List your company

Video Captioning on ActivityNet-Captions (test)

12.31METEORMV-GPT
7.44288.70649.9711.2336Jan 20, 2022Oct 12, 2022Jul 4, 2023Mar 25, 2024Dec 15, 2024Sep 6, 2025May 29, 2026

Evaluation results

MethodLinks
MV-GPTAction Proposals=Ground-truth2022.0112.316.84——
OracleVLM=Qwen2.5-VL-3B, k=82026.0511.644.5446.5822.95
OracleVLM=Qwen2.5-VL-3B, k=42026.0511.534.5145.2322.67
OracleVLM=Qwen2.5-VL-3B, k=22026.0511.254.3441.6722.05
HACAAction Proposals=Ground-truth2022.0111.162.71——
PEEKVLM=Qwen2.5-VL-3B, k=82026.0511.134.1842.4222.12
UniformVLM=Qwen2.5-VL-3B, k=82026.0511.124.1242.3322.11
MaxInfoVLM=Qwen2.5-VL-3B, k=82026.0511.084.0941.9122.08
RandomVLM=Qwen2.5-VL-3B, k=82026.0511.074.1141.9122.05
CSTAVLM=Qwen2.5-VL-3B, k=82026.0511.074.0942.0622.02
PEEKVLM=Qwen2.5-VL-3B, k=42026.0510.934.0640.5521.77
BMTAction Proposals=Ground-truth2022.0110.91.99——
Bi-SSTAction Proposals=Ground-truth2022.0110.89———
MaxInfoVLM=Qwen2.5-VL-3B, k=42026.0510.893.9339.4721.59
UniformVLM=Qwen2.5-VL-3B, k=42026.0510.88439.5121.6
OracleVLM=Qwen2.5-VL-3B, k=12026.0510.844.0437.3121.25
RandomVLM=Qwen2.5-VL-3B, k=42026.0510.843.9639.5421.59
CSTAVLM=Qwen2.5-VL-3B, k=42026.0510.793.8839.1621.46
PEEKVLM=Qwen2.5-VL-3B, k=22026.0510.613.837.0221.06
UniformVLM=Qwen2.5-VL-3B, k=22026.0510.473.7335.3620.81
MaxInfoVLM=Qwen2.5-VL-3B, k=22026.0510.413.6535.120.71
RandomVLM=Qwen2.5-VL-3B, k=22026.0510.393.6734.9920.71
CSTAVLM=Qwen2.5-VL-3B, k=22026.0510.293.634.5320.6
PEEKVLM=Qwen2.5-VL-3B, k=12026.0510.133.4532.3920.21
OracleVLM=Qwen3.5-4B, k=82026.0510.073.2640.320.37
OracleVLM=Qwen3.5-4B, k=42026.059.963.1240.1920.3
OracleVLM=Qwen3.5-4B, k=22026.059.763.0840.6420.35
UniformVLM=Qwen2.5-VL-3B, k=12026.059.723.1930.0519.58
OracleVLM=Qwen2.5-VL-7B, k=22026.059.642.6440.121.12
OracleVLM=SmolVLM2-2.2B, k=82026.059.543.3139.0520.63
RandomVLM=Qwen2.5-VL-3B, k=12026.059.543.1729.0319.35
CSTAVLM=Qwen2.5-VL-3B, k=12026.059.483.0928.5619.21
UniformVLM=Qwen3.5-4B, k=82026.059.372.7635.4219.3
PEEKVLM=Qwen3.5-4B, k=82026.059.372.835.3119.29
OracleVLM=SmolVLM2-2.2B, k=42026.059.332.8739.4120.79
CSTAVLM=Qwen3.5-4B, k=82026.059.332.7435.3919.26
DVCAction Proposals=Ground-truth2022.019.311.71——
RandomVLM=Qwen3.5-4B, k=82026.059.32.6934.9319.25
MaxInfoVLM=Qwen3.5-4B, k=82026.059.262.6835.2219.21
OracleVLM=Qwen2.5-VL-7B, k=12026.059.242.636.619.59
PEEKVLM=Qwen3.5-4B, k=42026.059.22.7334.9319.22
OracleVLM=Qwen2.5-VL-7B, k=42026.059.182.2738.0520.84
OracleVLM=SmolVLM2-2.2B, k=22026.059.142.6638.1920.5
RandomVLM=Qwen3.5-4B, k=42026.059.142.6434.7219.09
UniformVLM=Qwen3.5-4B, k=42026.059.122.6334.6619.08
CSTAVLM=Qwen3.5-4B, k=42026.059.12.6634.5419.02
MaxInfoVLM=Qwen3.5-4B, k=42026.059.082.634.4418.97
OracleVLM=Qwen3.5-4B, k=12026.059.072.5638.4419.35
