ResearchBenchmarksVision-Language Conversation on LLaVA-BenchVision-Language Conversation on LLaVA-BenchFollow106.8Overall AccuracyICLA58.75271.22683.796.174Feb 28, 2026Evaluation resultsMethodMethodLinksOverall AccuracyICLABackbone=Qwen3-VL-8BBackbone=Qwen3-VL-8B2026.02106.8VCDBackbone=Qwen3-VL-8BBackbone=Qwen3-VL-8B2026.02106.7VDDBackbone=Qwen3-VL-8BBackbone=Qwen3-VL-8B2026.02106.3DAMOBackbone=Qwen3-VL-8BBackbone=Qwen3-VL-8B2026.02102.8DoLABackbone=Qwen3-VL-8BBackbone=Qwen3-VL-8B2026.0299.8VanillaBackbone=Qwen3-VL-8BBackbone=Qwen3-VL-8B2026.0298.8DeCoBackbone=Qwen3-VL-8BBackbone=Qwen3-VL-8B2026.0260.6