🥯 BAGEL-NHR-Edit-V2

🌐 NHR Website | 📜 NHR Paper on arXiv | 🤗 NHR-Edit Dataset (part1) | 🤗 NHR-Edit Dataset (part2) |

This repository hosts the model weights for BAGEL, fine-tuned on the NHR-Edit + NHR-Edit-part2 dataset. For installation, usage instructions, and further documentation, please visit the official BAGEL GitHub repository.

🛠️ Training Setup

We performed parameter-efficient adaptation on the generation expert’s attention and FFN projection layers using LoRA.

LoRA parameters:

r = 16
lora_alpha = 16
dropout = 0.05
bias = "none"
target_modules = [
  "v_proj_moe_gen",
  "k_proj_moe_gen",
  "mlp_moe_gen.down_proj",
  "mlp_moe_gen.gate_proj",
  "q_proj_moe_gen",
  "mlp_moe_gen.up_proj",
  "o_proj_moe_gen"
]

Metrics for GEdit-Bench-EN:

Model	GEdit-Bench-EN (SC) ↑	GEdit-Bench-EN (PQ) ↑	GEdit-Bench-EN (O) ↑
BAGEL-7B-MoT	7.610 ± 0.150	6.180 ± 0.150	6.530 ± 0.140
BAGEL-NHR-Edit-V2	7.800 ± 0.070	6.560 ± 0.080	6.800 ± 0.070

Scoring model: gpt-4.1-2025-04-14 (with default temperature)

Metrics for ImgEdit-Bench:

Model	Style	Extract	Remove	Background	Action	Adjust	Add	Replace	Compose	Overall ↑
BAGEL-7B-MoT	4.20 ± 0.05	1.59 ± 0.10	3.16 ± 0.10	3.29 ± 0.06	3.96 ± 0.17	3.51 ± 0.20	3.98 ± 0.02	3.54 ± 0.11	2.93 ± 0.26	3.30 ± 0.03
BAGEL-NHR-Edit-V2	4.28 ± 0.04	1.65 ± 0.07	3.12 ± 0.06	3.31 ± 0.02	3.81 ± 0.17	3.48 ± 0.12	4.19 ± 0.03	3.51 ± 0.06	2.99 ± 0.21	3.33 ± 0.02

Scoring model: gpt-4o-2024-11-20 (with temperature = 0.0)

License

BAGEL-NHR-Edit-V2 is licensed under the Apache 2.0 license. It is finetuned from ByteDance-Seed/BAGEL-7B-MoT, which is also licensed under Apache 2.0.

✍️ Citation

@article{Layer2025NoHumansRequired,
    arxivId = {2507.14119},
    author = {Maksim Kuprashevich and Grigorii Alekseenko and Irina Tolstykh and Georgii Fedorov and Bulat Suleimanov and Vladimir Dokholyan and Aleksandr Gordeev},
    title = {{NoHumansRequired: Autonomous High-Quality Image Editing Triplet Mining}},
    year = {2025},
    eprint = {2507.14119},
    archivePrefix = {arXiv},
    primaryClass = {cs.CV},
    url = {https://arxiv.org/abs/2507.14119},
    journal={arXiv preprint arXiv:2507.14119}
}