Upload organize_model_results.json with huggingface_hub
Browse files- organize_model_results.json +15 -0
organize_model_results.json
CHANGED
|
@@ -149,6 +149,7 @@
|
|
| 149 |
"whisper_large_v3": 0.11863959266711877,
|
| 150 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.11416493424197618,
|
| 151 |
"phi_4_multimodal_instruct": 0.15921168191570967,
|
|
|
|
| 152 |
"WavLLM_fairseq": 0.6447482518259942,
|
| 153 |
"SALMONN_7B": 0.2577708974886327,
|
| 154 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.11773910240019567
|
|
@@ -174,6 +175,7 @@
|
|
| 174 |
"whisper_large_v3": 0.03660128246354058,
|
| 175 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.05307658841999735,
|
| 176 |
"phi_4_multimodal_instruct": 0.03879546787220762,
|
|
|
|
| 177 |
"WavLLM_fairseq": 0.04798834811886432,
|
| 178 |
"SALMONN_7B": 0.09671439650443565,
|
| 179 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.03714982881570734
|
|
@@ -187,6 +189,7 @@
|
|
| 187 |
"whisper_large_v3": 0.01878749009695552,
|
| 188 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.032349945297468596,
|
| 189 |
"phi_4_multimodal_instruct": 0.0167502923755989,
|
|
|
|
| 190 |
"WavLLM_fairseq": 0.02103218017882069,
|
| 191 |
"SALMONN_7B": 0.10270871845172973,
|
| 192 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.018334779492209605
|
|
@@ -267,6 +270,7 @@
|
|
| 267 |
"whisper_large_v3": 0.10001863741235596,
|
| 268 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.10600831614192711,
|
| 269 |
"phi_4_multimodal_instruct": 0.08262800367606891,
|
|
|
|
| 270 |
"WavLLM_fairseq": 0.14533325621300636,
|
| 271 |
"SALMONN_7B": 0.3062255383962828,
|
| 272 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.09876543209876543
|
|
@@ -599,6 +603,7 @@
|
|
| 599 |
"whisper_large_v3": 0.12359684029221357,
|
| 600 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.20886539565639167,
|
| 601 |
"phi_4_multimodal_instruct": 0.07466690423868068,
|
|
|
|
| 602 |
"WavLLM_fairseq": 0.7054601967888183,
|
| 603 |
"SALMONN_7B": 0.8259290055631446,
|
| 604 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.12450753301261111
|
|
@@ -902,6 +907,7 @@
|
|
| 902 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 46.7,
|
| 903 |
"gemini-1.5-flash": 31.4,
|
| 904 |
"phi_4_multimodal_instruct": 54.50000000000001,
|
|
|
|
| 905 |
"SALMONN_7B": 40.5,
|
| 906 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 51.0
|
| 907 |
},
|
|
@@ -912,6 +918,7 @@
|
|
| 912 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 52.7,
|
| 913 |
"gemini-1.5-flash": 58.199999999999996,
|
| 914 |
"phi_4_multimodal_instruct": 59.4,
|
|
|
|
| 915 |
"SALMONN_7B": 48.4,
|
| 916 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 55.60000000000001
|
| 917 |
}
|
|
@@ -924,6 +931,7 @@
|
|
| 924 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.39520958083832336,
|
| 925 |
"gemini-1.5-flash": 0.2904191616766467,
|
| 926 |
"phi_4_multimodal_instruct": 0.6377245508982036,
|
|
|
|
| 927 |
"SALMONN_7B": 0.4820359281437126,
|
| 928 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.5
|
| 929 |
},
|
|
@@ -934,6 +942,7 @@
|
|
| 934 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.44011976047904194,
|
| 935 |
"gemini-1.5-flash": 0.5868263473053892,
|
| 936 |
"phi_4_multimodal_instruct": 0.688622754491018,
|
|
|
|
| 937 |
"SALMONN_7B": 0.5598802395209581,
|
| 938 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.5359281437125748
|
| 939 |
}
|
|
@@ -946,6 +955,7 @@
|
|
| 946 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.45045045045045046,
|
| 947 |
"gemini-1.5-flash": 0.3483483483483483,
|
| 948 |
"phi_4_multimodal_instruct": 0.5975975975975976,
|
|
|
|
| 949 |
"SALMONN_7B": 0.4594594594594595,
|
| 950 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.46546546546546547
|
| 951 |
},
|
|
@@ -956,6 +966,7 @@
|
|
| 956 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.5345345345345346,
