Project Greenhouse: Progress Toward Fully Open and Sovereign Agentic Search

Key points

• As a first milestone, we describe how to build a competitive pointwise decoder-only reranker using a simple two-step recipe comprising pre-training from scratch followed by supervised fine-tuning, starting only from commonly available datasets.1
• Contrary to the dominant approach in the literature, we do not rely on existing open-weight backbones from third parties, and thus we are fully in control of model training, from end to end.1
• This report articulates the importance and benefits of our approach, and we share artifacts that enable transparent, independent reproduction of all aspects of model training.1
• Beyond data, code, and configurations that capture our efforts, we also release checkpoints for our family of Gaggle models, demonstrating the feasibility of our approach and providing a first step toward validating our broader thesis.1

Sources (1)

• [1]Project Greenhouse: Progress Toward Fully Open and Sovereign Agentic Search
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 01:19 PM

  Project Greenhouse represents our exploration of a simple thesis: We believe that it is possible to build fully open and sovereign models for agentic search with only modest computational resources.

  As a first milestone, we describe how to build a competitive pointwise decoder-only reranker using a simple two-step recipe comprising pre-training from scratch followed by supervised fine-tuning, starting only from commonly available datasets.

Extractive summary: sentences quoted from the sources.