Upload organize_model_results.json with huggingface_hub
Browse files- organize_model_results.json +49 -34
organize_model_results.json
CHANGED
@@ -1,4 +1,39 @@
|
|
1 |
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
"voxceleb_accent_test": {
|
3 |
"llama3_70b_judge": {
|
4 |
"Qwen-Audio-Chat": 48.05088223225277,
|
@@ -606,22 +641,6 @@
|
|
606 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 52.252747252747255
|
607 |
}
|
608 |
},
|
609 |
-
"cn_college_listen_mcq_test": {
|
610 |
-
"llama3_70b_judge": {
|
611 |
-
"Qwen-Audio-Chat": 63.232056362835756,
|
612 |
-
"MERaLiON-AudioLLM-Whisper-SEA-LION": 88.50726552179657,
|
613 |
-
"Qwen2-Audio-7B-Instruct": 74.7247908410392,
|
614 |
-
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 91.85380889476001,
|
615 |
-
"gemini-1.5-flash": 89.25583443416997,
|
616 |
-
"phi_4_multimodal_instruct": 73.18361955085865,
|
617 |
-
"WavLLM_fairseq": 66.31439894319684,
|
618 |
-
"SALMONN_7B": 50.99075297225891,
|
619 |
-
"cascade_whisper_large_v3_llama_3_8b_instruct": 85.2928225451343
|
620 |
-
},
|
621 |
-
"gpt4o_judge": {
|
622 |
-
"MERaLiON-AudioLLM-Whisper-SEA-LION": 88.77146631439894
|
623 |
-
}
|
624 |
-
},
|
625 |
"dream_tts_mcq_test": {
|
626 |
"llama3_70b_judge": {
|
627 |
"Qwen-Audio-Chat": 59.749085206481965,
|
@@ -1006,6 +1025,9 @@
|
|
1006 |
"seallms_audio_7b": 60.199999999999996,
|
1007 |
"SALMONN_7B": 48.4,
|
1008 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 55.60000000000001
|
|
|
|
|
|
|
1009 |
}
|
1010 |
},
|
1011 |
"mmau_mini_music": {
|
@@ -1030,6 +1052,9 @@
|
|
1030 |
"seallms_audio_7b": 0.6646706586826348,
|
1031 |
"SALMONN_7B": 0.5598802395209581,
|
1032 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.5359281437125748
|
|
|
|
|
|
|
1033 |
}
|
1034 |
},
|
1035 |
"mmau_mini_sound": {
|
@@ -1054,6 +1079,9 @@
|
|
1054 |
"seallms_audio_7b": 0.6486486486486487,
|
1055 |
"SALMONN_7B": 0.5105105105105106,
|
1056 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.5105105105105106
|
|
|
|
|
|
|
1057 |
}
|
1058 |
},
|
1059 |
"mmau_mini_speech": {
|
@@ -1078,6 +1106,9 @@
|
|
1078 |
"seallms_audio_7b": 0.4924924924924925,
|
1079 |
"SALMONN_7B": 0.3813813813813814,
|
1080 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.6216216216216216
|
|
|
|
|
|
|
1081 |
}
|
1082 |
},
|
1083 |
"imda_part5_30s_sqa_test": {
|
@@ -1108,23 +1139,6 @@
|
|
1108 |
"MERaLiON-AudioLLM-Whisper-SEA-LION": 66.9
|
1109 |
}
|
1110 |
},
|
1111 |
-
"slue_p2_sqa5_test": {
|
1112 |
-
"llama3_70b_judge": {
|
1113 |
-
"Qwen-Audio-Chat": 79.36274509803921,
|
1114 |
-
"MERaLiON-AudioLLM-Whisper-SEA-LION": 86.76470588235293,
|
1115 |
-
"Qwen2-Audio-7B-Instruct": 80.04901960784315,
|
1116 |
-
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 88.57843137254902,
|
1117 |
-
"phi_4_multimodal_instruct": 88.33333333333334,
|
1118 |
-
"seallms_audio_7b": 83.52941176470588,
|
1119 |
-
"WavLLM_fairseq": 83.92156862745098,
|
1120 |
-
"SALMONN_7B": 83.48039215686273,
|
1121 |
-
"cascade_whisper_large_v3_llama_3_8b_instruct": 82.99019607843137
|
1122 |
-
},
|
1123 |
-
"gpt4o_judge": {
|
1124 |
-
"MERaLiON-AudioLLM-Whisper-SEA-LION": 88.23529411764707,
|
1125 |
-
"cascade_whisper_large_v3_llama_3_8b_instruct": 87.79411764705883
|
1126 |
-
}
|
1127 |
-
},
|
1128 |
"tedlium3_long_form_test": {
|
1129 |
"wer": {
|
1130 |
"Qwen-Audio-Chat": 0.2911540507002305,
|
@@ -1286,7 +1300,8 @@
|
|
1286 |
},
|
1287 |
"gigaspeech2_thai": {
|
1288 |
"wer": {
|
1289 |
-
"MERaLiON-AudioLLM-Whisper-SEA-LION": 0.8901628256099774
|
|
|
1290 |
}
|
1291 |
},
|
1292 |
"gigaspeech2_indo": {
|
|
|
1 |
{
|
2 |
+
"cn_college_listen_mcq_test": {
|
3 |
+
"llama3_70b_judge": {
|
4 |
+
