List your company

Audio Captioning Evaluation on BRACE Main 1.0

67.63AudioCaps-Main HH ScoreCAF-Score
13.75827.74441.7355.716Mar 20, 2026

Evaluation results

MethodLinks
CAF-ScoreMetric Architecture=CLAP + LALM, Reference Protocol=Reference-free, Backbone LALM=Qwen3-Omni-Instruct, Backbone CLAP=M2D-CLAP2026.0367.6386.9767.9175.2870.6677.866.5271.2873.11
CLAIR-AReference Protocol=Reference-based2026.0366.1990.7967.7476.5156.2991.9465.6374.3975.36
CLAPScoreMetric Architecture=CLAP-ONLY, Reference Protocol=Reference-free, Backbone=M2D-CLAP2026.0365.4780.2263.9970.4867.6666.2162.8164.6967.35
S-CLAPScoreMetric Architecture=CLAP-ONLY, Reference Protocol=Reference-free, Backbone=M2D-CLAP2026.0365.4780.2263.9970.4870.0669.9463.8566.9168.59
CAF-ScoreMetric Architecture=CLAP + LALM, Reference Protocol=Reference-free, Backbone LALM=AudioFlamingo3-Base, Backbone CLAP=MGA-CLAP2026.0365.4771.0167.9168.8270.0670.9263.2666.9967.83
CLAPScoreMetric Architecture=CLAP-ONLY, Reference Protocol=Reference-free, Backbone=MS-CLAP2026.0364.0348.7655.0853.7167.0770.9265.9367.9561.42
S-CLAPScoreMetric Architecture=CLAP-ONLY, Reference Protocol=Reference-free, Backbone=MS-CLAP2026.0364.0346.0753.6551.9765.8772.8967.2669.2161.3
CAF-ScoreMetric Architecture=CLAP + LALM, Reference Protocol=Reference-free, Backbone LALM=AudioFlamingo3-Base, Backbone CLAP=M2D-CLAP2026.0364.0370.3467.3868.1267.6671.7163.767.2167.63
CAF-ScoreMetric Architecture=CLAP + LALM, Reference Protocol=Reference-free, Backbone LALM=Qwen3-Omni-Instruct, Backbone CLAP=MS-CLAP2026.0364.0360.6763.9962.7167.6679.1770.0773.2168.39
CLAPScoreMetric Architecture=CLAP-ONLY, Reference Protocol=Reference-free, Backbone=LAION-CLAP2026.0363.3177.365.4269.7859.2875.6465.6368.6269.15
S-CLAPScoreMetric Architecture=CLAP-ONLY, Reference Protocol=Reference-free, Backbone=LAION-CLAP2026.0363.3177.365.4269.7859.8875.8366.9669.4369.59
CAF-ScoreMetric Architecture=CLAP + LALM, Reference Protocol=Reference-free, Backbone LALM=AudioFlamingo3-Base, Backbone CLAP=LAION-CLAP2026.0363.3176.6367.3870.4858.6877.0167.8570.1770.31
CAF-ScoreMetric Architecture=CLAP + LALM, Reference Protocol=Reference-free, Backbone LALM=Qwen3-Omni-Instruct, Backbone CLAP=LAION-CLAP2026.0363.3181.5768.0972.7561.6880.166871.872.23
CAF-ScoreMetric Architecture=CLAP + LALM, Reference Protocol=Reference-free, Backbone LALM=Qwen3-Omni-Instruct, Backbone CLAP=MGA-CLAP2026.0363.3181.1267.7472.473.6575.4463.8569.4370.79
CLAPScoreMetric Architecture=CLAP-ONLY, Reference Protocol=Reference-free, Backbone=MGA-CLAP2026.0362.5973.4865.2468.1271.2667.5863.4165.9566.95
S-CLAPScoreMetric Architecture=CLAP-ONLY, Reference Protocol=Reference-free, Backbone=MGA-CLAP2026.0362.5973.4865.2468.1272.4668.9662.5266.1767.51
CAF-ScoreMetric Architecture=CLAP + LALM, Reference Protocol=Reference-free, Backbone LALM=AudioFlamingo3-Base, Backbone CLAP=MS-CLAP2026.0362.5948.0960.0755.7264.6773.8768.7470.1763.54
FENSEReference Protocol=Reference-based2026.0361.1584.4967.3873.2856.8984.6864.1570.9872.04
FLEURMetric Architecture=LALM-ONLY, Reference Protocol=Reference-free, Backbone=AudioFlamingo3-Base2026.0353.2452.5866.4959.4863.4768.3761.1964.1762.02
FLEURMetric Architecture=LALM-ONLY, Reference Protocol=Reference-free, Backbone=Qwen3-Omni-Instruct2026.0353.2478.6566.3169.5267.6674.6661.6367.2868.31
FLEURMetric Architecture=LALM-ONLY, Reference Protocol=Reference-free, Backbone=Qwen3-Omni-Thinking2026.0351.871.4658.8262.8858.6868.9656.7461.5862.18
FLEURMetric Architecture=LALM-ONLY, Reference Protocol=Reference-free, Backbone=AudioFlamingo3-Think2026.0341.7352.5856.5153.1959.2854.8155.1155.1154.45
Raw ScoreMetric Architecture=LALM-ONLY, Reference Protocol=Reference-free, Backbone=AudioFlamingo3-Base2026.0333.8128.0935.1232.2313.7722.423.2621.7626.56
Raw ScoreMetric Architecture=LALM-ONLY, Reference Protocol=Reference-free, Backbone=Qwen3-Omni-Thinking2026.0325.955.7343.1445.9441.9255.442.2247.1546.59
Raw ScoreMetric Architecture=LALM-ONLY, Reference Protocol=Reference-free, Backbone=Qwen3-Omni-Instruct2026.0316.5548.544140.9635.9354.6245.0447.5244.51
Raw ScoreMetric Architecture=LALM-ONLY, Reference Protocol=Reference-free, Backbone=AudioFlamingo3-Think2026.0315.8322.4719.072021.5621.2221.3321.3220.71