VeritasPP

KO | EN

gitlite — search

#open-source #windows #sql #rust #http #nextjs #linux #pytorch #video #shell

(★ 9)

No description available.

File Explorer

ZIP Download

- banner.png
- ding.png
- discord_qr.jpg
- wechat.png
- - asyncengine.png
  - deepeyes.png
  - dpo_data.png
  - grpo.png
  - grpo_clevr_count.png
  - grpo_code.png
  - grpo_countdown.png
  - grpo_countdown_1.png
  - grpo_geoqa.png
  - grpo_multi_turn.png
  - grpo_openr1_multimodal.png
  - gym_env.png
  - kto_data.png
  - multiturn_pipeline.png
  - sapo_tau.png
  - treepo.png
  - web-ui-en.jpg
  - web-ui.jpg
- - .readthedocs.yaml
  - conf.py
  - index.rst
- - .readthedocs.yaml
  - conf.py
  - index.rst
- make.bat
- Makefile
- README.md
- - mllm.sh
- - dataset.py
  - infer.sh
  - model.py
  - model_hf.py
  - sft.sh
- - README.md
  - sglang.sh
  - vllm.sh
  - vllm_dp.sh
- - merge_lora.sh
  - ollama.sh
  - push_to_hub.sh
- - cli_demo.sh
  - demo.py
  - demo_mllm.py
  - demo_reward_model.py
  - demo_vllm_reasoning_parser.py
- - base_to_chat.sh
  - long_text.sh
  - muon.sh
  - pretrain.sh
  - sft.sh
- - infer.sh
  - lora_sft.sh
  - on_policy_distillation.sh
- README.md
- method_overview2.png
- vaopd.png
- dependency_links.txt
- entry_points.txt
- not-zip-safe
- PKG-INFO
- requires.txt
- SOURCES.txt
- top_level.txt
- docs.txt
- eval.txt
- framework.txt
- install_all.sh
- ray.txt
- swanlab.txt
- tests.txt
- - exp.py
  - exp_utils.py
  - generate_report.py
- - plot_loss.py
  - run_dataset_info.py
  - run_model_info.py
  - run_template.py
  - test_link_valid.py
- - deploy_model.sh
- - infer_vllm_single.py
- - __init__.py
  - base.py
  - deepseek_v3_1.py
  - extra.py
  - glm4.py
  - hermes.py
  - llama.py
  - mapping.py
  - minimax_m2.py
  - mistral.py
  - qwen.py
  - qwen3_coder.py
  - react.py
  - seed_oss.py
  - toolbench.py
  - youtu.py
- - __init__.py
  - app_args.py
  - deploy_args.py
  - eval_args.py
  - export_args.py
  - infer_args.py
  - merge_args.py
  - pretrain_args.py
  - sft_args.py
  - tuner_args.py
  - webui_args.py
- - __init__.py
  - activation_cpu_offload.py
  - adalora.py
  - base.py
  - deepspeed_elastic.py
  - early_stop.py
  - lisa.py
  - mapping.py
  - perf_log.py
- - __init__.py
  - app.py
  - deploy.py
  - eval.py
  - export.py
  - infer.py
  - main.py
  - merge_lora.py
  - pt.py
  - sft.py
  - utils.py
  - web_ui.py
- - fsdp2.json
  - zero0.json
  - zero1.json
  - zero2.json
  - zero2_offload.json
  - zero3.json
  - zero3_offload.json
- - __init__.py
  - dispatcher.py
  - shard.py
- - __init__.py
  - dataset_meta.py
  - dataset_syntax.py
  - indexed_dataset.py
  - loader.py
  - media.py
  - packing.py
  - register.py
  - utils.py
- - __init__.py
  - constant.py
  - hub.py
- - __init__.py
  - base.py
  - grpo_vllm_engine.py
  - infer_client.py
  - infer_engine.py
  - lmdeploy_engine.py
  - patch.py
  - protocol.py
  - sglang_engine.py
  - transformers_engine.py
  - utils.py
  - vllm_engine.py
- - __init__.py
  - base.py
  - causal_lm.py
  - embedding.py
  - mapping.py
  - reranker.py
- - __init__.py
  - agent.py
  - base.py
  - mapping.py
  - other.py
  - utils.py
- - __init__.py
  - convert.py
  - init.py
- - __init__.py
  - acc.py
  - base.py
  - embedding.py
  - mapping.py
  - nlg.py
  - reranker.py
  - utils.py
- - __init__.py
  - constant.py
  - model_arch.py
  - model_meta.py
  - npu_patcher.py
  - patcher.py