"gpt-4o-audio": 90.88507265521797,
|
5 |
+
"Qwen-Audio-Chat": 63.232056362835756,
|
6 |
+
"MERaLiON-AudioLLM-Whisper-SEA-LION": 88.50726552179657,
|
7 |
+
"Qwen2-Audio-7B-Instruct": 74.7247908410392,
|
8 |
+
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 91.85380889476001,
|
9 |
+
"gemini-1.5-flash": 89.25583443416997,
|
10 |
+
"phi_4_multimodal_instruct": 73.18361955085865,
|
11 |
+
"WavLLM_fairseq": 66.31439894319684,
|
12 |
+
"SALMONN_7B": 50.99075297225891,
|
13 |
+
"cascade_whisper_large_v3_llama_3_8b_instruct": 85.2928225451343
|
14 |
+
},
|
15 |
+
"gpt4o_judge": {
|
16 |
+
"MERaLiON-AudioLLM-Whisper-SEA-LION": 88.77146631439894
|
17 |
+
}
|
18 |
+
},
|
19 |
+
"slue_p2_sqa5_test": {
|
20 |
+
"llama3_70b_judge": {
|
21 |
+
"gpt-4o-audio": 89.41176470588235,
|
22 |
+
"Qwen-Audio-Chat": 79.36274509803921,
|
23 |
+
"MERaLiON-AudioLLM-Whisper-SEA-LION": 86.76470588235293,
|
24 |
+
"Qwen2-Audio-7B-Instruct": 80.04901960784315,
|
25 |
+
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 88.57843137254902,
|
26 |
+
"phi_4_multimodal_instruct": 88.33333333333334,
|
27 |
+
"seallms_audio_7b": 83.52941176470588,
|
28 |
+
"WavLLM_fairseq": 83.92156862745098,
|
29 |
+
"SALMONN_7B": 83.48039215686273,
|
30 |
+
"cascade_whisper_large_v3_llama_3_8b_instruct": 82.99019607843137
|
31 |
+
},
|
32 |
+
"gpt4o_judge": {
|
33 |
+
"MERaLiON-AudioLLM-Whisper-SEA-LION": 88.23529411764707,
|
34 |
+
"cascade_whisper_large_v3_llama_3_8b_instruct": 87.79411764705883
|
35 |
+
}
|
36 |
+
},
|
37 |
"voxceleb_accent_test": {
|
38 |
"llama3_70b_judge": {
|
39 |
"Qwen-Audio-Chat": 48.05088223225277,
|
|
|
641 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 52.252747252747255
|
642 |
}
|
643 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
644 |
"dream_tts_mcq_test": {
|
645 |
"llama3_70b_judge": {
|
646 |
"Qwen-Audio-Chat": 59.749085206481965,
|
|
|
1025 |
"seallms_audio_7b": 60.199999999999996,
|
1026 |
"SALMONN_7B": 48.4,
|
1027 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 55.60000000000001
|
1028 |
+
},
|
1029 |
+
"gpt4o_judge": {
|
1030 |
+
"MERaLiON-AudioLLM-Whisper-SEA-LION": 63.9
|
1031 |
}
|
1032 |
},
|
1033 |
"mmau_mini_music": {
|
|
|
1052 |
"seallms_audio_7b": 0.6646706586826348,
|
1053 |
"SALMONN_7B": 0.5598802395209581,
|
1054 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.5359281437125748
|
1055 |
+
},
|
1056 |
+
"gpt4o_judge": {
|
1057 |
+
"MERaLiON-AudioLLM-Whisper-SEA-LION": 0.6347305389221557
|
1058 |
}
|
1059 |
},
|
1060 |
"mmau_mini_sound": {
|
|
|
1079 |
"seallms_audio_7b": 0.6486486486486487,
|
1080 |
"SALMONN_7B": 0.5105105105105106,
|
1081 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.5105105105105106
|
1082 |
+
},
|
1083 |
+
"gpt4o_judge": {
|
1084 |
+
"MERaLiON-AudioLLM-Whisper-SEA-LION": 0.6996996996996997
|
1085 |
}
|
1086 |
},
|
1087 |
"mmau_mini_speech": {
|
|
|
1106 |
"seallms_audio_7b": 0.4924924924924925,
|
1107 |
"SALMONN_7B": 0.3813813813813814,
|
1108 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.6216216216216216
|
1109 |
+
},
|
1110 |
+
"gpt4o_judge": {
|
1111 |
+
"MERaLiON-AudioLLM-Whisper-SEA-LION": 0.5825825825825826
|
1112 |
}
|
1113 |
},
|
1114 |
"imda_part5_30s_sqa_test": {
|
|
|
1139 |
"MERaLiON-AudioLLM-Whisper-SEA-LION": 66.9
|
1140 |
}
|
1141 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1142 |
"tedlium3_long_form_test": {
|
1143 |
"wer": {
|
1144 |
"Qwen-Audio-Chat": 0.2911540507002305,
|
|
|
1300 |
},
|
1301 |
"gigaspeech2_thai": {
|
1302 |
"wer": {
|
1303 |
+
"MERaLiON-AudioLLM-Whisper-SEA-LION": 0.8901628256099774,
|
1304 |
+
"seallms_audio_7b": 0.3332398502070376
|
1305 |
}
|
1306 |
},
|
1307 |
"gigaspeech2_indo": {
|