|
| 957 |
"gemini-1.5-flash": 0.5885885885885885,
|
| 958 |
"phi_4_multimodal_instruct": 0.6456456456456456,
|
|
|
|
| 959 |
"SALMONN_7B": 0.5105105105105106,
|
| 960 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.5105105105105106
|
| 961 |
}
|
|
@@ -968,6 +979,7 @@
|
|
| 968 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.5555555555555556,
|
| 969 |
"gemini-1.5-flash": 0.3033033033033033,
|
| 970 |
"phi_4_multimodal_instruct": 0.3993993993993994,
|
|
|
|
| 971 |
"SALMONN_7B": 0.2732732732732733,
|
| 972 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.5645645645645646
|
| 973 |
},
|
|
@@ -978,6 +990,7 @@
|
|
| 978 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.6066066066066066,
|
| 979 |
"gemini-1.5-flash": 0.5705705705705706,
|
| 980 |
"phi_4_multimodal_instruct": 0.44744744744744747,
|
|
|
|
| 981 |
"SALMONN_7B": 0.3813813813813814,
|
| 982 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.6216216216216216
|
| 983 |
}
|
|
@@ -1029,6 +1042,7 @@
|
|
| 1029 |
"whisper_large_v3": 0.03208650948413402,
|
| 1030 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.04396383619925545,
|
| 1031 |
"phi_4_multimodal_instruct": 0.0381847190214501,
|
|
|
|
| 1032 |
"WavLLM_fairseq": 0.4536784258110264,
|
| 1033 |
"SALMONN_7B": 0.14231519234178336,
|
| 1034 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.04754476156709803
|
|
@@ -1054,6 +1068,7 @@
|
|
| 1054 |
"whisper_large_v3": 0.037649480146197796,
|
| 1055 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.04900464852205386,
|
| 1056 |
"phi_4_multimodal_instruct": 0.028494375643163834,
|
|
|
|
| 1057 |
"WavLLM_fairseq": 0.06621482559171073,
|
| 1058 |
"SALMONN_7B": 0.0459884319222171,
|
| 1059 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.038146268762641496
|
|
|
|
| 149 |
"whisper_large_v3": 0.11863959266711877,
|
| 150 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.11416493424197618,
|
| 151 |
"phi_4_multimodal_instruct": 0.15921168191570967,
|
| 152 |
+
"seallms_audio_7b": 0.5115646296316884,
|
| 153 |
"WavLLM_fairseq": 0.6447482518259942,
|
| 154 |
"SALMONN_7B": 0.2577708974886327,
|
| 155 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.11773910240019567
|
|
|
|
| 175 |
"whisper_large_v3": 0.03660128246354058,
|
| 176 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.05307658841999735,
|
| 177 |
"phi_4_multimodal_instruct": 0.03879546787220762,
|
| 178 |
+
"seallms_audio_7b": 0.09453912648722265,
|
| 179 |
"WavLLM_fairseq": 0.04798834811886432,
|
| 180 |
"SALMONN_7B": 0.09671439650443565,
|
| 181 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.03714982881570734
|
|
|
|
| 189 |
"whisper_large_v3": 0.01878749009695552,
|
| 190 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.032349945297468596,
|
| 191 |
"phi_4_multimodal_instruct": 0.0167502923755989,
|
| 192 |
+
"seallms_audio_7b": 0.0509676689176444,
|
| 193 |
"WavLLM_fairseq": 0.02103218017882069,
|
| 194 |
"SALMONN_7B": 0.10270871845172973,
|
| 195 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.018334779492209605
|
|
|
|
| 270 |
"whisper_large_v3": 0.10001863741235596,
|
| 271 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.10600831614192711,
|
| 272 |
"phi_4_multimodal_instruct": 0.08262800367606891,
|
| 273 |
+
"seallms_audio_7b": 0.14429855849255468,
|
| 274 |
"WavLLM_fairseq": 0.14533325621300636,
|
| 275 |
"SALMONN_7B": 0.3062255383962828,
|
| 276 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.09876543209876543
|
|
|
|
| 603 |
"whisper_large_v3": 0.12359684029221357,
|
| 604 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.20886539565639167,
|
| 605 |
"phi_4_multimodal_instruct": 0.07466690423868068,
|
| 606 |
+
"seallms_audio_7b": 0.11804359446457208,
|
| 607 |
"WavLLM_fairseq": 0.7054601967888183,
|
| 608 |
"SALMONN_7B": 0.8259290055631446,
|