  - register.py
  - utils.py
- - __init__.py
  - base.py
  - lorap.py
  - mapping.py
  - multimodal.py
  - muon.py
  - muonclip.py
- - __init__.py
  - base.py
  - utils.py
- - __init__.py
  - arguments.py
  - base.py
  - resource_manager.py
- - __init__.py
  - orm.py
  - prm.py
  - rm_plugin.py
- - __init__.py
  - args_mixin.py
  - arguments.py
  - cpo_trainer.py
  - dpo_trainer.py
  - gkd_trainer.py
  - grpo_trainer.py
  - kto_trainer.py
  - orpo_trainer.py
  - ppo_trainer.py
  - reward_trainer.py
  - rlhf_mixin.py
  - rollout_mixin.py
  - utils.py
  - vllm_client.py
- - __init__.py
  - gym_env.py
  - multi_turn.py
- - __init__.py
  - ulysses.py
  - utils.py
  - zigzag_ring_attn.py
- - __init__.py
  - base.py
  - constant.py
  - grounding.py
  - register.py
  - template_inputs.py
  - template_meta.py
  - utils.py
  - vision_utils.py
- - __init__.py
  - arguments.py
  - embedding_trainer.py
  - mixin.py
  - patcher.py
  - reranker_trainer.py
  - seq2seq_trainer.py
  - trainer.py
  - trainer_factory.py
  - utils.py
- - __init__.py
  - base.py
  - dummy.py
  - ia3.py
  - lora_llm.py
  - mapping.py
- - __init__.py
  - adapter.py
  - base.py
  - llamapro.py
  - lora.py
  - lora_layers.py
  - mapping.py
  - neftune.py
  - part.py
  - peft.py
  - prompt.py
  - reft.py
  - restuning.py
  - restuning_components.py
  - side.py
  - utils.py
- - __init__.py
  - app.py
  - base.py
- - __init__.py
  - constants.py
  - env.py
  - hf_config.py
  - hub_utils.py
  - import_utils.py
  - io_utils.py
  - logger.py
  - np_utils.py
  - processor_utils.py
  - shutdown_manager.py
  - tb_utils.py
  - transformers_utils.py
  - utils.py
- __init__.py
- version.py
- - test_app.py
- - test_dataset.py
  - test_logprobs.py
- - test_eval.py
- - test_quant.py
- - test_arch.py
  - test_dataset.py
  - test_model.py
  - test_stream.py
  - test_template.py
- - __init__.py
  - test_check_model.py
- - test_agent.py
  - test_infer.py
  - test_logprobs.py
  - test_main.py
  - test_max_memory.py
  - test_mllm.py
  - test_sglang.py
- - __init__.py
  - test_custom.py
  - test_dataset.py
  - test_ollama_export.py
  - test_run.py
  - test_template.py
- - test_embedding.py
  - test_export.py
  - test_gkd.py
  - test_grpo.py
  - test_kto.py
  - test_lora.py
  - test_rlhf.py
- - test_flash_attn.py
  - test_llm.py
  - test_mllm.py
- - test_client.py
- - test_cls.py
  - test_lmdeploy_vlm.py
  - test_padding_side.py
  - test_rlhf_loss.py
- - test_channel.py
  - test_cls.py
  - test_embedding.py
  - test_export_cached_dataset.py
  - test_freeze.py
  - test_gkd.py
  - test_grounding.py
  - test_grpo.py
  - test_kto.py
  - test_liger.py
  - test_multilabel.py
  - test_packing.py
  - test_ppo.py
  - test_pt.py
  - test_resume_from_checkpoint.py
  - test_rlhf.py
  - test_sample.py
  - test_sft.py
  - test_train_eval.py
  - test_vit_lr.py
  - test_vllm_importance_sampling_basic.py
- - __init__.py
  - test_extra_state_dict.py
  - test_merged_linear.py
  - test_neft.py
  - test_peft.py
  - test_scetuning.py
  - test_swift_base.py
  - test_swift_device_map.py
  - test_swift_restuning.py
- - __init__.py
  - test_async_rewards.py
  - test_file_utils.py
  - test_io_utils.py
  - test_rewards.py
  - test_split_str_parts_by.py
  - test_torch_utils.py
- __init__.py
- model_tag.py
- run.py
- run_config.yaml
LICENSE
Makefile
MANIFEST.in
README.md
requirements.txt
setup.cfg
setup.py

