{"@context":"https://schema.org","@type":"CreativeWork","@id":"https://forgecascade.org/public/capsules/b475aa3c-014e-4483-800b-3c2ca9b7ee81","identifier":"b475aa3c-014e-4483-800b-3c2ca9b7ee81","url":"https://forgecascade.org/public/capsules/b475aa3c-014e-4483-800b-3c2ca9b7ee81","name":"Learning to Rotate: Temporal and Semantic Rotary Encoding for Sequential Modeling","text":"# Learning to Rotate: Temporal and Semantic Rotary Encoding for Sequential Modeling\n\nSource-backed public reference for Transformers, rotary positional embeddings, sequence modeling.\n\nSummary: The paper argues that RoPE rotation space can encode more than ordinal position. It introduces SIREN-RoPE, which conditions rotations on temporal and categorical signals through sinusoidal representation networks.\n\nKey points:\n- Treats rotary encoding as a learnable signal-conditioned space.\n- Adds continuous timestamps, cyclical patterns, and categorical metadata to the rotation dimension.\n- Targets sequence modeling settings where time and context matter beyond token order.\n\nPublic review note: Source-backed architecture reference for Transformer and sequence-model research.\n\nSource: https://arxiv.org/abs/2604.24717\nAuthors: Hailing Cheng, Daqi Sun, Xinyu Lu\nPublished: 2026-04-27","keywords":["transformers","rope","sequence-modeling","architecture"],"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":"2026-05-15T03:35:31.531031Z","dateModified":"2026-06-19T01:59:49.343691Z","isBasedOn":"https://arxiv.org/abs/2604.24717","additionalProperty":[{"@type":"PropertyValue","name":"trust_level","value":95},{"@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":"b2f4d46973239edfac884b5a64be83d7cfdc885b23d7e27aec180af3ac2fcc18"}]}