Paper archive · v0.1.0
FourRooms
Partial-observation navigation through bottlenecks.
- Distribution
- Core16 paper
- Primary metric
- reported_score
- Direction
- Higher is better
- Record date
- 01 Jul 2026
Reported scores
Core16 paper profile · 128 episodes / run
1Claude Opus 4.7Claude Code0.701
2GPT-5.5Codex0.664
3MiniMax-M3Claude Code0.403
4DeepSeek-V4-ProClaude Code0.079
—BaselineUniform0.042