# Use via CDN

jsDelivr

jsDelivr serves any public GitHub repository as a CDN with zero setup. Pick a version and a file to get a ready-to-paste link and snippet.

Version

File

Link

Example

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

[ 📄[**Paper**](https://arxiv.org/abs/2607.27113) | 🚀[**Model**](https://www.modelscope.cn/models/EricTanh/VeritasPP) ] | 🚀[**Model (HF)**](https://huggingface.co/EricTan7/VeritasPP) ]

**Veritas++** is an interpretable multimodal large language model (MLLM) for
generalizable AI-generated image (AIGI) detection. It treats reliable visual
perception as the foundation of authenticity reasoning: the model first learns
to perceive fine-grained details, semantic anomalies, and pixel-level
discrepancies, and then internalizes these abilities through
**Value-aware On-Policy Distillation (VaOPD)**.

## Highlights

- **Perception-oriented Learning (PoRL)** strengthens fine-grained visual,
  semantic-anomaly, and pixel-level perception with verifiable rewards.
- **Value-aware On-Policy Distillation (VaOPD)** internalizes
  perception-aware reasoning through a privileged EMA self-teacher.
- Veritas++ achieves strong generalization on standard OOD, in-the-wild, and
  emerging AIGI detection benchmarks.

## Method

<p align="center">
  <img src="figs/method_overview2.png" alt="Overview of the three-stage Veritas++ training pipeline" width="100%">
</p>
The pipeline contains a high-quality cold start, PoRL-based perception
learning, and VaOPD-based perception-enhanced reasoning. The model is built on
Qwen3-VL-8B.

### Value-aware On-Policy Distillation

VaOPD scores student-generated trajectories with a privileged EMA self-teacher
and adapts the distillation signal at three levels: it prioritizes incorrect
trajectories, emphasizes informative tokens, and adjusts the distillation
direction for corrective signals.

## Installation

```bash
conda create -n veritaspp python=3.10 -y
conda activate veritaspp

pip install -e .
```

## 🔎 Inference on a single image

Download **Veritas++** from [ModelScope](https://www.modelscope.cn/models/EricTanh/VeritasPP) or [Huggingface](https://huggingface.co/EricTan7/VeritasPP) 🔥🔥🔥. We recommend using `vLLM` for model deployment:

```bash
sh self_scripts/deploy/deploy_model.sh /path/to/Veritas++
```

Set the checkpoint path in `vllm_server_config` inside
`self_scripts/infer/infer_vllm_single.py`, then run:

```bash
python self_scripts/infer/infer_vllm_single.py \
  --model model \
  --image_path /path/to/image.jpg
```

## Citation

If you find this work useful, please cite:

```bibtex
@article{tan2026veritaspp,
  title   = {Veritas++: Value-aware On-Policy Distillation for
             Perception-Enhanced AIGI Detection},
  author  = {Tan, Hao and Lan, Jun and Tan, Zichang and Liu, Ajian and
             Yu, Zijian and Song, Chuanbiao and Zhu, Huijia and
             Wang, Weiqiang and Wan, Jun and Lei, Zhen},
  journal={arXiv preprint arXiv:2607.27113},
  year={2026}
}
```

## License

This repository is released under the
[Apache License 2.0](LICENSE).

## Acknowledgements

This project is built upon
[ms-swift](https://github.com/modelscope/ms-swift) and
[Qwen3-VL](https://github.com/QwenLM/Qwen3-VL). We thank their authors and the
open-source community for their contributions.

VeritasPP

Owner Repos

File Explorer

# Use via CDN

Was this content helpful?

Recent Feedback

VeritasPP

Owner Repos

File Explorer

# Use via CDN

Was this content helpful?

Recent Feedback

Download README