Upload 15 files
Browse files- app.py +95 -4
- file/result_v2.csv +27 -27
- file/result_v2_task.csv +27 -27
app.py
CHANGED
@@ -125,6 +125,7 @@ def add_new_eval(
|
|
125 |
content = input_file.decode("utf-8")
|
126 |
prediction = prediction_analyse(content)
|
127 |
csv_data = pd.read_csv(CSV_DIR)
|
|
|
128 |
|
129 |
Start_dimension, End_dimension = 1, 13
|
130 |
if Evaluation_dimension == 'Image':
|
@@ -142,18 +143,24 @@ def add_new_eval(
|
|
142 |
|
143 |
if Evaluation_dimension != 'Video':
|
144 |
average_accuracy_image = round(total_correct_image / total_image * 100, 1)
|
|
|
145 |
else:
|
146 |
average_accuracy_image = 0
|
|
|
147 |
|
148 |
if Evaluation_dimension != 'Image':
|
149 |
average_accuracy_video = round(total_correct_video / total_video * 100, 1)
|
|
|
150 |
else:
|
151 |
average_accuracy_video = 0
|
|
|
152 |
|
153 |
if Evaluation_dimension == 'All':
|
154 |
overall_accuracy = round((total_correct_image + total_correct_video) / (total_image + total_video) * 100, 1)
|
|
|
155 |
else:
|
156 |
overall_accuracy = 0
|
|
|
157 |
|
158 |
if LLM_type == 'Other':
|
159 |
LLM_name = LLM_name_textbox
|
@@ -202,11 +209,37 @@ def add_new_eval(
|
|
202 |
]
|
203 |
csv_data.loc[col] = new_data
|
204 |
csv_data = csv_data.to_csv(CSV_DIR, index=False)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
205 |
# v2 evaluation
|
206 |
else:
|
207 |
content = input_file.decode("utf-8")
|
208 |
prediction = prediction_analyse_v2(content)
|
209 |
csv_data = pd.read_csv(CSV_V2_DIR)
|
|
|
210 |
|
211 |
Start_dimension, End_dimension = 1, 28
|
212 |
if Evaluation_dimension_2 == 'L1':
|
@@ -217,14 +250,27 @@ def add_new_eval(
|
|
217 |
End_dimension = 28
|
218 |
# pdb.set_trace()
|
219 |
each_task_accuracy = {i: round(prediction[i]["correct"] / prediction[i]["total"] * 100, 1) if i >= Start_dimension and i < End_dimension else 0 for i in range(1, 28)}
|
220 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
221 |
|
222 |
if Evaluation_dimension_2 == 'L2':
|
223 |
-
average_p2 = round(sum(
|
|
|
224 |
average_p3 = 0
|
|
|
225 |
else:
|
226 |
-
average_p2 = round(sum(
|
227 |
-
|
|
|
|
|
228 |
|
229 |
if LLM_type == 'Other':
|
230 |
LLM_name = LLM_name_textbox
|
@@ -254,6 +300,9 @@ def add_new_eval(
|
|
254 |
LLM_name,
|
255 |
model_size,
|
256 |
Evaluation_method,
|
|
|
|
|
|
|
257 |
average_p1,
|
258 |
average_p2,
|
259 |
average_p3,
|
@@ -287,6 +336,48 @@ def add_new_eval(
|
|
287 |
]
|
288 |
csv_data.loc[col] = new_data
|
289 |
csv_data = csv_data.to_csv(CSV_V2_DIR, index=False)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
290 |
return 0
|
291 |
|
292 |
def get_baseline_df(average_type):
|
|
|
125 |
content = input_file.decode("utf-8")
|
126 |
prediction = prediction_analyse(content)
|
127 |
csv_data = pd.read_csv(CSV_DIR)
|
128 |
+
csv_task_data = pd.read_csv(CSV_TASK_DIR)
|
129 |
|
130 |
Start_dimension, End_dimension = 1, 13
|
131 |
if Evaluation_dimension == 'Image':
|
|
|
143 |
|
144 |
if Evaluation_dimension != 'Video':
|
145 |
average_accuracy_image = round(total_correct_image / total_image * 100, 1)
