File size: 12,187 Bytes
e05599b 3dae150 38c1507 e05599b 3dae150 e05599b 3dae150 e05599b 3dae150 e05599b 3dae150 46491b5 3dae150 46491b5 3d38f58 38c1507 46491b5 ead974a 46491b5 79c202e e37d027 79c202e e37d027 79c202e 25235d6 79c202e 3dae150 7178658 3dae150 76ebefc e05599b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 | ---
license: apache-2.0
pipeline_tag: feature-extraction
library_name: sentence-transformers
tags:
- transformers
- sentence-transformers
- feature-extraction
- multimodal-embedding
---
# LCO-Embedding: Scaling Language-Centric Omnimodal Representation Learning
We are thrilled to release LCO-Embedding - a language-centric omnimodal representation learning framework and the LCO-Embedding model families!
This model implements the framework presented in the paper [Scaling Language-Centric Omnimodal Representation Learning](https://huggingface.co/papers/2510.11693), accepted by NeurIPS 2025.
**Project Page:** https://huggingface.co/LCO-Embedding
**Github Repository:** https://github.com/LCO-Embedding/LCO-Embedding
## Quick Start
Note: We are only using the `thinker` component of Qwen2.5 Omni and drops the `talker` component.
### Using Sentence Transformers
Install Sentence Transformers:
```bash
pip install "sentence_transformers[image]"
```
```python
import torch
from sentence_transformers import SentenceTransformer
model = SentenceTransformer(
"LCO-Embedding/LCO-Embedding-Omni-7B",
trust_remote_code=True,
model_kwargs={"dtype": torch.bfloat16},
)
# The same "Summarize the above <modality> in one word:" instruction used in
# the paper is baked into the chat template, so encode() takes plain text or
# multimodal dicts directly.
texts = [
"The capital of France is Paris.",
"Paris is the capital city of France.",
"The Eiffel Tower is located in Paris.",
"Berlin is the capital of Germany.",
]
text_embeddings = model.encode(texts)
print(text_embeddings.shape)
# (4, 3584)
text_similarities = model.similarity(text_embeddings, text_embeddings)
print(text_similarities)
# tensor([[1.0000, 0.9453, 0.6885, 0.5223],
# [0.9453, 1.0000, 0.7283, 0.5434],
# [0.6885, 0.7283, 1.0000, 0.3772],
# [0.5223, 0.5434, 0.3772, 1.0000]])
# Encoding images (text, audio, and video also work, individually or combined using a dict input):
image_embeddings = model.encode([
"path/to/image_1.png",
"path/to/image_2.png",
])
print(image_embeddings.shape)
# (2, 3584)
# Multimodal inputs can mix modalities via dicts (text + image + audio + video):
queries = ["A diagram of the Qwen2.5-Omni architecture"]
documents = [
{"image": "path/to/qwen_diagram.png"},
{"text": "Llama 4 architecture overview", "image": "path/to/llama_diagram.png"},
]
query_embeddings = model.encode(queries)
document_embeddings = model.encode(documents)
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities.shape)
# torch.Size([1, 2])
```
### Using Transformers
```python
from transformers import Qwen2_5OmniThinkerForConditionalGeneration, Qwen2_5OmniProcessor
from qwen_omni_utils import process_mm_info
processor = Qwen2_5OmniProcessor.from_pretrained("LCO-Embedding/LCO-Embedding-Omni-7B") # or add a `max_pixels = 1280*28*28' for efficient encoding
model = Qwen2_5OmniThinkerForConditionalGeneration.from_pretrained("LCO-Embedding/LCO-Embedding-Omni-7B",
torch_dtype=torch.bfloat16,
device_map="auto")
```
#### Text Batch Encodings:
```python
texts = ["some random text", "a second random text", "a third random text"] * 30
batch_size = 8
text_prompt = "{}\nSummarize the above text in one word:"
all_text_embeddings = []
with torch.no_grad():
for i in tqdm(range(0, len(texts), batch_size)):
batch_texts = texts[i : i + batch_size]
batch_texts = [text_prompt.format(text) for text in batch_texts]
messages = [[
{
"role": "user",
"content": [
{"type": "text", "text":text},
],
}
] for text in batch_texts]
text_inputs = processor.apply_chat_template(messages, tokenize = False, add_generation_prompt = True)
text_inputs = processor(
text = text_inputs,
padding = True,
return_tensors = "pt",
)
text_inputs = text_inputs.to("cuda")
text_outputs = model(
**text_inputs, output_hidden_states=True, return_dict=True
).hidden_states[-1][:, -1, :]
all_text_embeddings.append(text_outputs.to(torch.float16).cpu())
all_text_embeddings = torch.cat(all_text_embeddings, dim=0)
```
#### Image Batch Encodings:
```python
images = [some random PIL.Image] * 100 # will be good to load them using dataloader; see MIEB evaluation pipeline
image_prompt = "\nSummarize the above image in one word:"
batch_size = 8
all_image_embeddings = []
with torch.no_grad():
for i in tqdm(range(0, len(images), batch_size)):
batch_images = images[i : i + batch_size]
messages = [[
{
"role": "user",
"content": [
{"type": "image", "image":image},
{"type": "text", "text": image_prompt},
],
}
] for image in batch_images]
text = processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
audio_inputs, image_inputs, video_inputs = process_mm_info(messages, use_audio_in_video=True)
inputs = processor(
text=text,
audio=audio_inputs,
images=image_inputs,
videos=video_inputs,
return_tensors="pt",
padding=True
)
inputs = inputs.to("cuda")
image_outputs = model(
**inputs, output_hidden_states=True, return_dict=True
).hidden_states[-1][:, -1, :]
all_image_embeddings.append(image_outputs.to(torch.float16).cpu())
all_image_embeddings = torch.cat(all_image_embeddings, dim=0)
```
#### Audio Batch Encoding:
```python
import logging
logging.getLogger("root").setLevel(logging.ERROR)
