Update README.md

README.md

@@ -45,10 +45,12 @@ license_link: LICENSE
 
 # LFM2.5-ColBERT-350M
 
-We release two new multilingual retrieval models built on [LFM2.5-350M-Base](https://huggingface.co/LiquidAI/LFM2.5-350M-Base): **LFM2.5-ColBERT-350M**, a late-interaction retriever, and **[LFM2.5-Embedding-350M](https://huggingface.co/LiquidAI/LFM2.5-Embedding-350M)**, a dense bi-encoder. Both are 350M params and the first bidirectional members of the LFM family, targeting fast, cheap, and reliable multilingual / cross-lingual search across 11 languages.
+We release two new multilingual retrieval models:
 
-- **[LFM2.5-Embedding-350M](https://huggingface.co/LiquidAI/LFM2.5-Embedding-350M)** — one vector per document. Smallest, fastest index.
-- **LFM2.5-ColBERT-350M** — one vector per *token*, matched via MaxSim. Higher accuracy and better generalization at the cost of index size.
+- **[LFM2.5-Embedding-350M](https://huggingface.co/LiquidAI/LFM2.5-Embedding-350M)** — A dense bi-encoder, one vector per document. Smallest, fastest index.
+- **LFM2.5-ColBERT-350M** — A late-interaction model. One vector per *token*, matched via MaxSim. Higher accuracy and better generalization at the cost of index size.
+
+Both models are 350M params and the first bidirectional members of the LFM family, built on [LFM2.5-350M-Base](https://huggingface.co/LiquidAI/LFM2.5-350M-Base). They target fast, cheap, and reliable multilingual / cross-lingual search across 11 languages.
 
 For the full story on architecture (bidirectional patches over LFM2.5), training recipe, and benchmarks, see the [release blog post](https://www.liquid.ai/blog) *(link to be added)*.
 
@@ -58,7 +60,7 @@ For the full story on architecture (bidirectional patches over LFM2.5), training
 | --------------------- | -------------------------------------- | ----------------------------------- |
 | **Type**              | Late interaction (per-token vectors)   | Dense bi-encoder (single vector)    |
 | **Total parameters**  | ~353M                                  | ~350M                               |
-| **Layers**            | 17 (10 conv + 6 attn + 1 dense)        | 16 (10 conv + 6 attn)               |
+| **Layers**            | 17 (10 conv + 6 attn + 1 dense)        | 16 (10 conv + 6 attn + 1 pool)               |
 | **Context length**    | 32,768 tokens                          | 32,768 tokens                       |
 | **Vocabulary size**   | 64,402                                 | 64,402                              |
 | **Output**            | 128-dim per token                      | 1024-dim CLS vector                 |
@@ -66,7 +68,7 @@ For the full story on architecture (bidirectional patches over LFM2.5), training
 | **Training precision**| BF16                                   | BF16                                |
 | **License**           | LFM Open License v1.0                  | LFM Open License v1.0               |
 
-**Document length:** 512 tokens    **Query length:** 32 tokens (ColBERT)
+**Document length:** 512 tokens    **Query length:** 32 tokens
 
 **Supported languages:** English, Spanish, German, French, Italian, Portuguese, Arabic, Swedish, Norwegian, Japanese, Korean.