Algorithmic Reasoning on CLRS-30 (test) benchmark leaderboard · SOTA2 Research

Algorithmic Reasoning on CLRS-30 (test)

96.01Kruskal MST AccuracyHint-ReLIC

Updated 2mo ago

Evaluation Results

Method
Hint-ReLIC2023.06	96.01	98.45	99.32	—	76.79	96.59	95.54	99	98.17	97.74	72.23	87.97	92.7	92.94	93.3	95.16	89.68	99.37	—	—	—	—	—	—	—	—	—	—	—
contrastive regularization2023.06	93.15	—	—	—	—	—	—	—	—	—	—	85.2	98.74	98.74	98.74	98.74	87.12	99.99	—	—	—	—	—	—	—	—	—	—	—
Proposed no-hint mode2023.06	92.89	94.55	98.91	34.52	66.38	94.58	95.25	99.95	98.18	96.97	42.16	85.7	90.67	90.67	90.67	90.67	85.29	99.98	—	—	—	—	—	—	—	—	—	—	—
No hint2023.06	92.28	81.97	95.62	33.94	57.63	84.29	93.26	99.89	97.62	95.01	40.8	85.33	77.29	81.32	71.6	68.5	93.21	99.24	—	—	—	—	—	—	—	—	—	—	—
EP2026.04	91.5	93.5	97.6	68.1	41.9	88.1	97.5	100	98	97.3	61.4	88.7	89.5	88.9	0.8	69.9	78	98.4	95.1	75.4	85.8	88.2	90.8	98.2	93.8	90.4	16.3	79.8	78.7
Baseline2023.06	91.18	88.93	93.75	39.71	38.53	87.27	96.67	99.64	88.12	93.41	46.51	87.64	75.28	79.87	70.53	32.12	74.6	97.78	—	—	—	—	—	—	—	—	—	—	—
CLRS2026.04	89.8	88.3	94	47.8	43.4	87.3	97.4	99.7	98.2	96.1	48.5	86.4	78.1	67.7	0.5	31	77.6	97.8	95.2	76.4	80.5	87.3	91.7	97.6	93.6	91	19.5	78.7	73.8
SPARK2026.04	86.2	97.9	98.2	83.7	77.3	99.7	96	70.6	97.8	99.2	32.9	88.7	97.5	98.1	10.7	75.6	87.6	99.1	98	85.6	87.6	87.4	87.3	97.1	96.4	94.9	14.9	77.3	96.1