For a long time, AI researchers argued that logic and math are easy for computers, while common sense is hard. But it turns out that common sense comes down to a simple statistical learning algorithm and lots of data, while deeply understanding logic and math is still hard
MACHINE LEARNING
-
Smaller Models Emerge as Big AI Model Pipeline Slows
By
–
I don’t think any major new big models will come out – it’s takes years to train them and nothing new seems to be in the pipeline. We already have quite a few. But there could be advances in smaller models (SLMs) – that more suited to specific domains and tasks. For legal,
-
Vibe Coding: Unusual Approach to AI-Assisted Programming
By
–
This is from a unusual sample (people who posted about vibe coding online): https://
arxiv.org/pdf/2510.00328
v1
… -
How LLMs Work: Weights, Architecture, and Running Models
By
–
– you are
– a normal dev with zero clue how LLMs actually work
– want to know how it all runs together
– suddenly: you see the layers, the tradeoffs, the forbidden math – first: what even is “running a model”? – model = weights (giant files, 2–140GB) + model architecture -
CS professionals can master LLMs AI in two years
By
–
any cs person can go from zero to deeply knowledgeable in llms and ai in ~2 years, top to bottom the elite don't want you to know this
-
Building Systems for AI-Driven Scientific Discovery Integration
By
–
Science isn't just a thing that happens. We can have novel discoveries flowing from AI-human collaboration every day (and soon, AI-led science), and we really have not built the system to absorb those results and translate them into streams of inquiry and translations to practice
-
LLM Capabilities Fragmented: Vision, Tools, and Heavy Thinking Models
By
–
The state of LLMs is messy: Some AI features (like vision) lag others (like tool use) while others have blind spots (imagegen and clocks). And the expensive "heavy thinking" models are now very far ahead of all the other AIs that most people use. None of this is well-documented.
-
Multiple Choice Questions as Model Training for Real-World Performance
By
–
What would be useful is when including more MC questions would also improve the answer performance on similar-topic questions where answer choices are not shown (like in a real world context). That being said I can see models who do well on MC settings being useful as verifiers
-
Internal Sanity Checks vs External Evaluations for LLM Benchmarking
By
–
Yes, it’s cheaper and easier, but it’s more of an internal sanity check than outward facing eval to report imho. Btw, spot on regarding including it for the sake of benchmarks. You can tell based on how sensitive some LLMs are to the exact MC prompt format.
-
Open-ended Evaluations: Free-form Answers and Conversational Performance
By
–
With open-ended evals you mean free-form answers (and measuring conversational performance)?