Multimodal Report Generation on TVIR-BENCH benchmark leaderboard · SOTA2 Research

Multimodal Report Generation on TVIR-BENCH

68.64Content Score (CS)TVIR-Agent

Evaluation Results

Method	Links
TVIR-Agent2026.06		68.64	71.98	69.2	68.16	80.2	84.61	62.55	70.13	73.39	77.35	71.64	73.61	72.62
TVIR-Agent2026.06		53.68	76.69	69.3	67.48	83	91.71	67.8	72.44	74.56	78.63	70.03	77.03	73.53
TVIR-Agent2026.06		51.2	81.09	69.88	72.22	76.2	87.17	77.8	74.49	76.75	77.58	70.12	78.76	74.44
Claude-4.5-Sonnet2026.06		47.53	79.32	69.37	70.52	84	90.24	63.85	61.47	53.43	67.49	70.15	67.3	68.72
Manus-1.62026.06		45.57	74.12	72.15	62.84	92.4	86.27	70.75	66.14	71.03	56.02	69.42	70.04	69.73
Perplexity Deep Research2026.06		44.6	81.03	67.6	70.64	80.9	73.62	62.9	59.02	63.41	8.35	68.95	53.46	61.2
Genspark Deep Research2026.06		35.27	83.71	69.28	70.64	84.6	92.87	70.7	63.58	59	40.28	68.7	65.29	66.99
Grok-4.1-Thinking DeepSearch2026.06		17.72	60.65	67.68	57.58	89.2	80.43	52.15	47.04	46.76	5.75	58.56	46.43	52.49
Gemini-3-Pro Deep Research2026.06		14.96	58.31	66.88	63.94	88.5	—	—	—	—	—	58.52	—	—