BREAKING : OpenAI released a new GPT-5-Codex model, optimised for coding and refactoring tasks. Codex is gaining momentum by being available almost everywhere – CLI, IDEs, Mobile and Web apps. Merged
MACHINE LEARNING
-
Machines Working Enable Humans to Play More
By
–
In the world of humachineology and humachinekind, machines are working and, therefore, humans have more time for play.
-
Inpainting-Guided Policy Optimization for Diffusion Language Models
By
–
Inpainting-Guided Policy Optimization for Diffusion Large Language Models
-
High Fidelity Structure Coherent Shape Decomposition
By
–
X-Part high fidelity and structure coherent shape decomposition
-
Search Space Optimization in Deep Learning: Beyond Polyhedral Constraints
By
–
polyhederal search spaces are okay but 1. for DL they dont even need to be poly, just rectangular is usually sufficient 2. the search space is highly discontinuous and non-linear, so the search is timetaking and highly unpredictable as a function of shape size
-
Royal Flush vs Tweaking AI Models: Which is More Likely?
By
–
What’s more likely: a royal flush or never having to "just tweak the model a bit"
-
Skepticism about the article on Clermont and machine learning in 2025
By
–
Wait, is this a joke? There have been programs in universities and engineering schools teaching machine learning for at least 10 years. And now there's an article claiming that the University of Clermont wants to be among the first in 2025? The level of
-
Generalist VLA Models and Specialist Models for Data Collection
By
–
I agree. Generalist VLA models can be much more versatile—given enough data. I believe specialist models can be useful for collecting this data.
-
Reinforcement Learning for Reasoning: A Category Error
By
–
Using reinforcement learning for reasoning is a category error.
-
Training Models: Balancing Demonstrations with Real-World Performance
By
–
See my original comment Sure, it's easy to deliberately demonstrate on small datasets, but for it to affect real-world models the people training them would have to ignore the problem entirely What makes you think they're just going to let their models get worse?