DexHoldem Policy Bench - Scene-Preserving Success: leaderboard

Metric: Scene-preserving success rate (%; the primitive completes without disturbing other cards or chips, over 80 real-robot primitive trials (10 per pickup primitive, 5 per other primitive, 14 primitives) per policy, each policy trained as one multi-task policy on the 1,470 DexHoldem teleoperated demonstrations). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 9 models tracked.

Top models

#ModelScore
1π0.547.5
2π047.5
3RDT30
4DP (DINO)26.2
5DP-Transformer13.8
6RDT-small13.8
7ACT10
8BAKU6.2
9DP-UNet1.2

Interactive version: theaggregate.ai/benchmark?slug=dexholdem-policy-bench-scene-preserving-success · How It Works · Data refreshed daily, snapshot 2026-09-26.