Show HN: Valen – Open multimodal decision models for text, images and video (valen-team.github.io)

A new project called Valen has been unveiled, introducing open multimodal decision models capable of processing text, images, and video. Designed for real-world applications, particularly in automating tasks like ticket intent routing, Valen aims to optimize the balance between automation and accuracy. In its initial evaluation, Valen automated 11.5% of inquiries with an observed error rate of just 0.7%, showcasing impressive performance through a rigorous validation process that included 1,276 test questions across 77 candidate intents. This development is significant for the AI/ML community as it emphasizes the potential for deploying multimodal systems to enhance decision-making in complex environments. With validation metrics like Expected Calibration Error (ECE) showing substantial improvement (from 0.0819 to 0.0518), Valen highlights the importance of model refinement and calibration in achieving reliable outcomes. The open model allows developers to customize and fit the system for their workflows, paving the way for broader adoption and innovation in automating decision processes across various industries.