Fotor Web-CogReasoner Accepted at ICLR 2026: Pure Pixel Visual Agent for Cross-Platform Workflows

GlobeNewswire·medium signalXBlueskyLinkedInCopy link

Fotor announced on April 2 that its joint research paper 'Web-CogReasoner: Towards Multimodal Knowledge-Induced Cognitive Reasoning for Web Agents' was accepted at ICLR 2026. The Fotor Agent moves beyond dependence on a webpage's underlying DOM/code by using 'pure pixel visual perception' — a multimodal reasoning approach that can seamlessly manage workflows across web, desktop, and mobile applications. This represents a significant step toward UI-agnostic computer use agents that don't require accessibility APIs or structured page representations, addressing a key limitation of current browser automation approaches.