No official torrent yet β requesting it tells the flock what to host next. You cannot seed a catalog-only page.
Cataloged from tencent/EVIE-8B
All models by tencent β
[blocked]
[blocked]
4096-Dimensional Token Representation β’ Bidirectional Full-Attention β’ Teacher for EVIE-4.5B Prefix-MRL
π€ EVIE-8B (Flagship Teacher) Β β’Β π€ EVIE-4.5B (Prefix-MRL & HAC) Β β’Β π GitHub: Tencent/EVIE
π’ Release Announcement: All model weights, inference pipelines, and evaluation suites have been fully open-sourced. Full technical details, architectural ablations, and the formal research paper will be updated in an upcoming release. Training code and distillation recipes are available in the Tencent/EVIE repository.
Official ranking on the ViDoRe leaderboard (ViDoRe V3, Mean Task).
Performance comparison across modern multi-vector late-interaction visual document retrievers on ViDoRe:
Protocol paired-all-pages-dedup+process_queries+ndcg2r-20260827 ($\text{MVT} = 1024$, bidirectional attention):
Load EVIE with Sentence Transformers to encode queries and document images and compute MaxSim scores. Bidirectional attention is configured automatically.
Documents can be URLs, local image paths, or PIL.Image objects. The output contains 4096-dimensional token embeddings. Scores can vary slightly with dtype and attention backend.
The default page budget allows up to 16,384 visual tokens. Encoding images one at a time reduces peak memory use. To use the 1,024-token budget from the evaluation protocol above, set the image processor's pixel budget:
EVIE-8B was trained on 775,635 document-query pairs across multilingual domains. Hard negatives were mined and classified into answerable (promoted to positives), ambiguous (masked), and strict negatives (negative cross-entropy loss).
The released a40 checkpoint represents a weight-space mix ($\alpha = 0.40$) of two independently trained arms, achieving the peak 66.75 nDCG@10 on ViDoRe V3. Full training reproduction scripts are provided in the companion repository Tencent/EVIE.
This model and codebase are released under the Apache-2.0 License [blocked].
Query Text βββββββββΊ ColQwen3.5 (BiDir Attention) βββββΊ 4096D Per-Token Embeddings β MaxSim Matching β Doc Image βββββββββΊ ColQwen3.5 (Vision Encoder) βββββΊ Full Multi-Vector Patch Embeddingsgit clone https://github.com/Tencent/EVIE.gitcd EVIEpip install -r requirements.txtexport PYTHONPATH="$(pwd)/colpali${PYTHONPATH:+:$PYTHONPATH}"import torchfrom PIL import Imagefrom colpali_engine.models import ColQwen3_5, ColQwen3_5Processor
model_id = "tencent/EVIE-8B"
# 1. Load model with FlashAttention and bidirectional attentionmodel = ColQwen3_5.from_pretrained( model_id, torch_dtype=torch.bfloat16, device_map="cuda", attn_implementation="flash_attention_2",).eval()model.enable_bidirectional_attention()
# 2. Process query and document imageprocessor = ColQwen3_5Processor.from_pretrained(model_id)images = [Image.open("examples/demo/pages/q3_revenue.png").convert("RGB")]queries = ["What is the total quarterly revenue?"]
image_batch = processor.process_images(images).to(model.device)query_batch = processor.process_queries(queries).to(model.device)
# 3. Generate 4096D multi-vectors and late-interaction scorewith torch.inference_mode(): image_embeddings = model(**image_batch) model.rope_deltas = None # Reset RoPE deltas before text query forward query_embeddings = model(**query_batch)
scores = processor.score(query_embeddings, image_embeddings)print("Late-interaction MaxSim Relevance Score:", scores)pip install -U "sentence-transformers[image]>=6.0.0"from sentence_transformers import MultiVectorEncoder
model = MultiVectorEncoder("tencent/EVIE-8B")
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, batch_size=1)print(query_embeddings[0].shape, document_embeddings[0].shape)# torch.Size([23, 4096]) torch.Size([3161, 4096])
scores = model.similarity(query_embeddings, document_embeddings)print(scores)# tensor([[14.3594, 6.1758, 4.5137, 3.2979],# [ 1.8079, 10.9180, 1.9668, 1.9541]])model = MultiVectorEncoder( "tencent/EVIE-8B", processor_kwargs={"size": {"longest_edge": 1024 * 32 * 32, "shortest_edge": 65536}},)Evie-8B/βββ model.safetensors # EVIE-8B weights (safetensors)βββ config.json # Model configuration (4096D)βββ 1_Dense/ # Sentence Transformers 4096D projectionβββ modules.json # MultiVectorEncoder module graphβββ infer.py # Standalone inference & scoring CLIβββ colpali/ # ColQwen3.5 inference engineβββ code/shared/ # 138-task eval harnessβββ examples/demo/ # 8-page retrieval demo (run.sh)βββ env.sh.example # Environment variables template@misc{tencent2026evie8b, title = {EVIE-8B: High-Capacity Visual Document Retrieval with 4096-Dimensional Embeddings}, author = {Wang, Zifei and Wen, Wei}, year = {2026}, howpublished = {\url{https://github.com/Tencent/EVIE}}, note = {Corresponding author: Wei Wen <jawnrwen@tencent.com>}}