very tiny Transformers, but it loops over itself at two different frequencies, updating two latent variables: one for the proposed answer and one for the reasoning.
@AlexiaJM started from the paper Hierarchical Reasoning Model, published a few months ago, that already showed breakthrough improvement on AGI for its small size (27M)
Hierarchical Reasoning Model had introduced one main feature:
š Deep supervision
In their model, one part (here one layer) would run at high frequency, and another would be lower frequency, running only every n steps.
They had used a recurrent architecture, where these layers would repeat many times ; but to make it work they had to do many approximations, including not fully backpropagating the loss through all layers.
Alexia studied what was useful and what wasn't, and cleaned the architecture as follows :
Why use a recurrent architecture, when you can just make it a loop?
ā”ļø She made the network recursive, looping over itself
Why use 2 latent variables ?
ā”ļø She provides a crystal clear explanation : the one that changes frequently is the reasoning, the one that changes at low frequency is the proposed answer.
ā”ļø She runs ablation studies to validate that 2 is indeed optimal.
This new setup is a much more elegant way to process reasoning than generating huge chains of tokens as all flagship models currently do.
This might be the breakthrough we've been awaiting for so long!
A lone scientist just proposed Tiny Recursive Model (TRM), and it is literally the most impressive model that I've seen this year.
ā”ļø Tiny Recursive Model is 7M parameters
ā”ļø On ARC-AGI, it beats flagship models like Gemini-2.5-pro
Consider how wild this is: Gemini-2.5-pro must be over 10,000x bigger
and had 1,000 as many authors š (Alexia is alone on the paper)
What's this sorcery?
In short: it's a very tiny Transformers, but it loops over itself at two different frequencies, updating two latent variables: one for the proposed answer and one for the reasoning.
@AlexiaJM started from the paper Hierarchical Reasoning Model, published a few months ago, that already showed breakthrough improvement on AGI for its small size (27M)
Hierarchical Reasoning Model had introduced one main feature:
š Deep supervision
In their model, one part (here one layer) would run at high frequency, and another would be lower frequency, running only every n steps.
They had used a recurrent architecture, where these layers would repeat many times ; but to make it work they had to do many approximations, including not fully backpropagating the loss through all layers.
Alexia studied what was useful and what wasn't, and cleaned the architecture as follows :
Why use a recurrent architecture, when you can just make it a loop?
ā”ļø She made the network recursive, looping over itself
Why use 2 latent variables ?
ā”ļø She provides a crystal clear explanation : the one that changes frequently is the reasoning, the one that changes at low frequency is the proposed answer.
ā”ļø She runs ablation studies to validate that 2 is indeed optimal.
This new setup is a much more elegant way to process reasoning than generating huge chains of tokens as all flagship models currently do.
This might be the breakthrough we've been awaiting for so long!
- 4 replies
liked
a
Space
5 months ago
Transparent bacground?
12
#13 opened 5 months ago
by
shtefcs
updated
a
Space
5 months ago
Change output video codec
3
#12 opened 12 months ago
by
tigra03
posted
an
update
5 months ago
Post
580
I realized i have github and i was not really posting anything interesting there so its been a few days i started uploading my (kinda) useful projects. some of them are going to be ai related too, so check out my profile if you like!
https://github.com/amirgame197
At the end, its your smile that is appreciated.
https://github.com/amirgame197
At the end, its your smile that is appreciated.
- 1 reply
liked
a
Space
5 months ago
liked
a
Space
7 months ago
Ų§Ų¬Ų±Ų§ Ų±ŁŪ ŁŁŚ©Ų§Ł
6
#1 opened 8 months ago
by
aminahmadi0101
updated
a
Space
9 months ago
published
a
Space
9 months ago
liked
a
Space
9 months ago
updated
4
Spaces
about 1 year ago