GeoVLM-R1

Disclaimer

THIS MODEL IS PROVIDED “AS IS,” WITHOUT WARRANTIES OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, ACCURACY, OR NON-INFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR THEIR AFFILIATES BE LIABLE FOR ANY DAMAGES ARISING FROM ACCESS TO OR USE OF THIS MODEL.

These models are released without a license. They are not open-source software.

Model weights for GeoVLM-R1: Reinforcement Fine-Tuning for Improved Remote Sensing Reasoning. Each folder is a separate Qwen2.5-VL-3B checkpoint trained for one Earth-observation task.

Model	Task	Benchmarks
GeoVLM_R1_Classification	Scene classification	AID, UCMerced, WHU-RS19, BigEarthNet, fMoW
GeoVLM_R1_VQA	Visual question answering	RSVQA-LR, RSVQA-HRBEN
GeoVLM_R1_Image_caption	Image captioning	NWPU-Captions, RSICD, RSITMD
GeoVLM_R1_Region_Captioning	Region captioning	GeoChat region captioning, NWPU VHR-10
GeoVLM_R1_grounding	Visual grounding (grounding description)	NWPU VHR-10 grounding
GeoVLM_R1_Refer	Referred object detection	GeoChat referring detection, NWPU VHR-10
GeoVLM_R1_CD_Caption	Change-detection captioning	LEVIR-CC, Dubai-CC, MUDS, SYSU-CC
GeoVLM_R1_xBD_object_Detection	xBD object detection	xBD object detection
GeoVLM_R1_xBD_refer_object_detection	xBD referred object detection	xBD referred object detection

See each folder's README.md for usage. Code for training, inference and evaluation is in the GeoVLM-R1 Toolkit.

Citation

@article{fiaz2025geovlmr1,
  title   = {GeoVLM-R1: Reinforcement Fine-Tuning for Improved Remote Sensing Reasoning},
  author  = {Fiaz, Mustansar and Debary, Hiyam and Fraccaro, Paolo and Paudel, Danda and
             Van Gool, Luc and Khan, Fahad and Khan, Salman},
  journal = {arXiv preprint arXiv:2509.25026},
  year    = {2025}
}

Contact

Mustansar Fiaz mustansar.fiaz@ibm.com