---
## ๐ฅ ViDoRe V3 โ Rank #1
8 public domains ร 6 query languages, nDCG@10.
| # | Model | Params | Token Dim | **V3 public** |
| :---: | :--- | ---: | ---: | ---: |
| ๐ฅ **1** | **EVIE-Preview-4.5B** | 4.54B | 128D | **65.36** |
| ๐ฅ 2 | webAI-ColVec1.1-8b | 8.40B | 640D | 65.32 |
| ๐ฅ 3 | webAI-ColVec1.1-4b | 4.54B | 640D | 63.90 |
| 4 | nemotron-colembed-vl-8b-v2 | 8B | โ | 63.54 |
| 5 | tomoro-colqwen3-embed-8b | 8B | โ | 61.60 |
| 6 | nemotron-colembed-vl-4b-v2 | 4B | โ | 61.42 |
| 7 | tomoro-colqwen3-embed-4b | 4B | โ | 60.16 |
| 8 | llama-nemotron-colembed-vl-3b-v2 | 3B | โ | 59.70 |
| 9 | colnomic-embed-multimodal-7b | 7B | โ | 57.64 |
| 10 | jina-embeddings-v4 | ~3.8B | โ | 57.54 |
### Two deployment tiers, one checkpoint
| Visual tokens / page | **V3 public** | Vectors / page | Raw index / 1M pages (BF16) |
| :--- | ---: | ---: | ---: |
| 768 (training budget) | 64.56 | **751.62** | **179.2 GiB** |
| 1,792 (extrapolated) | **65.36** | 1,763.58 | 420.5 GiB |
**EVIE was trained at 768 visual tokens per page.** The 1,792 tier is pure test-time extrapolation โ the same weights, never trained or fine-tuned at that budget, and never re-exported. That the model does not merely hold up but *gains* 0.80 nDCG@10 at more than twice its training budget, improving in 7 of the 8 domains, is a direct read on how well its page representation generalises beyond the resolution it was fit to.
Pick whichever tier fits your compute budget.
The lighter tier holds a million pages in under 180 GiB.
### Per-domain breakdown
| Model | **Avg** | CompSci | Energy | Finance EN | Finance FR | HR | Industrial | Pharma | Physics |
| :--- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| ๐ฅ **EVIE-Preview-4.5B** | **65.36** | 80.65 | 71.36 | 70.50 | 54.44 | 67.34 | 58.76 | 69.20 | 50.62 |
| webAI-ColVec1.1-8b | 65.32 | 80.08 | 70.12 | 71.90 | 54.87 | 68.55 | 57.65 | 67.88 | 51.50 |
| webAI-ColVec1.1-4b | 63.90 | 80.34 | 69.50 | 69.18 | 53.13 | 66.90 | 56.36 | 67.25 | 51.24 |
| nemotron-colembed-vl-8b-v2 | 63.54 | 79.30 | 69.82 | 67.29 | 51.54 | 66.32 | 56.03 | 67.19 | 50.84 |
| tomoro-colqwen3-embed-8b | 61.60 | 75.35 | 68.41 | 65.08 | 49.10 | 63.98 | 54.41 | 66.36 | 50.13 |
| nemotron-colembed-vl-4b-v2 | 61.42 | 78.56 | 67.48 | 65.02 | 49.01 | 62.39 | 53.91 | 66.10 | 48.86 |
| llama-nemotron-colembed-vl-3b-v2 | 59.70 | 77.09 | 64.88 | 64.23 | 44.41 | 62.28 | 51.71 | 66.04 | 46.93 |
| colnomic-embed-multimodal-7b | 57.64 | 76.20 | 63.58 | 56.57 | 45.46 | 58.67 | 50.13 | 62.26 | 48.25 |
| jina-embeddings-v4 | 57.54 | 71.81 | 63.50 | 59.30 | 46.10 | 59.53 | 50.38 | 63.09 | 46.63 |
*EVIE rows measured with [`reproduce.sh`](reproduce.sh). Comparison rows are the vendors' published ViDoRe V3 public scores.*
---
## ๐ฅ ViDoRe V1 + V2 โ Rank #1
14 tasks, nDCG@5. **First place on the classic boards too.**
| # | Model | **Avg** | ArxivQA | DocVQA | InfoVQA | ShiftProj | SynAI | SynEnergy | SynGov | SynHealth | Tabfquad | Tatdqa | BioMed | ESGHL | ESG | Econ |
| :---: | :--- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| ๐ฅ **1** | **EVIE-Preview-4.5B** | **85.77** | 90.73 | 64.53 | 93.26 | 93.85 | 99.63 | 98.26 | 98.89 | 98.89 | 97.32 | 81.93 | 70.17 | 79.84 | 64.95 | 68.53 |
| ๐ฅ 2 | Ops-Colqwen3-4B | 84.90 | 91.80 | 66.50 | 94.00 | 90.80 | 99.60 | 97.30 | 98.00 | 99.60 | 93.60 | 82.40 | 65.50 | 78.60 | 66.00 | 64.50 |
| ๐ฅ 3 | nemotron-colembed-vl-8b-v2 | 84.80 | 93.10 | 68.10 | 94.60 | 93.30 | 100.0 | 97.90 | 98.90 | 99.60 | 97.70 | 83.40 | 66.20 | 73.20 | 60.60 | 60.80 |
| 4 | nemotron-colembed-vl-4b-v2 | 83.90 | 92.00 | 67.40 | 93.30 | 92.30 | 99.30 | 96.20 | 98.00 | 98.50 | 98.10 | 81.20 | 64.30 | 71.40 | 61.50 | 60.80 |
| 5 | colqwen3.5-4.5B-v3 | 83.70 | 91.90 | 66.60 | 93.60 | 90.20 | 100.0 | 97.10 | 97.30 | 98.90 | 95.90 | 84.00 | 65.30 | 73.80 | 58.00 | 59.90 |
| 6 | llama-nemotron-colembed-vl-3b-v2 | 83.60 | 90.40 | 67.20 | 94.70 | 92.00 | 100.0 | 98.00 | 98.00 | 98.90 | 97.30 | 81.00 | 63.20 | 73.10 | 58.60 | 58.60 |
| 7 | tomoro-colqwen3-embed-8b | 83.50 | 91.20 | 66.40 | 94.50 | 87.90 | 99.30 | 96.70 | 97.60 | 99.10 | 94.20 | 80.90 | 65.50 | 76.00 | 60.70 | 59.50 |
| 8 | EvoQwen2.5-VL-Retriever-7B-v1 | 83.40 | 91.50 | 65.10 | 94.10 | 88.80 | 99.60 | 96.60 | 96.30 | 98.90 | 93.60 | 82.30 | 65.20 | 77.00 | 59.70 | 59.10 |
| 9 | tomoro-colqwen3-embed-4b | 83.20 | 90.60 | 66.30 | 94.30 | 87.40 | 99.30 | 96.90 | 97.20 | 99.60 | 94.30 | 79.90 | 65.40 | 74.60 | 62.40 | 56.30 |
| 10 | SauerkrautLM-ColQwen3-8b-v0.1 | 82.90 | 93.80 | 64.70 | 94.50 | 90.40 | 98.60 | 96.50 | 96.80 | 99.30 | 92.20 | 84.00 | 63.30 | 70.80 | 57.90 | 58.00 |
*Tasks 1โ10: ViDoRe V1. Tasks 11โ14: ViDoRe V2. Board aggregates: **V1 91.73** ยท **V2 70.87**.*
---
## ๐พ Index Cost
Index size is what decides whether multi-vector retrieval actually ships. EVIE emits native **128D** token vectors, so the index stays compact at both page budgets.
| Raw BF16 index | 768 tokens/page | 1,792 tokens/page |
| :--- | ---: | ---: |
| 1M pages | **179.2 GiB** | 420.5 GiB |
| 10M pages | **1.8 TB** | 4.1 TB |
```text
1,763.58 vectors/page ร 128 dim ร 2 bytes ร 1,000,000 pages รท 2^30 = 420.5 GiB
```
Scoring stays cheap for the same reason: MaxSim is a late-interaction dot product over the token vectors, so a narrower vector cuts the scoring work exactly as it cuts storage.
---
## ๐ง Architecture
```text
Text Query โโโโโโโโบ ColQwen3_5 (BiDir Attn) โโโโโโบ Query Token Embeddings (128D)
โ
Late Interaction (MaxSim) โโโบ Relevance Score
โ
Document Image โโโโโโบ ColQwen3_5 (Dynamic Vision) โโโบ Doc Token Embeddings (128D)
```
1. **Vision-Language Backbone** โ `Qwen3.5-4B` with interleaved GatedDeltaNet linear attention and full attention.
