World Labs' Atlas Tops Rapidata Camera-Movement Benchmark Across 513K Human Votes
theworldlabs · x · 2026-11-10
Rapidata's Camera Movement Benchmark ranks world and video generation models on whether they can execute plain-language camera instructions (move forward, tilt down, roll clockwise) starting from a single still image.
- Compares 19 world/video models on a fixed set of 600 prompts (30 camera instructions mixed across 22 spaces)
- Built from 513,079 human votes across 102,595 pairwise matches, fitted with a Bradley–Terry model on an Elo scale stretched by ln10≈2.3 to highlight capability gaps
- Annotator pre-filtering via gold-standard questions and majority voting over ≥5 responses per match ensure rating quality
World Labs' Atlas ranks #1. All data and judgments are public, and the leaderboard updates continuously.
Related event: World Labs' Atlas Tops Rapidata Camera Movement Benchmark(4 posts)→
More from Models
- Why did Anthropic ship 5.2 instead of jumping to 5.5? Two theories — haider1 · 2026-11-10
- Scientist: Anthropic's refusals make automated AI research loops nearly unusable — omarsar0 · 2026-11-10
- Thinking Machines' Inkling 2 Small lands on OpenCode free, ranks #9 with 1.4T tokens — thinkymachines · 2026-11-10
- Mythos 5.2 hits 41% on Remote Labor Index, up from Opus 4.8's 8% — peterwildeford · 2026-11-10
- Frontier model audits 20k-commit FrankenSim project brutally — and the author loves it — doodlestein · 2026-11-10
- Matt Shumer mocks local models as 'basically useless,' says buy Claude Ultra instead — mattshumer_ · 2026-11-10