Rebrand card to Merino-Pro-4bit

MODEL_CARD.md

@@ -1,42 +1,47 @@
-# baa-ai-Embedding-Reranker-v1
+---
+license: other
+license_name: baa-proprietary
+library_name: sentence-transformers
+tags:
+- retrieval
+- embeddings
+- reranker
+- cross-encoder
+- rag
+- sentence-similarity
+pipeline_tag: sentence-similarity
+---
 
-Unified **embedder + reranker** in one package, sharing a single XLM-RoBERTa-large word-embedding table.
-By BAA AI (Black Sheep AI).
+# baa.ai · Merino-Pro-4bit
+
+**The premium unified retrieval model — bi-encoder embedding *and* cross-encoder reranking in one package, over a single shared word-embedding table.** A 1024-dimensional multilingual model, by BAA AI (Black Sheep AI). ~872M params, 4-bit quantized (~0.5 GB).
+
+## Get the optimal model for *your* data
+
+Merino-Pro-4bit is baa.ai's flagship default. But the best embedder + reranker is **corpus-specific** — the ideal choice depends on your documents and your notion of relevance. **baa.ai offers exclusive tooling that identifies the optimal embedding and reranking models for your specific data**, so you ship the smallest models that maximize document recovery on your corpus. For a tailored recommendation, **reach out to baa.ai**.
 
 ## What it is
-A two-role retrieval model over a **shared input word-embedding matrix** (250002x1024 ~ 256M params, stored
-once). Validated finding: the word-embedding table is fully shareable between a bi-encoder embedder and a
-cross-encoder reranker (both built on `xlm-roberta-large`) at **zero quality loss**, while the transformer
-*layers* are NOT mergeable. So we keep each role's native layers + head and dedupe only the embedding table.
-
-- **Embed role:** bi-encoder. 1024-d, CLS-pool, L2-normalized. Use `"query: "` prefix on queries.
-- **Rerank role:** cross-encoder. Single relevance logit per (query, doc) pair.
-- **Router:** trivial — call `.embed(...)` or `.rerank(...)`.
-
-## Footprint
-~**0.77x** the two separate models (the 256M word-embedding table is stored once and tied at load), with
-**no quality loss** and **no retraining**. Disk: the reranker's word-embedding copy is stripped and injected
-from the embedder at load.
-
-## Eval (450-query ML-PDF holdout)
-| Role | hit@3 | note |
-|---|---|---|
-| Embed (dense retrieval) | 0.9511 | == standalone bi-encoder baseline |
-| Rerank (full pipeline)  | 0.9511 | == standalone cross-encoder baseline |
-Both identical to the two-model baseline; sharing the word-embedding table is a no-op on quality.
+
+A two-role retrieval model over a **shared input word-embedding matrix** (~256M params, stored once). The bi-encoder embedder and cross-encoder reranker are both built on the `xlm-roberta-large` backbone and **co-trained** to share a single word-embedding table at **no quality loss**, while each role keeps its native layers and head. 4-bit weight quantization is lossless on this stack (the embedder is the limiter; rerank tolerates lower bits).
+
+- **Embed role:** bi-encoder, 1024-d, L2-normalized. Prepend `"query: "` to queries.
+- **Rerank role:** cross-encoder, single relevance logit per (query, document) pair.
+- **Router:** call `.embed(...)` or `.rerank(...)`.
 
 ## Usage
+
 ```python
-from modeling_baa import BaaEmbeddingReranker
-m = BaaEmbeddingReranker()                       # loads embedder + reranker (shared word-emb)
-qv = m.embed(["my query"], is_query=True)        # 1024-d normalized
-dv = m.embed(["doc a", "doc b"])                 # doc embeddings
-ranked = m.rerank("my query", ["doc a","doc b"], top_k=10)   # [(doc, score), ...]
+from modeling_baa import BaaEmbeddingReranker   # included in this repo
+
+m = BaaEmbeddingReranker("baa-ai/Merino-Pro-4bit")
+qv = m.embed(["my query"], is_query=True)[0]     # 1024-d normalized
+dv = m.embed(["doc a", "doc b"])
+ranked = m.rerank("my query", ["doc a", "doc b"], top_k=10)   # [(doc, score), ...]
 ```
 
-## License
-- **BAA Contributions** (shared-embedding architecture, router/loader code, packaging, weights, docs) are
-  **proprietary to BAA AI (Black Sheep AI)** — see `LICENSE`.
+## License & attribution
+
+- **BAA Contributions** (shared-embedding architecture, router/loader code, packaging, weights, docs) are **proprietary to BAA AI (Black Sheep AI)** — see `LICENSE`.
 - Incorporates the `xlm-roberta-large` backbone under the **MIT License** — see `LICENSE-xlm-roberta-large.txt`.
 
-Build: shared-embedding MoE (Experiment A, 2026-06-22).
+© 2026 BAA AI (Black Sheep AI) — baa.ai. Provided "as is" without warranty.