SOTA Multi-task reinforcement learning benchmarks and papers with code

Benchmarks

Dataset Name	SOTA Method	Metric
Meta-World MT50 V2	TOPPO	Overall Success Rate90.9	16	21d ago
Meta-World MT10 V2		Success Rate99.5	15	21d ago
Meta-World MT10 v1 (Fixed)	MT-MH-SAC	Success Rate88	12	3mo ago
Meta-World MT50 V1 (final-checkpoint)	TOPPO	Success Rate (IQM)79.3	11	21d ago
Metaworld MT10 v2 (train/eval)	FAMO	Time4.2	11	3mo ago
Meta-World MT10 v1 (train test)	Nash-MTL	Average Success91	9	3mo ago
Meta-World MT50 (MT50-rand) V2 (Near-optimal)	MTDIFF-P-ONEHOT	Avg Success Rate61.32	8	3mo ago
Meta-World MT50 v1 (Fixed)	Soft Modularization (Deep)	Success Rate60	8	3mo ago
DMLab-30 (test)	PopArt-IMPALA	Mean Capped Human Score72.8	8	3mo ago
ALFWorld	Complementary RL	Validation Score82	6	2mo ago
Minihack Room	Complementary RL	Success Rate78	6	2mo ago
Meta-World MT50-rand V2 (Sub-optimal)	MTDIFF-P-ONEHOT	Average Success Rate48.94	6	3mo ago
Meta-World MT50 Conditioned v1	Soft Modularization (Deep)	Average Success Rate61	6	3mo ago
Meta-World MT10 Conditioned v1	Soft Modularization (Shallow)	Average Success Rate71.8	6	3mo ago
Meta-World MT10 5 x 10^6 steps	GIRL	IQM79	5	1mo ago
DIBS Multi-environment Suite (train)	MAML	Relative tasks solved ratio (vs DIBS)10.83	4	1d ago
DMLab-30 Multi-task Standard	LASER: Shared Experience Replay	Mean-Capped Human-Normalized Score81.7	4	3mo ago
Atari-57 unclipped	PopArt-IMPALA	Median Human Normalised Score10,700	4	3mo ago
Atari-57 clipped	PopArt-IMPALA	Median Human Normalized Score110.7	4	3mo ago
DIBS Multi-environment Suite Unseen tasks	MAML	Relative Solved Ratio (vs DIBS)14.04	3	1d ago
DmLab-30 (train)	PopArt-IMPALA	Mean capped human normalised score0.735	2	3mo ago
Meta-World	-	-	0	3mo ago

Showing 22 of 22 rows