| 609 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.12450753301261111
|
|
|
|
| 907 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 46.7,
|
| 908 |
"gemini-1.5-flash": 31.4,
|
| 909 |
"phi_4_multimodal_instruct": 54.50000000000001,
|
| 910 |
+
"seallms_audio_7b": 51.5,
|
| 911 |
"SALMONN_7B": 40.5,
|
| 912 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 51.0
|
| 913 |
},
|
|
|
|
| 918 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 52.7,
|
| 919 |
"gemini-1.5-flash": 58.199999999999996,
|
| 920 |
"phi_4_multimodal_instruct": 59.4,
|
| 921 |
+
"seallms_audio_7b": 60.199999999999996,
|
| 922 |
"SALMONN_7B": 48.4,
|
| 923 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 55.60000000000001
|
| 924 |
}
|
|
|
|
| 931 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.39520958083832336,
|
| 932 |
"gemini-1.5-flash": 0.2904191616766467,
|
| 933 |
"phi_4_multimodal_instruct": 0.6377245508982036,
|
| 934 |
+
"seallms_audio_7b": 0.6047904191616766,
|
| 935 |
"SALMONN_7B": 0.4820359281437126,
|
| 936 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.5
|
| 937 |
},
|
|
|
|
| 942 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.44011976047904194,
|
| 943 |
"gemini-1.5-flash": 0.5868263473053892,
|
| 944 |
"phi_4_multimodal_instruct": 0.688622754491018,
|
| 945 |
+
"seallms_audio_7b": 0.6646706586826348,
|
| 946 |
"SALMONN_7B": 0.5598802395209581,
|
| 947 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.5359281437125748
|
| 948 |
}
|
|
|
|
| 955 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.45045045045045046,
|
| 956 |
"gemini-1.5-flash": 0.3483483483483483,
|
| 957 |
"phi_4_multimodal_instruct": 0.5975975975975976,
|
| 958 |
+
"seallms_audio_7b": 0.5165165165165165,
|
| 959 |
"SALMONN_7B": 0.4594594594594595,
|
| 960 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.46546546546546547
|
| 961 |
},
|
|
|
|
| 966 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.5345345345345346,
|
| 967 |
"gemini-1.5-flash": 0.5885885885885885,
|
| 968 |
"phi_4_multimodal_instruct": 0.6456456456456456,
|
| 969 |
+
"seallms_audio_7b": 0.6486486486486487,
|
| 970 |
"SALMONN_7B": 0.5105105105105106,
|
| 971 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.5105105105105106
|
| 972 |
}
|
|
|
|
| 979 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.5555555555555556,
|
| 980 |
"gemini-1.5-flash": 0.3033033033033033,
|
| 981 |
"phi_4_multimodal_instruct": 0.3993993993993994,
|
| 982 |
+
"seallms_audio_7b": 0.42342342342342343,
|
| 983 |
"SALMONN_7B": 0.2732732732732733,
|
| 984 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.5645645645645646
|
| 985 |
},
|
|
|
|
| 990 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.6066066066066066,
|
| 991 |
"gemini-1.5-flash": 0.5705705705705706,
|
| 992 |
"phi_4_multimodal_instruct": 0.44744744744744747,
|
| 993 |
+
"seallms_audio_7b": 0.4924924924924925,
|
| 994 |
"SALMONN_7B": 0.3813813813813814,
|
| 995 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.6216216216216216
|
| 996 |
}
|
|
|
|
| 1042 |
"whisper_large_v3": 0.03208650948413402,
|
| 1043 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.04396383619925545,
|
| 1044 |
"phi_4_multimodal_instruct": 0.0381847190214501,
|
| 1045 |
+
"seallms_audio_7b": 0.3208650948413402,
|
| 1046 |
"WavLLM_fairseq": 0.4536784258110264,
|
| 1047 |
"SALMONN_7B": 0.14231519234178336,
|
| 1048 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.04754476156709803
|
|
|
|
| 1068 |
"whisper_large_v3": 0.037649480146197796,
|
| 1069 |
"cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct": 0.04900464852205386,
|
| 1070 |
"phi_4_multimodal_instruct": 0.028494375643163834,
|
| 1071 |
+
"seallms_audio_7b": 0.04829495049856286,
|
| 1072 |
"WavLLM_fairseq": 0.06621482559171073,
|
| 1073 |
"SALMONN_7B": 0.0459884319222171,
|
| 1074 |
"cascade_whisper_large_v3_llama_3_8b_instruct": 0.038146268762641496
|