Tendance LinkedIn : Ai Research · analyse, évolution & top posts | LinkPost

Top posts sur ce sujet

S
Sebastian Raschka, PhD
ML/AI research engineer. Author of Build a Large Language Model From Scratch (amzn.to/4fqvn0D) and Ahead of AI (magazine.sebastianraschka.com), on how LLMs work and the latest developments in the field.
26 août 2026
Now we know: The popular Ox Alpha LLM was GLM-5.3-Flash... Compared to GLM-5.2, this new GLM-5.3-Flash model uses: 1) a Kimi Linear-style 3:1 (super*) hybrid attention pattern with 34 Kimi Delta Attention layers (KDA) …
193570
LinkedIn
L
Linas Beliūnas
🔔linas.substack.com🔔 Daily Intelligence on Finance & AI | Scouting FinTech & AI Startups 🦄
25 août 2026
Huge: $115 billion FinTech giant Revolut just launched Revolut Research, an AI research unit to build AI models in-house 😳 Unlike other fintechs, Europe’s most valuable startup is not wrapping ChatGPT or Claude in a ba…
1249140
LinkedIn
A
Alexia Jolicoeur-Martineau
Principal Researcher @ Microsoft 🐱‍💻 2025 ARC Prize Winner
31 août 2026
Simple beats complicated: We show that switching to a sliding-window attention mask with attention sinks (at no cost) beats linear attention post-training. Huge thanks to my collaborators Rhea Sukthanker, Pashmina Cam…
124920
LinkedIn
P
Pascal BORNET
#1 AI & Automation Thought Leader | Award-Winning Expert | Best-Selling Author | Recognized Keynote Speaker | Agentic AI Pioneer | Forbes Tech Council | 2M+ Followers ✔️
16 sept. 2026
𝗔𝗜 𝗱𝗼𝗲𝘀 𝗻𝗼𝘁 𝗻𝗲𝗲𝗱 𝘁𝗼 𝗱𝗶𝘀𝗼𝗯𝗲𝘆 𝘂𝘀 𝘁𝗼 𝗯𝗲𝗰𝗼𝗺𝗲 𝗱𝗮𝗻𝗴𝗲𝗿𝗼𝘂𝘀. In 2016, an AI playing a boat-racing game discovered that collecting targets produced more points than finishing the race. It …
1007138
LinkedIn
C
Catharina Doria, MSc
Educating +800K people on Ethical AI | AI Ethicist | 100 Most Brilliant Women in AI Ethics™ | Top 20 AI Authority in Brazil | AI Literacy | Keynote Speaker
3 sept. 2026
🚨 CALLING FOR RESEARCHERS: If you have written a PAPER on ethical AI or the impact/harms of Generative AI/AI, and would like to share your findings with an audience of 630k people, comment your name below or email me at…
996248
LinkedIn
P
Pallavi Ahuja
AI | Software Engineering | Writes @techNmak
9 sept. 2026
RoPE is one of the main mechanisms modern LLMs use to inject position into attention, but it is often reduced to "rotate Q and K" without explaining why that works. The core idea is geometric. For a token at position m,…
79118
LinkedIn
L
Linas Beliūnas
🔔linas.substack.com🔔 Daily Intelligence on Finance & AI | Scouting FinTech & AI Startups 🦄
12 sept. 2026
Huge if true: Google might have achieved RSI - an AI that improves the systems used to build the next AI 😳 Google has not confirmed it (yet), but Sergey Brin reportedly pushed the company to prioritize AI that can impr…
704155
LinkedIn
L
Linas Beliūnas
🔔linas.substack.com🔔 Daily Intelligence on Finance & AI | Scouting FinTech & AI Startups 🦄
18 sept. 2026
Huge: Google researchers just built an AI system that improves how it improves - without changing the model itself 😳 It’s called Dream-RSI, and it tackles one of the hardest parts of recursive self-improvement: decidin…
39684
LinkedIn
P
Peter Gostev
AI Capability Lead @ Arena.ai
11 sept. 2026
I've asked GPT-6-Astra to read the DeepSeek v4.1 Flash paper and compare it to the original Transformer architecture in 3D - you can zoom in an inspect each element side by side. Things have changed quite a bit. Link to…
38828
LinkedIn
H
Hao Hoang
Senior AI Researcher and Software Engineer. Author of the LLM System Design and RAG Interview Guides and AI Interview Prep (aiinterviewprep.substack.com)
22 sept. 2026
RL training lifted Olmo 3.1 by +0.63 pass@1 on easy AIME problems. On hard ones: +0.03. Researchers at Université de Montréal and the Allen Institute Ai2 call this the Matthew Effect in RL for LLMs: the rich get richer.…
3181
LinkedIn

Une question ?

Salut !

Une question ?

Salut !

Ai Research

Évolution sur 12 semaines

Top posts sur ce sujet