Procedural Graph Extraction on PAGED (test) benchmark leaderboard · SOTA2 Research

Procedural Graph Extraction on PAGED (test)

78.6Action F1text2flow

Evaluation Results

Method	Links
text2flow2026.01		78.6	64.2	79.1	0.452	83.9	0.345	64.8	47.9	34.2	68.2
text2flow2026.01		78.2	68.8	85.9	0.485	83.1	0.356	69.2	48.4	35.7	71
text2flow2026.01		78.1	72.3	81.2	0.46	83.2	0.282	67.7	47.6	31.7	65.3
Few-Shot2026.01		77.8	71.1	77.6	0.461	76.5	0.192	61.3	42.3	28	61.1
Self-Refine2026.01		77.8	71.7	79	0.455	80.4	0.249	63.9	45	28.4	61.9
text2flow2026.01		77.6	64.4	84.3	0.411	79.2	0.238	58.1	42.9	27.6	68.4
Few-Shot2026.01		77.4	62.1	77.1	0.445	78.9	0.28	60.1	37.1	27.6	47.7
Self-Refine2026.01		77.4	61.2	73.4	0.434	78.4	0.239	64.2	47.5	28	63.8
Few-Shot2026.01		76.9	68.4	84.1	0.481	79.2	0.296	64.8	43.4	30.2	59.4
Self-Refine2026.01		76.9	69.1	83.8	0.483	79.6	0.299	67.3	47.6	30.6	67.3
Few-Shot2026.01		76.2	62.3	81.6	0.421	76.2	0.201	55.7	33.9	23.2	49.9
Self-Refine2026.01		76.2	65.3	84.6	0.421	79.2	0.201	58.7	33.9	25.2	49.9
Actor-Critic2026.01		75.4	66	70.9	0.335	81.5	0.256	63.3	45.5	29.1	55.6
Actor-Critic2026.01		75	51.8	71.4	0.388	73.4	0.237	58.4	34.4	24.6	55.5
text2flow2026.01		74.2	57.7	64.4	0.178	70.3	0.144	37	39.5	23.8	43.6
Actor-Critic2026.01		74.2	46.9	62.5	0.339	72.8	0.217	57.4	41.7	24.8	47.2
Self-Refine2026.01		70.8	56.8	56.2	0.172	70.2	0.152	34.7	36.8	20.4	39.9
Actor-Critic2026.01		70.6	61.7	76	0.441	72.6	0.225	65.1	42.5	23.7	58.1
Actor-Critic2026.01		67	50.4	55.6	0.165	67.3	0.15	32.7	32.1	18.2	37.1
Few-Shot2026.01		64.9	40.3	53.4	0.151	70.1	0.141	36.6	31.5	18.6	28.8
CIS2026.01		63.9	63.3	—	—	45.5	—	—	20.3	15.7	—
PET2026.01		43	8.5	6.9	—	49.3	—	—	16.4	2.6	—
BPMN-Gen2026.01		38.7	—	—	—	46.3	—	19.8	9.1	2.2	—
MT-BPMN2026.01		30.8	2.8	21.3	—	48.5	—	27.9	5.6	4.7	1.7
BRP2026.01		27.6	2.7	—	—	46.9	—	33.7	7.4	6.1	—