Phoenix-VL: localized multimodal training and evaluation resources

What changed

Adapted a pretrained multimodal model using curated regional/general replay data, long-context continued training and local multimodal/alignment data, introducing Singapore-focused evaluation and behavior frameworks.

Scientific scope

Adapted a pretrained multimodal model using curated regional/general replay data, long-context continued training and local multimodal/alignment data, introducing Singapore-focused evaluation and behavior frameworks. Inherits MistralMedium and continual pretraining/ODPO. Localized empirical performance does not prove no capability loss, regulatory compliance or correct legal advice.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Inherits MistralMedium and continual pretraining/ODPO. Localized empirical performance does not prove no capability loss, regulatory compliance or correct legal advice.

Contributors

• Source-listed authors and technical contributors · SCIENCE: All 27 named technical contributors are listed. Two executive sponsors receive separate support credit, not scientific-person associations.

• Phoenix-VL 1.5 Medium Technical Report
• Mistral research publications index

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction