Procedural Generalization by Planning with Self-Supervised World Models
Open paper
Benchmark
Model
Rank
Results
meta-learning-on-ml10
MZ
–
Meta-train success rate: 97.6%
Meta-test success rate (zero-shot): 26.5
meta-learning-on-ml10
MZ+Recon
–
Meta-train success rate: 97.8%
Meta-test success rate (zero-shot): 25
Rank counts only results with a code link.