Name: LinkedIn · Ai-research
Creator: LinkPost

Top posts sur ce sujet

M
Maxime Labonne
Head of Post-Training @ Liquid AI
11 juin 2026
🐳 Revisiting DeepSeek-V4: how RL got replaced with on-policy distillation DeepSeek released V4 in late April with a huge tech report! I finally found time to finish it and finalize my article on Substack. There are thr…
38619
LinkedIn
M
Maxime Labonne
Head of Post-Training @ Liquid AI
1 juil. 2026
⚗️ Multi-teacher on-policy distillation (MOPD) Xiaomi's MiMo team presents a clean recipe for combining several specialist models into one, without the usual cross-domain interference. Instead of pooling all the trainin…
32118
LinkedIn
C
Clem Delangue 🤗
Co-founder & CEO at Hugging Face
14 mai 2026
Are scaling laws finally working for time series foundation models? Today, Datadog is releasing Toto 2.0 weights in Apache 2.0 on Hugging Face. It's a family of open-weights TSFMs from 4M to 2.5B parameters, where every…
25120
LinkedIn
S
Sandipan Bhaumik
Data & AI Technical Lead | Production AI for Regulated Industries | Founder, AgentBuild
22 mai 2026
Claude Code Isn't a Chatbot. 𝐈𝐭'𝐬 𝟕 𝐀𝐠𝐞𝐧𝐭𝐬 𝐑𝐮𝐧𝐧𝐢𝐧𝐠 𝐚𝐬 𝐚 𝐏𝐢𝐩𝐞𝐥𝐢𝐧𝐞. Most engineers using Claude Code don't realize this. They think they're prompting one model. What's actually happening: A …
14493
LinkedIn
K
Kevin Degila
Head of Data and AI | Forbes 30 under 30
16 juil. 2026
Vous avez été nombreux à precommander le livre après son annonce hier. Merci pour cet accueil enthousiaste. « Construire un LLM de zéro - des maths aux agents IA », c'est 4 parties, 21 chapitres, et une promesse : parti…
1392
LinkedIn
G
Ghita Houir Alami
Co-Founder and CEO at ZeroEntropy (YC W25) - We're Hiring!
21 mai 2026
Mini models are supposed to get cheaper every release. They're getting more expensive. Gemini 1.5 Flash to 2.5 Flash: 4x on input, 8x+ on output. Gemini 3 Flash to 3.5 Flash: another 3x. And 3.5 Flash, the model that's…
1096
LinkedIn
G
Ghita Houir Alami
Co-Founder and CEO at ZeroEntropy (YC W25) - We're Hiring!
20 mai 2026
MTEB largely utilizes binary relevance scores: a document is either relevant or it isn't. But under binary labels, metrics like NDCG degenerate. It can't tell a perfect answer apart from a loosely related document. So…
10011
LinkedIn
B
Ben Burtenshaw
Community Education in AI @ Hugging Face
4 juin 2026
Had loads of fun setting up continuous learning through on agent traces. This post is based on a script that implements continuous learning with self distillation on agent traces. It's purely an experiment and not somet…
971
LinkedIn
P
Pini Reznik
CEO and Co-Founder @ re:cinq | AI Native Transformations
12 mai 2026
Last week we ran the 9th edition of AI Native Netherlands at Adyen, with more than 100 people in the room. Many thanks to Kwok He Chu and Ayodeji Ogundare at Adyen to help organise this event. 𝗪𝗲 𝗵𝗮𝗱 𝘁𝘄𝗼 𝗴𝗿𝗲…
971
LinkedIn
M
Max Buckley
Founding something new
13 mai 2026
Exa at AI Tinkerers Dublin on Thursday May 28th 🇮🇪 I'll be representing Exa in my home country of Ireland very soon. I'll be presenting at AI Tinkerers Dublin, hosted by Baseline, giving a deep dive on building AI res…
816
LinkedIn

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

Une question ?

Salut !

Une question ?

Salut !

Ai-research

Évolution sur 12 semaines

Top posts sur ce sujet