|
146 |
+
average_task_accuracy_image = round(sum(each_task_accuracy[key] for key in range(1,10)) / 9, 1)
|
147 |
else:
|
148 |
average_accuracy_image = 0
|
149 |
+
average_task_accuracy_image = 0
|
150 |
|
151 |
if Evaluation_dimension != 'Image':
|
152 |
average_accuracy_video = round(total_correct_video / total_video * 100, 1)
|
153 |
+
average_task_accuracy_video = round(sum(each_task_accuracy[key] for key in range(10,13)) / 3, 1)
|
154 |
else:
|
155 |
average_accuracy_video = 0
|
156 |
+
average_task_accuracy_video = 0
|
157 |
|
158 |
if Evaluation_dimension == 'All':
|
159 |
overall_accuracy = round((total_correct_image + total_correct_video) / (total_image + total_video) * 100, 1)
|
160 |
+
overall_task_accuracy = round(sum(each_task_accuracy[key] for key in range(1,13)) / 12, 1)
|
161 |
else:
|
162 |
overall_accuracy = 0
|
163 |
+
overall_task_accuracy = 0
|
164 |
|
165 |
if LLM_type == 'Other':
|
166 |
LLM_name = LLM_name_textbox
|
|
|
209 |
]
|
210 |
csv_data.loc[col] = new_data
|
211 |
csv_data = csv_data.to_csv(CSV_DIR, index=False)
|
212 |
+
|
213 |
+
new_task_data = [
|
214 |
+
model_type,
|
215 |
+
model_name,
|
216 |
+
LLM_name,
|
217 |
+
model_size,
|
218 |
+
Evaluation_method,
|
219 |
+
overall_task_accuracy,
|
220 |
+
average_task_accuracy_image,
|
221 |
+
average_task_accuracy_video,
|
222 |
+
each_task_accuracy[1],
|
223 |
+
each_task_accuracy[2],
|
224 |
+
each_task_accuracy[3],
|
225 |
+
each_task_accuracy[4],
|
226 |
+
each_task_accuracy[5],
|
227 |
+
each_task_accuracy[6],
|
228 |
+
each_task_accuracy[7],
|
229 |
+
each_task_accuracy[8],
|
230 |
+
each_task_accuracy[9],
|
231 |
+
each_task_accuracy[10],
|
232 |
+
each_task_accuracy[11],
|
233 |
+
each_task_accuracy[12],
|
234 |
+
]
|
235 |
+
csv_task_data.loc[col] = new_data
|
236 |
+
csv_task_data = csv_task_data.to_csv(CSV_TASK_DIR, index=False)
|
237 |
# v2 evaluation
|
238 |
else:
|
239 |
content = input_file.decode("utf-8")
|
240 |
prediction = prediction_analyse_v2(content)
|
241 |
csv_data = pd.read_csv(CSV_V2_DIR)
|
242 |
+
csv_task_data = pd.read_csv(CSV_V2_TASK_DIR)
|
243 |
|
244 |
Start_dimension, End_dimension = 1, 28
|
245 |
if Evaluation_dimension_2 == 'L1':
|
|
|
250 |
End_dimension = 28
|
251 |
# pdb.set_trace()
|
252 |
each_task_accuracy = {i: round(prediction[i]["correct"] / prediction[i]["total"] * 100, 1) if i >= Start_dimension and i < End_dimension else 0 for i in range(1, 28)}
|
253 |
+
|
254 |
+
average_single = round(sum(prediction[i]["correct"] for i in range(1, 17)) / sum(prediction[i]["total"] for i in range(1, 17)) * 100, 1)
|
255 |
+
average_multi = round(sum(prediction[i]["correct"] for i in range(17, 19)) / sum(prediction[i]["total"] for i in range(17, 19)) * 100, 1)
|
256 |
+
average_video = round(sum(prediction[i]["correct"] for i in range(19, 23)) / sum(prediction[i]["total"] for i in range(19, 23)) * 100, 1)
|
257 |
+
average_p1 = round(sum(prediction[i]["correct"] for i in range(1, 23)) / sum(prediction[i]["total"] for i in range(1, 23)) * 100, 1)
|
258 |
+
|
259 |
+