2. **Compact Projection** โ contextual token states projected directly into native 128-dimensional representations.
3. **Late-Interaction Retrieval** โ token-level MaxSim between query tokens and document visual tokens.
### ๐ Multilingual
Queries in **English, French, German, Italian, Spanish, Portuguese and Chinese**, retrieving over charts, tables, scientific reports, financial filings and scanned forms โ including Japanese-language pages.
### ๐ฆ Model Footprint
| | Value |
| :--- | ---: |
| Parameters | 4.54B |
| Checkpoint (BF16) | 8.5 GB |
| Token embedding | 128D |
| Max visual tokens | 768 / 1,792 |
---
## โก Quick Start
ColPali Engine is the reference path โ every number on this card comes from it. A [Sentence Transformers](#sentence-transformers) path is also available for late-interaction pipelines already built on that API.
### Installation
```bash
pip install "colpali-engine>=0.3.15" accelerate
```
Or `pip install -r requirements.txt` if you cloned the repository; that adds `pyarrow`, which only [`reproduce.py`](reproduce.py) needs.
### Python Inference
Self-contained โ nothing to clone, no local files to prepare.
```python
import torch
from huggingface_hub import hf_hub_download
from PIL import Image
from colpali_engine.models import ColQwen3_5, ColQwen3_5Processor
model_id = "tencent/EVIE-Preview-4.5B"
def enable_bidirectional_attention(model):
"""Encoder-ize the full-attention layers; the GatedDeltaNet layers stay recurrent."""
for cfg in (model.config, getattr(model.config, "text_config", None)):
if cfg is not None:
cfg.is_causal = False
for module in model.modules():
if module.__class__.__name__ in ("Qwen3_5Attention", "Qwen3Attention"):
if hasattr(module, "is_causal"):
module.is_causal = False
# 1. Load model and enable bidirectional attention
model = ColQwen3_5.from_pretrained(
model_id,
torch_dtype=torch.bfloat16,
device_map="cuda",
attn_implementation="flash_attention_2",
).eval()
enable_bidirectional_attention(model)
# 2. Load processor
processor = ColQwen3_5Processor.from_pretrained(model_id)
# 3. Prepare inputs โ four example document pages
pages = [
hf_hub_download("sentence-transformers/example-documents", f"doc{i}.jpg", repo_type="dataset")
for i in range(1, 5)
]
images = [Image.open(p).convert("RGB") for p in pages]
queries = [
"What is the variable represented on the y-axis of the graph?",
"Total outlay is maximum in which year?",
]
image_batch = processor.process_images(images).to(model.device)
query_batch = processor.process_queries(queries).to(model.device)
# 4. Generate multi-vector embeddings and score
with torch.inference_mode():
image_embeddings = model(**image_batch)
model.rope_deltas = None # required before query forward
query_embeddings = model(**query_batch)
scores = processor.score(query_embeddings, image_embeddings)
print(scores)
# tensor([[17.3750, 10.9375, 7.8750, 7.3438],
# [ 6.5938, 13.3750, 6.2188, 6.0938]])
print("Best page per query:", scores.argmax(dim=1))
# Best page per query: tensor([0, 1])
```
> โ ๏ธ Apply `enable_bidirectional_attention(model)` once after loading, and reset `model.rope_deltas = None` before every query forward pass. Both are required to reach the scores above โ released `colpali-engine` (through 0.3.17) builds ColQwen3.5 with causal masks, which costs about 1.1 on top-hit MaxSim. The same helper ships as [`bidirectional.py`](bidirectional.py) for `infer.py` and `reproduce.py`.
### CLI
```bash
python infer.py --query "Quarterly revenue report" --image page_1.png --image page_2.png
```
### Sentence Transformers
EVIE also loads as a [Sentence Transformers](https://www.sbert.net/) `MultiVectorEncoder`, exposing the familiar `encode_query` / `encode_document` / `similarity` API with MaxSim scoring built in. Bidirectional attention is baked into the shipped configuration, so no extra call is needed.
