ReefNet/reefnet-minicpm-v-4.5-lora

ReefNet LoRA adapter for MiniCPM-V 4.5

A LoRA adapter for openbmb/MiniCPM-V-4_5 (revision c1a6986b), fine-tuned to identify the hard-coral genus at the centre of a reef photo crop (39 labels: 38 genera plus the family Fungiidae). This is the model reported in Table 5 of the ReefNet paper (NeurIPS 2026, Evaluations & Datasets track). Use of the adapter is also subject to the MiniCPM model licence of the base model.

test set	macro recall (%)
ReefNet (within-source, 39 labels)	63.36
Al-Wajh, Red Sea (cross-source, 13 genera)	48.09

Usage

MiniCPM-V uses remote code. Load the base model with AutoModel.from_pretrained(base, revision=rev, trust_remote_code=True, torch_dtype=torch.bfloat16), wrap it with PeftModel.from_pretrained(model, "ReefNet/reefnet-minicpm-v-4.5-lora"), then query it via model.chat(...) with the prompt below. The reference inference script is in the reefnet-mllm repository.

Training

LoRA fine-tuning with LLaMA-Factory @ 589da21d. The exact config is in train_config.yaml: rank 8 on all linear layers, AdamW, lr 5e-5, cosine schedule, warmup ratio 0.03, bf16, per-device batch 8, no gradient accumulation, image_max_pixels 1024x1024, no augmentation. Software: transformers 4.57.1, peft 0.18.1, torch 2.9.1. Model selection used the ReefNet validation split. Al-Wajh is never used for training, tuning or selection.

Prompts (exact strings, <image> marks the image position)

split	user prompt	target
train / val / ReefNet test	What is the genus in the center of the image <image>? Answer with one word	genus name
Al-Wajh test	<image>Identify the hard coral genus in this image. Respond with only the genus name.	genus name

Decoding: greedy (top_k=1), max_new_tokens=32. The output is normalised (strip, capitalise) and must match a genus exactly.

Metric

Macro recall (balanced accuracy). ReefNet: 39 labels, 23,043 test crops. Al-Wajh: averaged over the 13 genera present (4,624 crops), while the model keeps the full 39-way output. A prediction outside the 13 counts as an error. predictions/ holds the per-sample outputs behind the reported numbers. Recompute them with tools/recompute_macro_recall.py from the reefnet-mllm repository.

Data

ReefNet-1.0: 1024x1024 crops centred on each point annotation. Training uses the high-confidence benchmark split (445,985 train / 9,999 val).

Citation

@inproceedings{felemban2026reefnet,
  title     = {ReefNet: A Large-Scale Dataset and Benchmark for Fine-Grained Coral Reef Recognition},
  author    = {Felemban, Abdulwahab and Battach, Yahia and Khan, Faizan Farooq and Fu, Yuqian and Liu, Xuhui and
               Khattab, Yesmeen and Radwan, Yousef A. and Li, Xiang and Dunn, Natalie and B{\"a}hr, Susanne and
               Terraneo, Tullia Isotta and Marchese, Fabio and Beery, Sara and Jones, Burton H. and
               Benzoni, Francesca and Elhoseiny, Mohamed},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS) Evaluations and Datasets Track},
  year      = {2026}
}