---
library_name: transformers
license: mit
base_model: Qwen/Qwen2.5-VL-3B-Instruct
arxiv: 2602.21202
datasets:
  - vidore/colpali_train_set
language:
  - en
tags:
  - visual-document-retrieval
  - multi-vector
  - late-interaction
  - colbert
  - index-compression
  - memory-tokens
  - text-to-image
  - feature-extraction
---

# MemTok-Qwen2.5-VL-3B (ViDoRe)

This model uses **Memory Tokens (MemTok)** to compress multi-vector visual document representations for efficient ColBERT-style late interaction retrieval. Model weights are initialized from [Qwen2.5-VL-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct) and finetuned on the [ColPali train set](https://huggingface.co/datasets/vidore/colpali_train_set) for text-to-visual-document retrieval with bidirectional attention.

MemTok compresses ~1300 visual document token vectors into a fixed budget of **64 vectors** (95.1% compression) via learnable memory tokens that aggregate document information through attention.

[![arXiv](https://img.shields.io/badge/arXiv-2602.21202-b31b1b.svg)](https://arxiv.org/abs/2602.21202)
[![GitHub](https://img.shields.io/badge/GitHub-omni--col--press-blue?logo=github)](https://github.com/hanxiangqin/omni-col-press)
[![License](https://img.shields.io/badge/License-MIT-green.svg)](https://opensource.org/licenses/MIT)

## Method Overview

MemTok appends a set of *m* learnable memory tokens to the document token sequence. The concatenated sequence is encoded with a bidirectional transformer; after self-attention, each memory token has attended over the full document. The final hidden states of the *m* memory tokens form the compressed multi-vector representation used for ColBERT-style MaxSim retrieval.

![Method](https://cdn-uploads.huggingface.co/production/uploads/683d3519fabaa8b10395bf8a/eC_E_E9kwuj_hp-5xwc2B.png)

## Results on ViDoRe v2

| Method | Tokens | nDCG@5 (Avg) | Bio | Econ | ESG-R | ESG-H |
|---|---|---|---|---|---|---|
| ColPali | – | 53.3 | 56.5 | 49.9 | 55.7 | 51.1 |
| ColQwenOmni | – | 56.5 | 56.5 | 53.2 | 54.2 | 62.2 |
| MetaEmbed | 64 | 58.8 | 58.7 | 55.5 | 57.4 | 63.7 |
| Baseline (Ours, uncompressed) | 1297 | 60.0 | 61.4 | 53.9 | 57.0 | 67.6 |
| SeqResize | 64 | 51.7 | 54.7 | 53.5 | 45.2 | 53.5 |
| **MemTok (This model)** | **64** | **54.3** | **56.8** | **53.0** | **46.4** | **61.4** |
| H-Pool | 64 | 56.4 | 59.6 | 52.1 | 53.4 | 60.6 |
| AGC | 64 | 56.7 | 59.0 | 54.5 | 55.8 | 57.3 |

## Model Details

| | |
|---|---|
| **Initial weights** | Qwen2.5-VL-3B-Instruct |
| **Architecture** | Qwen2.5-VL with bidirectional attention |
| **Hidden dimension** | 2048 |
| **Compression method** | MemTok (memory tokens) |
| **Memory tokens** | 64 learned tokens (`<\|mem0\|>` – `<\|mem63\|>`) appended to document |
| **Budget** | 64 vectors per document |
| **Scoring** | ColBERT-style MaxSim (late interaction) |
| **Normalization** | L2-normalized embeddings |
| **Query prefix** | `"Query: "` |
| **Passage prefix** | `"Passage: "` |
| **Precision** | bfloat16 |
| **Max image tokens** | 1280 |

## Usage

```python
import torch
from transformers import AutoProcessor
from qwen_vl_utils import process_vision_info

from src.arguments import ModelArguments
from src.encoder.multivec_encoder import MultiVecEncoder
from src.models.qwen2_5_vl_embed.qwen2_5_vl_embed import Qwen2_5ForEmbedding
from src.utils import get_appending_token_strings

MODEL_ID = "hltcoe/MemTok_qwen2.5-vl_colpali"
IMAGE_PATH = "PLACEHOLDER"
NUM_MEMORY_TOKENS = 64
APPENDING_SUFFIX = "".join(get_appending_token_strings(NUM_MEMORY_TOKENS))

# --- Setup ---
model_args = ModelArguments(
    model_name_or_path=MODEL_ID,
    pooling="memory",
    normalize=True,
    num_appending_token=NUM_MEMORY_TOKENS,
    use_parametric_appending_tokens=True,
    attn_implementation="flash_attention_2",
)

processor = AutoProcessor.from_pretrained(MODEL_ID)
model = MultiVecEncoder.load(
    Qwen2_5ForEmbedding,
    model_args,
    attn_implementation=model_args.attn_implementation,
    dtype=torch.bfloat16,
)
model = model.to("cuda").eval()

# --- Encode an image document ---
passage_messages = [
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "Passage: "},
            {"type": "image", "image": IMAGE_PATH, "max_pixels": 1003520, "min_pixels": 614656},
        ],
    }
]
text = processor.apply_chat_template(passage_messages, tokenize=False, add_generation_prompt=False)
text += APPENDING_SUFFIX
image_inputs, video_inputs = process_vision_info(passage_messages)
passage_inputs = processor(
    text=[text], images=image_inputs, videos=video_inputs, padding=True, return_tensors="pt",
).to("cuda")

with torch.amp.autocast(device_type="cuda", dtype=torch.bfloat16):
    with torch.inference_mode():
        doc_embeddings, doc_mask = model.encode(passage_inputs, is_query=False)
        print(doc_embeddings.shape)
        # doc_embeddings: (1, 64, 2048) — 64 MemTok vectors

# --- Encode a text query ---
query_messages = [{"role": "user", "content": [{"type": "text", "text": "Query: What types of tissues are unable to regenerate spontaneously?"}]}]
query_text = processor.apply_chat_template(query_messages, tokenize=False, add_generation_prompt=False)
query_inputs = processor(text=[query_text], padding=True, return_tensors="pt").to("cuda")

with torch.amp.autocast(device_type="cuda", dtype=torch.bfloat16):
    with torch.inference_mode():
        query_embeddings, query_mask = model.encode(query_inputs, is_query=True)
        print(query_embeddings.shape)

# --- ColBERT MaxSim scoring ---
score = model.compute_similarity(query_embeddings, doc_embeddings, query_mask, doc_mask)
print(f"Similarity score: {score.item():.4f}")
```

## Command line usage

For running inference and evaluation from the command line, see the [Quick Start](https://github.com/HanxiangQin/omni-col-press?tab=readme-ov-file#quick-start-1) section.

## Citation

```bibtex
@misc{qin2026multivectorindexcompressionmodality,
      title={Multi-Vector Index Compression in Any Modality}, 
      author={Hanxiang Qin and Alexander Martin and Rohan Jha and Chunsheng Zuo and Reno Kriz and Benjamin Van Durme},
      year={2026},
      eprint={2602.21202},
      archivePrefix={arXiv},
      primaryClass={cs.IR},
      url={https://arxiv.org/abs/2602.21202}, 
}
```