List your company
ResearchFormerly WizWand

Multimodal Reasoning on SEED-BENCH

78.1AccuracyC-RADIOv4-H
38.47648.76359.0569.337Aug 8, 2024Dec 1, 2024Mar 27, 2025Jul 21, 2025Nov 13, 2025Mar 9, 2026Jul 3, 2026
Updated 3mo ago

Evaluation results

MethodLinks
C-RADIOv4-HTokens per frame/tile=256, TTFT (ms)=468.2, LLM Backbone=9B Nemotron2026.0778.1
RADIO1DTokens per frame/tile=256, TTFT (ms)=452.8, LLM Backbone=9B Nemotron2026.0777.9
RADIO1DTokens per frame/tile=224, TTFT (ms)=410.2, LLM Backbone=9B Nemotron2026.0777.8
SigLIP2-SO400mTokens per frame/tile=256, TTFT (ms)=440.5, LLM Backbone=9B Nemotron2026.0777.7
RADIO1DTokens per frame/tile=128, TTFT (ms)=373.2, LLM Backbone=9B Nemotron2026.0777.6
RADIO1DTokens per frame/tile=192, TTFT (ms)=380.4, LLM Backbone=9B Nemotron2026.0777.6
CoLTSize=8B, Backbone=Qwen3-VL-8B2026.0677.5
SigLIP2-gTokens per frame/tile=256, TTFT (ms)=517.6, LLM Backbone=9B Nemotron2026.0777.4
LLaVA-OneVision-1.5Size=8B2026.0677.3
RADIO1DTokens per frame/tile=64, TTFT (ms)=346.1, LLM Backbone=9B Nemotron2026.0776.9
Qwen3-VL (Textual reasoning)Size=8B2026.0676.4
RADIO1DTokens per frame/tile=32, TTFT (ms)=335.1, LLM Backbone=9B Nemotron2026.0776.3
RADIO1DTokens per frame/tile=8, TTFT (ms)=329.7, LLM Backbone=9B Nemotron2026.0774.8
SIM-CoTSize=8B, Backbone=Qwen3-VL-8B2026.0672.4
LVRSize=7B2026.0671.2
RADIO1DTokens per frame/tile=1, TTFT (ms)=327.0, LLM Backbone=9B Nemotron2026.0770
InternVL2-8B + RPRP fine-tuning=Object Replacement Data2024.0869.9
InternVL2-8B-FTRP fine-tuning=None2024.0869.5
SoftCoTSize=8B, Backbone=Qwen3-VL-8B2026.0669.2
CODISize=8B, Backbone=Qwen3-VL-8B2026.0668.6
LaCoTSize=7B2026.0667.4
Qwen3-VL (Direct answer)Size=8B2026.0666.2
ICoTSize=8B, Backbone=Qwen3-VL-8B2026.0664.7
MGM-7BRP fine-tuning=None2024.0863.5
CCoTSize=8B, Backbone=Qwen3-VL-8B2026.0663.4
MGM-7B + RPRP fine-tuning=Object Replacement Data2024.0863.2
Multimodal CoTSize=8B, Backbone=Qwen3-VL-8B2026.0663.2
LLaVA-1.5-13B + MMFuserLLM=Vicuna-13B2024.1062
LLaVA-1.5-7B + RPRP fine-tuning=Object Replacement Data2024.0861.7
LLaVA-1.5-13BLLM=Vicuna-13B2024.1061.6
LLaVA-1.5-7B + MMFuserLLM=Vicuna-7B2024.1060.8
VanillaBackbone=LLaVA-1.5-7B, Visual Token Budget=5762026.0659.3
LLaVA-1.5-7BLLM=Vicuna-7B2024.1058.6
LLaVA-1.5-7BRP fine-tuning=None2024.0858.6
Qwen-VL-ChatLLM=Qwen-7B2024.1058.2
EvoCutBackbone=LLaVA-1.5-7B, Visual Token Budget=1922026.0657.3
FastVBackbone=LLaVA-1.5-7B, Visual Token Budget=1922026.0657.1
ApETBackbone=LLaVA-1.5-7B, Visual Token Budget=1922026.0656.8
VisionZipBackbone=LLaVA-1.5-7B, Visual Token Budget=1922026.0656.4
V2DropBackbone=LLaVA-1.5-7B, Visual Token Budget=1922026.0656.4
Qwen-VLLLM=Qwen-7B2024.1056.3
FastVBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.0655.9
SparseVLMBackbone=LLaVA-1.5-7B, Visual Token Budget=1922026.0655.8
EvoCutBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.0655.2
VisionZipBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.0654.9
PDropBackbone=LLaVA-1.5-7B, Visual Token Budget=1922026.0654.7
ApETBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.0654.7
V2DropBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.0653.8
InstructBLIPLLM=Vicuna-7B2024.1053.4
SparseVLMBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.0653.4
PDropBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.0653.3
EvoCutBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.0653.2
ApETBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.0652.4
VisionZipBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.0652.2
FastVBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.0651.9
V2DropBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.0651.4
SparseVLMBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.0651.1
BLIP-2LLM=Vicuna-13B2024.1046.4
PDropBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.0640