GPT-4.1 peut trouver plusieurs réponses cachées dans un long texte, même si elles se ressemblent presque à l’identique.
RESEARCH
-

GPT-4.1 Outperforms GPT-4o in Coding Benchmarks
By
–
GPT-4.1 écrase tout en programmation. 55 % de précision contre 33 % pour GPT-4o. Moins de bugs, code plus intelligent.
-

OpenAI Releases GPT-4.1, Mini, and Nano Models
By
–
OpenAI a sorti 3 modèles : GPT-4.1, Mini et Nano — plus rapides, plus intelligents, moins chers, et capables de gérer 1 million de tokens. Ils surpassent GPT-4o et GPT-4o mini sur tous les plans.
-
Lilian Weng announces Thinking Machines happy hour at ICLR2025
By
–
See you at #ICLR2025 soon. Excited about chatting with many of you about Thinking Machines and what we have been up to! Thinking Machines (@thinkymachines) Thinking Machines is hosting a happy hour in Singapore during #ICLR2025 on Friday, April 25: lu.ma/ecgmuhmx Come eat, drink, and learn more about us! — https://nitter.net/thinkymachines/status/1911844660190069117#m
→ View original post on X — @lilianweng, 2025-04-14 18:29 UTC
-

GPT-4.1 models available on OpenAI
By
–

GPT-4.1 models are now available on OpenAI Platform and APIs
-

GRPO Simplified: Implementation Guide for Language Models
By
–
GRPO,simplified. https://
k-a.in/grpo.html
and
Implementing GRPO https://
k-a.in/grpo-1B.html
by @arjunkocher -
Cool Research Links Earth Intelligence to Hitchhiker’s Guide Concept
By
–
This is really cool work! But also it's confirming that Hitchhiker's Guide to the Galaxy may've accurately described the nature of intelligence here on Earth
-
Google Launches DolphinGemma LLM to Decode Dolphin Language
By
–
The CEO of Google announces to us a new AI capable of understanding dolphin language. Yes, you're not dreaming. Translation: Introducing DolphinGemma, a language model (LLM) fine-tuned from several years of dolphin sound data to advance scientific research. We collaborated
-

OpenAI Launches o3 and o4-mini Reasoning Models for Autonomous Science
By
–
C'EST LE CHOC ! CETTE SEMAINE VA TOUT CHANGER OpenAI s’apprête à lancer de nouveaux modèles de raisonnement (o3 et o4-mini) capables, pour la première fois, de développer de nouvelles idées scientifiques de manière autonome.
Ces IA peuvent croiser des connaissances issues de -
Enterprise GenAI Evaluation: Beyond Traditional Metrics
By
–
At the #GenAISummit a few days back, our CEO @ajratner discussed why traditional AI metrics fall short & how enterprises can use business-specific evaluation criteria, scale SME feedback & apply data slicing & programmatic evaluation to get the most from #GenAI #SnorkelAI
