Paper Overview
Field: Machine Learning Authors: Tica Lin, Deepak Chandran, Gauri Jagatap, Chen Chen, Andrea Fanelli, David Gunawan, Josh Kimball Published: 2026-09-17 arXiv: 2609.20768
Summary
Generative agents are increasingly used to select and narrate video highlights, but they typically operate over unstructured or frame-level representations. Their output is consequently difficult for a viewer to verify and steer toward individual preferences. The authors present the semantic action graph, a lightweight domain schema that represents a sports match as performer, action, recipient, moment, and state nodes connected by role, temporal, and outcome edges.
The schema demonstrates three key properties:
1. Connected event sequences 2. A shared, closed vocabulary 3. Frame-addressable moments
These make it suitable to serve two consumers at once:
- An agentic pipeline that composes narrated highlights
- A visual interface through which viewers query and inspect the same structure
SportSAGE
The schema is instantiated in SportSAGE, a design exploration pairing a four-module highlights pipeline with a graph interface. The authors report feedback from 12 soccer fans: participants were satisfied with the quality of generated highlights and narration, and used the graph interface to search, navigate, and interpret match highlights.
These results provide early evidence that a small, human-readable schema can simultaneously support agent generation and human understanding.
--- *Auto-collected on 2026-09-19*