average_task_single = round(sum(each_task_accuracy[key] for key in range(1,17)) / 16, 1)
|
260 |
+
average_task_multi = round(sum(each_task_accuracy[key] for key in range(17,19)) / 2, 1)
|
261 |
+
average_task_video = round(sum(each_task_accuracy[key] for key in range(19,23)) / 4, 1)
|
262 |
+
average_task_p1 = round(sum(each_task_accuracy[key] for key in range(1,23)) / 22, 1)
|
263 |
|
264 |
if Evaluation_dimension_2 == 'L2':
|
265 |
+
average_p2 = round(sum(prediction[i]["correct"] for i in range(23, 25)) / sum(prediction[i]["total"] for i in range(23, 25)) * 100, 1)
|
266 |
+
average_task_p2 = round(sum(each_task_accuracy[key] for key in range(23,25)) / 2, 1)
|
267 |
average_p3 = 0
|
268 |
+
average_task_p3 = 0
|
269 |
else:
|
270 |
+
average_p2 = round(sum(prediction[i]["correct"] for i in range(23, 25)) / sum(prediction[i]["total"] for i in range(23, 25)) * 100, 1)
|
271 |
+
average_task_p2 = round(sum(each_task_accuracy[key] for key in range(23,25)) / 2, 1)
|
272 |
+
average_p3 = round(sum(prediction[i]["correct"] for i in range(25, 28)) / sum(prediction[i]["total"] for i in range(25, 28)) * 100, 1)
|
273 |
+
average_task_p3 = round(sum(each_task_accuracy[key] for key in range(25,28)) / 3, 1)
|
274 |
|
275 |
if LLM_type == 'Other':
|
276 |
LLM_name = LLM_name_textbox
|
|
|
300 |
LLM_name,
|
301 |
model_size,
|
302 |
Evaluation_method,
|
303 |
+
average_single,
|
304 |
+
average_multi,
|
305 |
+
average_video,
|
306 |
average_p1,
|
307 |
average_p2,
|
308 |
average_p3,
|
|
|
336 |
]
|
337 |
csv_data.loc[col] = new_data
|
338 |
csv_data = csv_data.to_csv(CSV_V2_DIR, index=False)
|
339 |
+
|
340 |
+
new_task_data = [
|
341 |
+
model_name,
|
342 |
+
LLM_name,
|
343 |
+
model_size,
|
344 |
+
Evaluation_method,
|
345 |
+
average_task_single,
|
346 |
+
average_task_multi,
|
347 |
+
average_task_video,
|
348 |
+
average_task_p1,
|
349 |
+
average_task_p2,
|
350 |
+
average_task_p3,
|
351 |
+
each_task_accuracy[1],
|
352 |
+
each_task_accuracy[2],
|
353 |
+
each_task_accuracy[3],
|
354 |
+
each_task_accuracy[4],
|
355 |
+
each_task_accuracy[5],
|
356 |
+
each_task_accuracy[6],
|
357 |
+
each_task_accuracy[7],
|
358 |
+
each_task_accuracy[8],
|
359 |
+
each_task_accuracy[9],
|
360 |
+
each_task_accuracy[10],
|
361 |
+
each_task_accuracy[11],
|
362 |
+
each_task_accuracy[12],
|
363 |
+
each_task_accuracy[13],
|
364 |
+
each_task_accuracy[14],
|
365 |
+
each_task_accuracy[15],
|
366 |
+
each_task_accuracy[16],
|
367 |
+
each_task_accuracy[17],
|
368 |
+
each_task_accuracy[18],
|
369 |
+
each_task_accuracy[19],
|
370 |
+
each_task_accuracy[20],
|
371 |
+
each_task_accuracy[21],
|
372 |
+
each_task_accuracy[22],
|
373 |
+
each_task_accuracy[23],
|
374 |
+
each_task_accuracy[24],
|
375 |
+
each_task_accuracy[25],
|
376 |
+
each_task_accuracy[26],
|
377 |
+
each_task_accuracy[27]
|
378 |
+
]
|
379 |
+
csv_task_data.loc[col] = new_task_data
|
380 |
+
csv_task_data = csv_task_data.to_csv(CSV_V2_TASK_DIR, index=False)
|
381 |
return 0
|
382 |
|
383 |
def get_baseline_df(average_type):
|
file/result_v2.csv
CHANGED
@@ -1,28 +1,28 @@
|