# set this to prevent getting the Qwen Omni system prompt mismatch warning.
batch_size = 4
audio_prompt = "\nSummarize the above audio in one word:"
audis = [some audios] * 1000
all_audio_embeddings = []
with torch.no_grad():
for i in tqdm(range(0, len(audios), batch_size)):
torch.cuda.empty_cache()
batch_audios = audios[i : i + batch_size]
messages = [[
{
"role": "user",
"content": [
{"type": "audio", "audio": audio},
{"type": "text", "text": audio_prompt},
],
}
] for audio in batch_audios]
text = self.processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
audio_inputs, image_inputs, video_inputs = process_mm_info(
messages, use_audio_in_video=False
)
inputs = self.processor(
text=text,
audio=audio_inputs,
images=image_inputs,
videos=video_inputs,
return_tensors="pt",
padding=True
)
inputs = inputs.to("cuda")
audio_outputs = self.model(
**inputs, output_hidden_states=True, return_dict=True
).hidden_states[-1][:, -1, :]
all_audio_embeddings.append(audio_outputs.to(torch.float16).cpu())
del inputs, audio_outputs
torch.cuda.empty_cache()
all_audio_embeddings = torch.cat(all_audio_embeddings, dim=0)
```
#### Video Batch Encoding:
```python
videos = [some videos] * 1000
video_prompt = "\nSummarize the above video in one word:"
batch_size = 4
long_video = False
# followed by some example hyperparameters to save RAM
# for long videos. Not optimal. Tune case by case.
all_video_embeddings = []
with torch.no_grad():
for i in tqdm(range(0, len(videos), batch_size)):
torch.cuda.empty_cache()
batch_videos = videos[i : i + batch_size]
if long_video:
messages = [[
{
"role": "user",
"content": [
{
"type": "video",
"video": video,
"max_pixels": 224 * 224,
"fps": 1,
"max_frames": 10
},
{"type": "text", "text": video_prompt},
],
}
] for video in batch_videos]
else:
messages = [[
{
"role": "user",
"content": [
{
"type": "video",
"video": video,
},
{"type": "text", "text": video_prompt},
],
}
] for video in batch_videos]
text = self.processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
audio_inputs, image_inputs, video_inputs = process_mm_info(
messages, use_audio_in_video=False
)
inputs = self.processor(
text=text,
audio=audio_inputs,
images=image_inputs,
videos=video_inputs,
return_tensors="pt",
padding=True
)
inputs = inputs.to("cuda")
video_outputs = self.model(
**inputs, output_hidden_states=True, return_dict=True
).hidden_states[-1][:, -1, :]
all_video_embeddings.append(video_outputs.to(torch.float16).cpu())
del inputs, video_outputs
torch.cuda.empty_cache()
all_video_embeddings = torch.cat(all_video_embeddings, dim=0)
```
## Overview
We introduce **LCO-Embedding**, a language-centric omnimodal representation learning method and the LCO-Embedding model families, setting a new state-of-the-art on [MIEB](https://huggingface.co/blog/isaacchung/introducing-mieb) (Massive Image Embedding Benchmark), while supporting audio and videos.
This work also introduces the **Generation-Representation Scaling Law**, connecting models' generative capabilities and their representation upper bound. Furthermore, we introduce **SeaDoc**, a challenging visual document retrieval task in Southeast Asian languages, and show that continual generative pretraining before contrastive learning raises the representation upper bound.
<div align='center'><img src="https://cdn-uploads.huggingface.co/production/uploads/604f67ef0fe8ff3ec13d71ef/4Wd8fDFBdT6GxqN6-KzZN.png" alt="overview" width="100%"/></div>
## Evaluation Results
We evaluate LCO-Embedding with state-of-the-art embedding models, including E5-V, Voyage Multimodal 3, mmE5, and GME, on a MIEB-Lite benchmark (51 tasks) broken down by task categories.
<div align='center'><img src="https://cdn-uploads.huggingface.co/production/uploads/63108cc834c7d77420b0fd68/63WBsKh57HbNwwe3bZ-oZ.png" alt="mieb_lite" width="100%"/></div>
LCO-Embedding is also SOTA on MAEB (massive audio embedding benchmark) without even training on audio. Screenshot from the MAEB paper.
Performance and efficiency comparisons of different training strategies using 3B and 7B variants of Qwen2.5-VL backbones.
<div align='center'><img src="https://github.com/LCO-Embedding/LCO-Embedding/raw/main/assets/lora_ablation.png" alt="lora_ablation" width="100%"/></div>
Scaling relationship between generation benchmark performance (X-axis) and representation benchmark performance after language-centric contrastive learning (Y-axis).
<div align='center'><img src="https://github.com/LCO-Embedding/LCO-Embedding/raw/main/assets/scaling.png" alt="scaling_law" width="100%"/></div>
## Citation
If you find LCO-Embedding useful for your research and applications, please cite using this BibTeX:
```bibtex
@article{xiao2025scaling,
title={Scaling Language-Centric Omnimodal Representation Learning},
author={Xiao, Chenghao and Chan, Hou Pong and Zhang, Hao and Xu, Weiwen and Aljunied, Mahani and Rong, Yu},
journal={arXiv preprint arXiv:2510.11693},
year={2025}
}
``` |