Sample, Don't Search: Rethinking Test-Time Alignment for Language Models Gonçalo Faria, Noah A. Smith: https://
arxiv.org/abs/2504.03790 #ArtificialIntelligence #DeepLearning #MachineLearning
MACHINE LEARNING
-
Sample-Based Approach Improves Language Model Test-Time Alignment
By
–
-
Sampling Over Search: New Test-Time Alignment for Language Models
By
–
Sample, Don't Search: Rethinking Test-Time Alignment for Language Models Gonçalo Faria, Noah A. Smith: https://
arxiv.org/abs/2504.03790 #ArtificialIntelligence #DeepLearning #MachineLearning -
Share insights on LLMs, AI Agents, and Machine Learning
By
–
If you found it insightful, reshare with your network.
— Akshay 🚀 (@akshay_pachaar) 12 avril 2025
Find me → @akshay_pachaar ✔️
For more insights and tutorials on LLMs, AI Agents, and Machine Learning! https://t.co/vn40hmVFRKIf you found it insightful, reshare with your network. Find me → @akshay_pachaar For more insights and tutorials on LLMs, AI Agents, and Machine Learning!
-
Vanna: Open-Source RAG Framework for Text-to-SQL Generation
By
–
Finally! A Text-to-SQL tool that actually works!
— Akshay 🚀 (@akshay_pachaar) 12 avril 2025
Vanna is an open-source RAG framework for complex Text-to-SQL generation. It manages dynamic data and allows custom RAG model training for greater accuracy.
100% open-source. pic.twitter.com/5mKAJUuAHBFinally! A Text-to-SQL tool that actually works! Vanna is an open-source RAG framework for complex Text-to-SQL generation. It manages dynamic data and allows custom RAG model training for greater accuracy. 100% open-source.
-
Sam Altman Envisions 10 Million GPUs for AI Pre-Training
By
–
« L’humanité fera-t-elle un jour un entraînement de pré-apprentissage avec 10 millions de GPU ? »
— VISION IA (@vision_ia) 12 avril 2025
Le PDG d’OpenAI, Sam Altman, soulève la question :
« Il y aura 10 millions de GPU travaillant ensemble sur un système d’IA capable d’apprendre et d’exécuter des tâches. »… pic.twitter.com/3gZrqijwQQ« Will humanity one day conduct a pre-training run with 10 million GPUs? » OpenAI CEO Sam Altman raises the question: « There will be 10 million GPUs working together on an AI system capable of learning and performing tasks. » However, the method could evolve: we would shift
-
Scaling Reasoning in Diffusion Large Language Models via RL
By
–
d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learning
Paper: https://
dllm-reasoning.github.io/media/preprint
.pdf
…
Project Page: https://
dllm-reasoning.github.io
Code: https://
github.com/dllm-reasoning
/d1
… -
GRPO Adapted for Diffusion LLMs: UCLA and Meta’s diffu-GRPO
By
–
Can GRPO work for diffusion LLMs (dLLMs)?
A team from UCLA and Meta says yes—and they’ve got the receipts.
They introduce d1, a framework that adapts pre-trained masked dLLMs into reasoning models via SFT + RL.
The twist? Their RL method is a fresh one: diffu-GRPO. -
DeepSeek-R1 Reasoning: New Thoughtology for Large Language Models
By
–
DeepSeek-R1 Thoughtology: Let's about LLM Reasoning Marjanović et al.: https://
arxiv.org/abs/2504.07128 #ArtificialIntelligence #Deeplearning #MachineLearning -
DeepSeek-R1 Thoughtology: Understanding LLM Reasoning Mechanisms
By
–
DeepSeek-R1 Thoughtology: Let's about LLM Reasoning Marjanović et al.: https://
arxiv.org/abs/2504.07128 #ArtificialIntelligence #Deeplearning #MachineLearning -
Statistical Code Generation vs Direct Program Space Search
By
–
Generating *code* using a statistical prior about which token sequences are more likely is a fundamentally different task than generating *programs* by searching directly in the space of program dynamics