Multi-task Dense Prediction on Taskonomy (test) benchmark leaderboard · SOTA2 Research

Multi-task Dense Prediction on Taskonomy (test)

0.016DE (L1)Finetuned

Evaluation Results

Method	Links
Finetuned2026.03		0.016	0.016	0.101	0.171	0.162	0.082	0.217	0.71	0.136	0.17	0.144	—
SVD2026.03		98	94.4	103.7	123.4	146.9	117.8	101.5	122.2	143.4	146.7	96.9	83
DARE-TIES2026.03		98.1	97.7	103.3	104.6	106.4	104.5	100.1	103.7	103.9	104.6	100.6	76.9
EMR-Merging2026.03		98.2	98.2	102.4	103	104.8	103.2	100.1	103.4	105.4	108.1	98.5	77
FR-Merging2026.03		99.7	95.6	101.5	125.8	150.6	117	101.7	122.7	146.6	149.7	97.3	83.4
RobustMerge2026.03		100	101	102.2	100.6	100.5	100.6	100	100.2	99.7	99.1	99.8	89.9
TA2026.03		100.1	99.8	101.7	104.6	106.4	102.2	100.2	103.6	106.2	107.4	98.8	77.2
KnOTS-TIES2026.03		100.1	99.6	101.3	102.8	105.2	102.4	100.1	102.7	104.9	104.8	97.9	76.6
NSC2026.03		100.1	101	102.2	100.8	100.8	100.7	100	100.4	100	99.6	99.6	92
TIES2026.03		102.5	102	100.6	103.7	105.4	99.5	100.1	102.4	105.2	104.3	99.5	77.3
KnOTS-DARE-TIES2026.03		103.9	102.8	97.7	107.8	107.3	96.1	100.2	102.4	103.1	101.7	100.2	76.8
LoRA-LEGO2026.03		103.9	104.1	96.7	106.7	108.7	99.1	100.2	103.6	110.4	102.2	104.5	76.4
Linear2026.03		108.3	105.9	95.9	120.5	122.1	117.1	100.4	98.1	126.4	75	104.9	74.3