MaxInfoVLM=Qwen2.5-VL-3B, k=12026.059.052.8226.4718.46
PEEKVLM=Qwen2.5-VL-7B, k=22026.059.052.3135.0420.06
PEEKVLM=SmolVLM2-2.2B, k=82026.059.022.9634.3319.7
OracleVLM=SmolVLM2-2.2B, k=12026.0592.6537.3620.32
UniformVLM=SmolVLM2-2.2B, k=82026.058.952.9433.8519.52
CSTAVLM=SmolVLM2-2.2B, k=82026.058.922.8833.6919.5
UniformVLM=Qwen2.5-VL-7B, k=22026.058.912.3334.4319.84
RandomVLM=SmolVLM2-2.2B, k=82026.058.92.9133.4719.49
DCEVAction Proposals=Ground-truth2022.018.881.6——
MaxInfoVLM=Qwen2.5-VL-7B, k=22026.058.882.2333.5619.73
PEEKVLM=Qwen3.5-4B, k=22026.058.842.6134.5118.94
MaxInfoVLM=SmolVLM2-2.2B, k=82026.058.812.7933.0619.31
RandomVLM=Qwen2.5-VL-7B, k=22026.058.82.233.6819.67
CSTAVLM=Qwen2.5-VL-7B, k=22026.058.782.2533.6319.63
MaxInfoVLM=Qwen2.5-VL-7B, k=42026.058.732.1233.9220.11
PEEKVLM=SmolVLM2-2.2B, k=42026.058.722.4833.4519.62
UniformVLM=Qwen3.5-4B, k=22026.058.722.5533.3518.67
RandomVLM=Qwen3.5-4B, k=22026.058.72.532.9318.67
MaxInfoVLM=Qwen3.5-4B, k=22026.058.72.532.9718.66
UniformVLM=Qwen2.5-VL-7B, k=42026.058.72.0633.5419.96
CSTAVLM=Qwen3.5-4B, k=22026.058.692.533.4718.59
RandomVLM=Qwen2.5-VL-7B, k=42026.058.692.0733.5520.01
PEEKVLM=Qwen2.5-VL-7B, k=42026.058.672.0333.1519.92
CSTAVLM=Qwen2.5-VL-7B, k=42026.058.622.0333.0219.89
UniformVLM=SmolVLM2-2.2B, k=42026.058.592.4332.7619.39
RandomVLM=SmolVLM2-2.2B, k=42026.058.582.432.3519.38
CSTAVLM=SmolVLM2-2.2B, k=42026.058.582.4132.6219.39
OracleVLM=Qwen2.5-VL-7B, k=82026.058.581.9233.8819.69
PEEKVLM=Qwen2.5-VL-7B, k=12026.058.582.2631.5418.54
PEEKVLM=SmolVLM2-2.2B, k=22026.058.542.3232.9819.39
MaxInfoVLM=SmolVLM2-2.2B, k=42026.058.472.3231.9119.23
PEEKVLM=SmolVLM2-2.2B, k=12026.058.372.2331.5319.12
UniformVLM=SmolVLM2-2.2B, k=22026.058.312.1931.2319
RandomVLM=SmolVLM2-2.2B, k=22026.058.252.1731.0618.88
UniformVLM=Qwen2.5-VL-7B, k=12026.058.18228.5417.88
CSTAVLM=SmolVLM2-2.2B, k=22026.058.172.1330.3818.72
PEEKVLM=Qwen3.5-4B, k=12026.058.132.0431.7317.71
UniformVLM=SmolVLM2-2.2B, k=12026.058.082.129.7918.55
RandomVLM=Qwen2.5-VL-7B, k=12026.058.082.0628.4417.79
PEEKVLM=Qwen2.5-VL-7B, k=82026.058.061.7130.0118.75
MaxInfoVLM=SmolVLM2-2.2B, k=22026.058.052.0729.4918.6
CSTAVLM=Qwen2.5-VL-7B, k=12026.058.032.0228.0217.71
UniformVLM=Qwen2.5-VL-7B, k=82026.058.011.7129.418.7
RandomVLM=Qwen2.5-VL-7B, k=82026.058.011.6829.5818.72
CSTAVLM=Qwen2.5-VL-7B, k=82026.057.991.6929.7718.75
MaxInfoVLM=Qwen2.5-VL-7B, k=82026.057.971.6729.4318.72
RandomVLM=SmolVLM2-2.2B, k=12026.057.852.0328.418.18
CSTAVLM=SmolVLM2-2.2B, k=12026.057.851.9928.3718.2
UniformVLM=Qwen3.5-4B, k=12026.057.851.9329.5517.28
MaxInfoVLM=Qwen2.5-VL-7B, k=12026.057.761.9926.4417.2
RandomVLM=Qwen3.5-4B, k=12026.057.691.9428.9417.06
CSTAVLM=Qwen3.5-4B, k=12026.057.631.928.5116.95