#reasoningmodels
October 11, 2026 at 10:59 PM
La semana en [Blue Chip]. Lunes 28-Sep: "Cómo hacemos que un modelo razone (IV): aprendizaje por refuerzo basado en recompensas verificables (RLVR)" ignaciogavilan.com/como-hacemos... #RLVR #GRPO #ReinforcementLearning #AprendizajePorRefuerzo #ModelosRazonadores #ReasoningModels #IA #AI #LLM
Cómo hacemos que un modelo razone (IV): aprendizaje por refuerzo basado en recompensas verificables (RLVR) | Ignacio G.R. Gavilán
Comentamos la estrategia de aprendizaje por refuerzo con recompensas verificables (RLVR) como forma de entrenamiento para modelos razonadores.
ignaciogavilan.com
October 4, 2026 at 7:19 PM
I've just finished "Build a Reasoning Model (from Scratch)" (www.amazon.es/Build-Reason...) by @rasbt.bsky.social Deep, development-oriented but with solid theoretical foundations #libros #books #AI #ReasoningModels #LLM #FrontierModels
September 21, 2026 at 6:45 AM
La semana en [Blue Chip]. Miércoles 16: "Cómo hacemos que un modelo razone (I): tres grandes estrategias" ignaciogavilan.com/como-hacemos... #AI #ArtificialIntelligence #LLM #ReasoningModels #ModelosRazonadores #AgenticAI
Cómo hacemos que un modelo razone (I): tres grandes estrategias | Ignacio G.R. Gavilán
Comentamos las tres estrategias básicas para hacer que hacemos que un modelo razone, pasando de la generación de texto a planes.
ignaciogavilan.com
September 20, 2026 at 8:09 PM
Poetiq’s AI Reasoning Layer Hits 54% on ARC-AGI-2 at Half the Cost: Meta-system breaks the 50% barrier on the benchmark while cutting per-problem costs by more than half.
Continue reading... #ainews #reasoningmodels
Poetiq’s AI Reasoning Layer Hits 54% on ARC-AGI-2 at Half the Cost
Meta-system breaks the 50% barrier on the benchmark while cutting per-problem costs by more than half. Continue reading...
www.vktr.com
December 29, 2025 at 4:00 PM
🚀 UC Berkeley’s Sky Computing Lab debuts Sky-T1-32B, a truly open-source reasoning AI trained for just $450! 🌟 Excelling in math & coding, it redefines affordable AI. What’s next for open models? #AI #ReasoningModels

Read more from @techcrunch.com: techcrunch.com/2025/01/11/r...
Researchers open source Sky-T1, a 'reasoning' AI model that can be trained for less than $450 | TechCrunch
Researchers at UC Berkeley have open sourced Sky-T1, a 'reasoning' AI model that can be trained for less than $450.
techcrunch.com
January 14, 2025 at 2:20 AM