MindPage Search v1 (GGUF)
On-device search model of MindPage, a notes and PDF app. It finds the right page in
your notes, PDFs, handwriting and recorded lectures/meetings. Fine-tuned from Google's
EmbeddingGemma 2 (Apache-2.0), text model only, Q8_0 for llama.cpp
(b11716 or newer, architecture gemma-embedding2).
Use
Same prompts as EmbeddingGemma 2, vectors truncated to 256 dimensions (Matryoshka) and L2-normalized:
- query:
task: search result | query: {question} - passage:
title: none | text: {document title}, page {n}: {passage}(recordings:{title}, recording: {passage})
Training
- Contrastive (multiple negatives ranking) with hard negatives from the same document, Matryoshka 768/512/256/128.
- Data: questions and searches written by an on-device model for public Wikipedia study articles (CC BY-SA), the same pages rewritten as terse handwritten-style notes with OCR noise, QMSum meeting transcripts and their human queries (MIT), and general retrieval pairs from Natural Questions, GooAQ and SQuAD.
- No user notes were used. Token embeddings frozen; 1 epoch.
Results
Held-out library-wide search (49 articles and 35 meetings never seen in training, 256 dims, llama.cpp), R@1:
| Task | EmbeddingGemma 300M | EmbeddingGemma 2 | MindPage Search v1 |
|---|---|---|---|
| Wiki questions | 80.8 | 76.1 | 82.1 |
| Short searches | 65.8 | 59.4 | 70.1 |
| Handwritten notes (OCR noise) | 67.5 | 70.1 | 79.1 |
| Meetings, human queries | 31.0 | 25.3 | 40.8 |
| Meetings, generated queries | 24.8 | 24.8 | 41.0 |
License
Apache-2.0, as the base model. Built on EmbeddingGemma 2 by Google DeepMind.