{"@context":"https://schema.org","@type":"CreativeWork","@id":"https://forgecascade.org/public/capsules/d3cbef5d-c8fd-4363-b72c-de5bc97da85d","identifier":"d3cbef5d-c8fd-4363-b72c-de5bc97da85d","url":"https://forgecascade.org/public/capsules/d3cbef5d-c8fd-4363-b72c-de5bc97da85d","name":"ATLAS: Functional Tokens for Visual Reasoning","text":"# ATLAS: Functional Tokens for Visual Reasoning\n\nSource-linked arXiv preprint reference. This capsule summarizes the paper at the abstract level and points to the primary source.\n\nAuthors: Ziyu Guo, Rain Liu, Xinyan Chen, Pheng-Ann Heng\nSource: https://arxiv.org/abs/2605.15198v1\n\n## What it covers\nThe paper proposes ATLAS, a visual-reasoning framework that represents visual operations as discrete functional tokens inside the normal token vocabulary. The design aims to combine aspects of agentic tool-like reasoning and latent reasoning while remaining compatible with standard next-token training.\n\n## Why it is useful\nThis is useful for teams following multimodal reasoning research because it gives a compact mechanism for representing intermediate visual operations without generating full visual states at every step.\n\n## Limits\nThe method and benchmark claims are from a single preprint. Treat this as a research reference rather than evidence of production reliability.\n\n## Sources\n- https://arxiv.org/abs/2605.15198v1","keywords":["arxiv","visual-reasoning","multimodal-ai","reinforcement-learning","public-reference"],"about":[],"citation":[],"isPartOf":{"@type":"Dataset","name":"Forge Cascade Knowledge Graph","url":"https://forgecascade.org"},"publisher":{"@type":"Organization","name":"Forge Cascade","url":"https://forgecascade.org"},"dateCreated":"2026-05-24T00:19:08.853118Z","dateModified":"2026-06-19T01:30:50.188141Z","isBasedOn":"https://arxiv.org/abs/2605.15198v1","additionalProperty":[{"@type":"PropertyValue","name":"trust_level","value":100},{"@type":"PropertyValue","name":"verification_status","value":"sources_verified"},{"@type":"PropertyValue","name":"provenance_status","value":"valid"},{"@type":"PropertyValue","name":"evidence_level","value":"primary_source"},{"@type":"PropertyValue","name":"content_hash","value":"38529da942c41f2c5bcec33fc5394c39f8bde0bc03954e9ca1d6fee880e839a6"}]}