EVA01 is a unified native 3D framework for mesh understanding, shape generation, and context-aware editing. It integrates 3D meshes as a native modality through a Mixture-of-Transformers architecture with an Understanding Expert, a structurally mirrored Generation Expert, shared global self-attention, and hard modality routing.

EVA01-2B-Instruct is the UND-side Full checkpoint release. It is intended for native 3D mesh understanding and open-ended question answering over .glb mesh input.

Model Details

Item	Description
Model	`SEELE-AI/EVA01-2B-Instruct`
Release type	UND-side Full checkpoint
Backbone	Qwen3-VL language backbone
3D input	`.glb` mesh input through the EVA01 mesh UND processor
Components	Qwen3-VL weights, EVA01 mesh UND encoder, connector, tokenizer, processor, and config files
Training recipe	Alignment followed by instruction tuning
Code	SeeleAI/OpenEVA
Project page	EVA01
Paper	arXiv:2605.16745

Installation

EVA01 requires the OpenEVA code package in addition to the checkpoint files.

bash

git clone https://github.com/SeeleAI/OpenEVA.git
cd OpenEVA/EVA01
bash install.sh
source .venv/bin/activate

If a different CUDA wheel index is required, set TORCH_INDEX_URL before running the install script.

TORCH_INDEX_URL=https://download.pytorch.org/whl/cu121 bash install.sh

Quick Inference

CLI:

bash

python infer.py \
  --checkpoint SEELE-AI/EVA01-2B-Instruct \
  --mesh assets/examples/construction_backhoe.glb \
  --question "Describe this 3D object in detail."

Python API:

python

import torch
from eva01 import EVA01ForConditionalGeneration, EVA01Processor

model = EVA01ForConditionalGeneration.from_pretrained(
    "SEELE-AI/EVA01-2B-Instruct",
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
processor = EVA01Processor.from_pretrained("SEELE-AI/EVA01-2B-Instruct")

messages = [{
    "role": "user",
    "content": [
        {"type": "mesh", "mesh": "assets/examples/construction_backhoe.glb"},
        {"type": "text", "text": "Describe this 3D object in detail."},
    ],
}]

inputs = processor.apply_chat_template(
    messages,
    tokenize=True,
    add_generation_prompt=True,
    return_dict=True,
    return_tensors="pt",
).to(model.device)

output_ids = model.generate(**inputs, max_new_tokens=128, do_sample=False)
text = processor.batch_decode(
    output_ids[:, inputs.input_ids.shape[1]:],
    skip_special_tokens=True,
    clean_up_tokenization_spaces=False,
)[0]
print(text)

The public mesh token is <|mesh_und_pad|>. The processor returns input_ids, attention_mask, and mesh_und_values.

Gradio Chat

python app.py --host 127.0.0.1 --port 7860

The app loads this Full checkpoint by default, supports uploaded .glb files, and includes 10 built-in TexVerse examples. The OpenEVA GitHub repository also contains a PBR-rendered example gallery.

PointLLM-200 Evaluation

The OpenEVA eval script downloads the PointLLM-200 benchmark files staged in this checkpoint repo and writes results under EVA01/outputs/pointllm200/.

python eval_pointllm200.py --variant full

The deterministic path computes BLEU, ROUGE, METEOR, Sentence-BERT, and SimCSE with fixed seed 20260615 and greedy generation. GPT-ref and GPT-img judge paths are available when OPENAI_API_KEY is set.

The PointLLM-200 benchmark source is cited as RunsenXu/PointLLM.

Metrics

All metrics below are reported on PointLLM-200 with 200 samples, fixed seed 20260615, and greedy decoding. GPT-ref and GPT-img are judge metrics and may vary with judge model and API settings.

Model	B-1	B-4	R-L	METEOR	SBERT	SimCSE	GPT-ref	GPT-img
PointLLM-13B	7.873	0.649	10.519	13.620	47.539	48.602	51.735	49.745
ShapeLLM-13B	10.542	1.050	12.954	14.234	39.935	40.728	33.925	35.870
ShapeLLM-Omni	11.326	1.197	14.190	13.276	34.617	35.115	25.625	20.190
EVA01-2B-Instruct	6.386	0.589	9.443	13.505	50.651	50.767	59.045	70.335
EVA01-2B-Instruct-LoRA	6.372	0.607	9.455	13.567	51.194	51.320	59.560	71.480

Intended Use

This checkpoint is intended for research and development around 3D mesh understanding, 3D asset captioning, and mesh-grounded question answering. It expects mesh input through the EVA01 processor and should be used with the OpenEVA runtime/API.

Limitations

The public release focuses on the UND-side path.
Model quality depends on mesh geometry, scale, topology, materials, and texture availability.
Outputs may contain incorrect or unsupported details when the mesh is ambiguous, incomplete, or visually underspecified.
GPT-ref and GPT-img scores are judge references and are not bitwise reproducible across judge model or API changes.

Citation

bibtex

@misc{eva01_2026,
  title  = {EVA01: Unified Native 3D Understanding and Generation via Mixture-of-Transformers},
  author = {Zongyuan Yang and Mingjing Yi and Wanli Ma and Chenzhuo Fan and Bocheng Li and Baolin Liu and Yuke Lou and Yingde Song and Yongping Xiong and Zhengdong Guo and Shimu Wang},
  year   = {2026},
  eprint = {2605.16745},
  archivePrefix = {arXiv},
  primaryClass = {cs.CV}
}