`MultiVectorEncoder` requires Sentence Transformers 6.0.0, which is not on PyPI yet โ install from source until it is released:
```bash
pip install "sentence-transformers[image] @ git+https://github.com/huggingface/sentence-transformers.git"
```
```python
from sentence_transformers import MultiVectorEncoder
model = MultiVectorEncoder("tencent/EVIE-Preview-4.5B")
queries = [
"What is the variable represented on the y-axis of the graph?",
"Total outlay is maximum in which year?",
]
documents = [
"https://huggingface.co/datasets/sentence-transformers/example-documents/resolve/main/doc1.jpg",
"https://huggingface.co/datasets/sentence-transformers/example-documents/resolve/main/doc2.jpg",
"https://huggingface.co/datasets/sentence-transformers/example-documents/resolve/main/doc3.jpg",
"https://huggingface.co/datasets/sentence-transformers/example-documents/resolve/main/doc4.jpg",
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings[0].shape, document_embeddings[0].shape)
# torch.Size([23, 128]) torch.Size([755, 128])
scores = model.similarity(query_embeddings, document_embeddings)
print(scores)
# tensor([[17.3457, 10.8008, 7.8613, 7.3174],
# [ 6.5547, 13.3828, 6.2207, 6.0771]])
print("Best page per query:", scores.argmax(dim=1))
# Best page per query: tensor([0, 1])
```
Both paths above run on the same four example pages, so they are directly comparable. The two sets of scores agree closely; the small differences come from the attention backend and dtype, and the ranking is identical.
Documents may be file paths, URLs or `PIL.Image` objects. Text passed to `encode_document` is rendered as a query, since this model has no separate text-document format.
The default page budget is the 768-token tier. To score the 1,792-token tier, raise the pixel budget through `processor_kwargs`:
```python
model = MultiVectorEncoder(
"tencent/EVIE-Preview-4.5B",
model_kwargs={"attn_implementation": "flash_attention_2", "device_map": "cuda:0"},
processor_kwargs={"size": {"longest_edge": 1792 * 32 * 32, "shortest_edge": 65536}},
)
```
---
## ๐ฌ Reproducing
Every number on this card is reproducible with the shipped script across all visible GPUs:
```bash
bash reproduce.sh
```
On the first run, [`download_data.py`](download_data.py) fetches the 22 public ViDoRe datasets (~55 GB) from Hugging Face. To reuse an existing directory:
```bash
bash reproduce.sh /path/to/vidore
```
### Target aggregates
```text
ViDoRe V1 nDCG@5 91.73 (10 tasks)
ViDoRe V2 nDCG@5 70.87 (4 tasks)
ViDoRe V1+V2 nDCG@5 85.77 (14 tasks)
ViDoRe V3 public nDCG@10 64.56 (8 domains x 6 languages, 768 visual tokens)
ViDoRe V3 public nDCG@10 65.36 (8 domains x 6 languages, 1792 visual tokens)
```
To score the 1,792-token tier directly:
```bash
python -m torch.distributed.run --nproc_per_node=$(nvidia-smi -L | wc -l) reproduce.py \
--boards v3 --max-visual-tokens 1792 --data-root /path/to/vidore
```
---
## ๐ Training Details
EVIE was trained on approximately **0.8 million high-quality image-query pairs** spanning multilingual documents, technical reports, complex financial tables, infographics and document visual QA.
### Hard Negative Mining & Evidence Judging
Every mined negative is re-judged by a large multimodal judge before it reaches the loss:
- ๐ข Candidates that actually answer the query are **promoted to positives**.
- ๐ก Partially relevant or ambiguous candidates are **masked out of the loss**.
- ๐ด Only strictly irrelevant pages survive as **true hard negatives**.
Multi-positive rows are group-aware weighted by `1/positive_count` so that positives from the same query never penalise each other in-batch. Rows with empty queries, corrupted images or degraded text are dropped.
---
## ๐ Acknowledgements
- Built on the [ColPali Engine](https://github.com/illuin-tech/colpali) by Illuin Technology.
- Powered by the [Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B) vision-language backbone.
- Evaluated on the [ViDoRe Benchmark](https://huggingface.co/vidore) family.
---
## ๐ Citation
```bibtex
@misc{tencent2026evie,
title = {EVIE-Preview-4.5B: Rank-1 Multilingual Visual Document Retrieval with 128-Dimensional Multi-Vector Embeddings},
author = {{Tencent}},
year = {2026},
howpublished = {\url{https://huggingface.co/tencent/EVIE-Preview-4.5B}}
}
```
**๐ Rank #1 on ViDoRe V3 ยท ๐ Rank #1 on ViDoRe V1+V2 ยท ๐พ native 128D token vectors**
