wx-b.bsky.social
LightNews LightNews
wx-b.bsky.social
@wx-b.bsky.social
7 followers 220 following 6 posts
Posts Replies Media Videos
New research on grokking suggests that Softmax Collapse (SC) from numerical instability & Naive Loss Minimization (NLM) are key hurdles to generalization. Proposed StableMax & ⊥Grad show promise in addressing these.

openreview.net/forum?id=Tvf...
Grokking at the Edge of Numerical Stability
Grokking, or sudden generalization that occurs after prolonged overfitting, is a surprising phenomenon that has challenged our understanding of deep learning. While a lot of progress has been made...
openreview.net
January 14, 2025 at 8:25 PM
Proud to share our new #CVPR2024 paper! Discover how RIOS and ANU improve NeRF by modeling transparent objects with our innovative neural surface refinement technique. #NeRF #AI #3DVision

Project page
tnsr.rios.ai
July 4, 2024 at 9:49 PM
An open source alternative to Devin has 1.3K stars in 2 days. Whether or not the current generation of agent-based software engineering tools turn out to have practical applications or are just part of inflated expectations around AI probably depends on (1/2)

github.com/stitionai/de...
GitHub - stitionai/devika: Devika is an Agentic AI Software Engineer that can understand high-level human instructions, break them down into steps, research relevant information, and write code to ach...
Devika is an Agentic AI Software Engineer that can understand high-level human instructions, break them down into steps, research relevant information, and write code to achieve the given objective...
github.com
March 22, 2024 at 5:54 PM
"LLMs are not genies that grant wishes. They are idiot savants that you can make on your computer."

bit.ly/LocalLLaMA-c...
March 22, 2024 at 5:52 PM