RSVLM-SFT: Instruction-Tuning Pairs for Remote Sensing Vision-Language Models

RSVLM-SFT is a remote sensing instruction-tuning dataset released by FelixKAI in 2026 for training the MF-RSVLM vision-language model. It contains image-text pairs for modality alignment and instruction following, although the specific record count is not disclosed in the metadata.

Use Cases

• Fine-tuning Vicuna-7B models using instruction-following prompts for satellite imagery
• Training VQA systems to identify specific features within remote sensing image-text pairs
• Performing modality alignment between CLIP vision encoders and language models using visual-language pairs

Strengths

• Licensed under Apache 2.0 for open research and commercial use
• Specifically formatted for two-stage VLM training workflows
• Optimized for integration with Vicuna-7B and CLIP architectures

Limitations

• Unknown total record count and storage size
• Lack of explicit documentation on the geographic diversity of the underlying remote sensing images

Provenance

Source

FelixKAI / MF-RSVLM research project (Arxiv: 2512.24022)

Freshness

Last updated on January 9, 2026.