Generating Human-level Text with Contrastive Search in Transformers đ¤
Weâre on a journey to advance and democratize artificial intelligence through open source and open science.
More from the blog
GPT-Lab
¡ gpt-lab.eu
Oct 8, 2026
GenEye in a Box: Making Machine Vision Something You Can Just Ask For
Exploring how generative AI could make machine vision more accessible to businesses. The post GenEye in a Box: Making Machine Vision Something You Can Just Ask For appeared first on GPT-Lab.
Discovering the value of humanistic inquiry
Students in MITâs Concourse program delve deeply into the human condition, debate challenging questions, and learn to develop judgment about issues that canât be quantified.
One Model Family, Two Gold-Level Results: Fine-Tuning Nemotron for IOI and IMO
Open-sourcing AstaBrief, the fast report-generation model in Asta
A Blog post by Ai2 on Hugging Face
Related papers
A Multiple Case Study of Artificial Intelligent System Development in Industry
This investigation revealed different types of AI systems and different AI development approaches, but it is common that business opportunities involving with AI systems are not validated and there is lack of business-driven metrics that guide the development ofAI systems.
Experimenting with Multi-Agent Software Development: Towards a Unified Platform
A unified platform that utilizes multiple artificial intelligence agents to automate the process of transforming user requirements into well-organized deliverables, including user stories, prioritization, and UML sequence diagrams, along with the modular approach to APIs, unit tests, and end-to-end tests.
Bridging Humans and LLMs: Investigating Human-AI Collaboration in Multi-agent Requirements Analysis for Organizational AI Adoption
LLM-based multi-agent systems can support strategic AI planning by enabling iterative refinement with human experts by supporting structured and collaborative Requirements Engineering processes for AI adoption planning.
Constitutional Midtraining: Content Presence Drives Alignment Gains
Post-training alignment is often shallow, eroding under fine-tuning. It remains untested as to whether constitutional midtraining interventions can produce durable alignment when cleanly isolated from post-training. We build a 394M-token constitutional corpus from Anthropic's Constitution and apply constitutional midtr...