This paper by Meta introduces Query-Only Test-Time Training for long contexts So by using a small amount of inference-time training that retunes how the model attends to the given input, it works far better than generating extra reasoning tokens
LLMS
-
RAG Systems Struggle with Multi-Hop Reasoning Problems
By
–
RAG systems struggle with multi-hop reasoning. In most cases, the problem isn't the LLMs. It's the retrieval system. Standard RAG treats each piece of evidence as equally reliable, ignoring how documents connect to each other. Why is this a problem? When questions require
-
DeepSeek rivals GPT-5 on LM Arena, released in August
By
–
For coding it is, but not for typical LLM outputs. DeepSeek is tied with GPT-5 on LM Arena and it was released in August.
-
Open-source AI on Mac Studios: Cheaper, faster, and free alternative to ChatGPT
By
–
Small companies could run DeepSeek-V3.2 on one of these souped-up Mac Studios instead of paying for dozens and dozens of ChatGPT licenses for employees. Open-source models are only about 6 months behind the proprietary models. And they’re free. Forever.
-
AI Solves Impossible Math Problems Better Than Top Mathematicians
By
–
AI is solving 'impossible' math problems. Can it beat the world's top mathematicians? | Live Science https://
share.google/jACbXTwJJIuKc3
ADE
… #AI #ArtificialInteligence #math #LLM #GenAI #GenerativeAI #mathematics -
Meta AI developing new Memories and Custom Prompts features
By
–
Meta likely aims to bridge a feature gap with top-tier AI labs in one shot, and is working on Memories and Custom Prompts for Meta AI. Soon?
-
Using ChatGPT to generate personalized Sora holiday videos
By
–
You can get a personalised Christmas Sora video on ChatGPT if you send emoji to the chat. Looking forward to ChatGPT wrapped
-
8 RAG Architectures Every AI Engineer Must Know
By
–
8 RAG architectures all AI Engineers should know:
-
The need for efficiency research in AI agent development
By
–
AI agents are getting slower and it might be the time they all stop scaling and start doing some research to make them fast and accessible to everyone.
-
AI 2025 Breakthroughs: RL, Reasoning Models, and Future Paradigms
By
–
Things move very very fast in AI. 2025 was a year of RL with verifiable rewards(RLVR), LLM ghosts/jagged intelligence, Cursor-like LLM apps, claude code/codex, vibe-coding, NanoBanana showing early glimpse of LLM promptable graphical interfaces(PGI, i just coined this lol), LLM reasoning models crushing olympiad competitions (maths, physics, code). Most altering releases tend to come early in a year, jan-feb, and then scaling-up and small fixes begin. Eagerly looking forward to new paradigm shifts. What will next NanoBanana look like, just bigger or new capabilities no one thought before? There are several stages of training now, RL(VR) being the most recent. What will be the RL successor? And continual learning, will it be fixed in 2026, or this is a problem we will live with for long? There are also world models, agents that actually work reliably in the wild for hours. Andrej Karpathy (@karpathy) x.com/i/article/200211463822… — https://nitter.net/karpathy/status/2002118205729562949#m