List your company

Image Captioning on COCO 2017 (val)

115.44CIDErLLaVA-v1.5-13B + Visual Lazy Attention
71.302482.761294.22105.6788Aug 2, 2023Jan 1, 2024Jun 2, 2024Nov 1, 2024Apr 3, 2025Sep 2, 2025Feb 2, 2026

Evaluation results

MethodLinks
LLaVA-v1.5-13B + Visual Lazy AttentionBase Model=LLaVA-v1.5-13B, Optimization Method=Visual Lazy Attention2026.02115.44——
LLaVA-v1.5-7BBase Model=LLaVA-v1.5-7B, Optimization Method=None2026.02110.43——
Flamingo-9BShots=322023.08106.3——
LLaVA-NEXT-13B + Visual Lazy AttentionBase Model=LLaVA-NEXT-13B, Optimization Method=Visual Lazy Attention2026.02101.91——
OpenFlamingo-9BShots=322023.0899.5——
OpenFlamingo-4B (Instruct)Shots=322023.0899.2——
Flamingo-3BShots=322023.0899——
OpenFlamingo-4BShots=322023.0895.1——
OpenFlamingo-3B (Instruct)Shots=322023.0894.8——
Flamingo-9BShots=42023.0893.1——
OpenFlamingo-3BShots=322023.0893——
OpenFlamingo-9BShots=42023.0889——
OpenFlamingo-4B (Instruct)Shots=42023.0885.8——
Flamingo-3BShots=42023.0885——
OpenFlamingo-3B (Instruct)Shots=42023.0882.7——
OpenFlamingo-4BShots=42023.0881.8——
OpenFlamingo-4B (Instruct)Shots=02023.0881.2——
OpenFlamingo-9BShots=02023.0879.5——
Flamingo-9BShots=02023.0879.4——
OpenFlamingo-3BShots=42023.0877.3——
OpenFlamingo-4BShots=02023.0876.7——
OpenFlamingo-3BShots=02023.0874.9——
OpenFlamingo-3B (Instruct)Shots=02023.0874.4——
Flamingo-3BShots=02023.0873——
Baseline ResNet+LSTMBackbone=ResNet, Decoder=LSTM2025.12—32.46—
Baseline ResNet+Transf.Backbone=ResNet, Decoder=Transformer2025.12—23.36—
C10C-RESNETLSTMDecoder Type=GRU(+Attn), Epochs=32025.12—9.14—
C10C-ResNetTransformerDecoder Type=GRU (feat-init), Epochs=32025.12—6.37—
C1C-RESNETLSTMDecoder Type=LSTM, Epochs=32025.12—10.5—
C5C-RESNETLSTMDecoder Type=LSTM, Epochs=32025.12—11.92—
C5C-ResNetTransformerDecoder Type=Transformer, Epochs=32025.12—8.62—
FROMAGE2023.05—10.2328.73
GILL2023.05—10.5925.29