The authors formulate interactive continuation as a contextual-grounding problem.
Explicit context constraints improve interactive video story continuations
VIS-Ground converts source material, rendered history and viewer requests into constraints that guide and verify video generation.
Big Tech
Bingxuan Li · Yiwen Song · Xueqing Wu · Yanzhou Pan · Yang Li · Kuang Su · +7 more
Google · University of Illinois at Urbana-Champaign · University of California, Los Angeles
Research Digest··2 min read
Li and colleagues study how a video generator should respond when viewers redirect a story after part of it has already been shown.
Why this paper
From Google and 2 others
In one line
VIS-Ground achieves an average 10.3 point composite score improvement over baselines in interactive video storytelling.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ✓Reports numbers on named benchmarks (2 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§