Congratulations to BAIR Faculty @negar_mehr and BAIR alumni @pathak2206 @ebiyik_ @mark_ho_ who have been awarded the 2026 ONR Young Investigator Award!
RESEARCH
-
Understanding SiLU/GELU Activation Performance Loss in RL Networks
By
–
I always lost performance when I tried to use silu/gelu activations in my RL value networks, and I finally understand why. If the pre-activation values are small, the smooth curve through zero is basically a linear activation, destroying the representation power of the network. You need a batch/layer/rms norm on the preactivations to put them in the range the smooth activations are designed for. Internal norms generally hurt performance on our RL tasks, but combining them with a smooth activation at least works basically as well as a raw relu (but slower). So, not actually a win, but the lightbulb of understanding was good!
→ View original post on X — @id_aa_carmack, 2026-02-23 16:54 UTC
-
Debating the relationship between Reinforcement Learning and reasoning
By
–
People are trying to use RL for reasoning when what they should do is use reasoning for RL.
-
Applied Causal Inference Powered by ML and AI (arXiv eBook)
By
–
[Download 496-page PDF eBook] Applied Causal Inference Powered by #MachineLearning and #AI: http://
arxiv.org/abs/2403.02467
————
#ML #DataScience #Algorithms #Statistics -
Book: Data Without Labels — Practical Unsupervised Machine Learning
By
–
Data Without Labels — Models and Algorithms for Practical Unsupervised #MachineLearning: http://
amzn.to/4q5bbYz 𝓦𝓱𝓪𝓽 𝓨𝓸𝓾 𝓦𝓲𝓵𝓵 𝓛𝓮𝓪𝓻𝓷: Fundamental building blocks and concepts of machine learning and unsupervised learning
Data cleaning for structured and -
AI progress in text and images, from blobs to SAT scores
By
–
I don't agree, I've seen all of this play out first for text, where in 2020 GPT-1 could basically not even write a simple coherent sentence. Now well, LLM's are scoring better on SATs than humans. Secondly it happened to photos, same there, started with 124×124 pixel blobs
-
Review: Mathematics of Machine Learning — book on mathematical algorithms
By
–
Mathematics of Machine Learning: http://
amzn.to/4eN7i52 by @TivadarDanka v/ @PacktDataML —
Author’s GitHub: https://
github.com/cosmic-cortex?
tab=achievements
…
—
My review: 𝗧𝗵𝗲 𝗦𝗲𝘁 𝗢𝗳 𝗠𝗮𝘁𝗵𝗲𝗺𝗮𝘁𝗶𝗰𝗮𝗹 𝗔𝗹𝗴𝗼𝗿𝗶𝘁𝗵𝗺𝘀 𝗧𝗵𝗮𝘁 𝗟𝗲𝗮𝗿𝗻 𝗙𝗿𝗼𝗺 𝗘𝘅𝗽𝗲𝗿𝗶𝗲𝗻𝗰𝗲 This massive book -
Fusing Street View and Aerial Imagery into SF Radiance Fields
By
–
So much fun I spent a good visiting my old haunts. Now imagine fusing all ground street view and aerial oblique, and select drone imagery into one massive radiance field of SF
-
Top AI Research Papers: GLM-5, Agents, and Context Management
By
–
The Top AI Papers of the Week (February 16-22) – GLM-5
– SkillsBench
– MemoryArena
– Team of Thoughts
– AI Delegation Framework
– Lossless Context Management Read on for more: -
Augmented Data Drives AI Pattern Recognition in Modern Manufacturing
By
–
Modern manufacturing involves a complex web of overlapping processes. Augmented data provides the necessary clues to navigate this landscape. It allows AI models to distinguish between noise and meaningful patterns, making it a critical differentiator for any facility pursuing… pic.twitter.com/YySnfdjd8D
— Lucian Fogoros (@fogoros) 22 février 2026Modern manufacturing involves a complex web of overlapping processes. Augmented data provides the necessary clues to navigate this landscape. It allows AI models to distinguish between noise and meaningful patterns, making it a critical differentiator for any facility pursuing