|
1 |
Model,Language Model,Model Size,Evaluation Method,Avg. Single,Avg. Multi,Avg. Video,Avg. P1,Avg. P2,Avg. P3,Scene Understanding,Instance Identity,Instance Attribute,Instance Location,Instance Counting,Spatial Relation,Instance Interaction,Visual Reasoning,Text Recognition,Celebrity Recognition,Landmark Recognition,Chart Understanding,Visual Referring Expression,Science Knowledge,Emotion Recognition,Visual Mathematics,Difference Spotting,Meme Comprehension,Global Video Understanding,Action Recognition,Action Predicion,Procedure Understanding,In-Context Captioning,Interleaved Image-Text Analysis,Text-to-Image Generation,Next Image Prediction,Text-Image Creation
|
2 |
-
[BLIP-2](https://github.com/salesforce/LAVIS),Flan-T5-XL,3B,PPL,46.8,22.8,36.0,44.2,37.3,0.0,58.5,48.6,49,39.1,43.4,36.2,48.5,52.9,60.7,51.8,51.4,19.2,43.2,52.4,29.3,22,17.8,38.6,42.5,37.7,36.2,22.9,40,30.6,0,0,0
|
3 |
-
[InstructBLIP](https://github.com/salesforce/LAVIS),Flan-T5-XL,3B,PPL,52.4,25.8,36.5,48.6,36.1,0.0,58.9,49.7,61.7,35.1,58.1,34.9,47.4,55.9,61.4,48.5,45.4,26.4,41.7,47.7,34.5,21.2,22.8,35.2,41.5,36.1,40.5,24.5,36.7,34.7,0,0,0
|
4 |
-
[InstructBLIP-Vicuna](https://github.com/salesforce/LAVIS),Vicuna-7B,7B,PPL,47.5,41.1,33.0,44.2,28.4,0.0,53.6,43.9,49,37.8,56.5,35.8,43.3,56.2,57.2,60.3,44.4,27.9,39.2,39.4,23,26.5,36.5,55.4,40.4,38.6,31.2,15.6,26.7,32.7,0,0,0
|
5 |
-
[LLaVA](https://github.com/haotian-liu/LLaVA),LLaMA-7B,7B,PPL,42.4,32.5,32.6,40.2,34.3,0.0,53.8,47.5,38.3,34.2,42,34.7,40.2,52.9,46.4,51.8,45.6,30.3,40.2,37.6,34.3,20.5,27,50,44.1,36.2,25.1,18.6,40,20.4,0,0,0
|
6 |
-
[MiniGPT-4](https://github.com/Vision-CAIR/MiniGPT-4),Vicuna-7B,7B,PPL,45.0,25.7,34.3,42.5,39.0,0.0,56.3,49.2,45.8,37.9,45.3,32.6,47.4,57.1,41.8,55.2,45.2,20.2,41.2,43.3,24.2,25,19,46.7,39,38.7,27.4,28.6,45.8,22.5,0,0,0
|
7 |
-
[VPGTrans](https://github.com/VPGTrans/VPGTrans),LLaMA-7B,7B,PPL,36.6,29.3,33.2,35.8,21.9,0.0,46.9,38.6,33.6,35.6,27.5,34.4,33,50.8,47.6,52.4,38.2,30.1,34.7,36.1,31.5,27.3,24.6,44,37.8,38.2,20.9,33.5,19.2,28.6,0,0,0
|
8 |
-
[MultiModal-GPT](https://github.com/open-mmlab/Multimodal-GPT),LLaMA-7B,7B,PPL,36.7,44.1,32.6,35.9,36.7,0.0,46.9,42.5,32,32.3,27.7,29.7,29.9,48.3,35.2,60.9,50.4,24.2,42.2,37.6,32.1,27.3,40.1,56.5,37.6,38.7,25.3,24.4,39.2,30.6,0,0,0
|
9 |
-
[Otter](https://github.com/Luodian/Otter),LLaMA-7B,7B,PPL,36.0,32.0,32.3,35.2,39.0,0.0,45.9,39.7,31.9,31.6,26.4,32,33,49.2,39.3,59.7,53,23.6,41.2,36.1,37.3,22,27.4,46.7,36.6,37.9,26,24.8,42.5,30.6,0,0,0
|
10 |
-
[OpenFlamingo](https://github.com/mlfoundations/open_flamingo),LLaMA-7B,7B,PPL,36.6,43.5,32.4,35.8,36.7,0.0,46.7,42.3,31.7,33.4,27.4,29.8,29.9,47.7,35.6,60.3,49.8,24.2,42.2,39,32.1,27.3,39.9,54.9,37.6,38.4,25.2,24.1,38.3,32.7,0,0,0
|
11 |
-
[LLaMA-AdapterV2](https://github.com/OpenGVLab/LLaMA-Adapter),LLaMA-7B,7B,PPL,36.0,34.7,31.4,35.1,0.0,0.0,45.2,38.5,29.3,33,29.7,35.5,39.2,52,48.7,58.5,46.4,24.2,41.2,40.1,39.7,23.5,29.1,52.2,41.9,38.2,18.8,20.3,0,0,0,0,0
|
12 |
-
[GVT](https://github.com/TencentARC/GVT),Vicuna-7B,7B,PPL,34.9,45.8,31.0,34.2,40.2,0.0,41.7,35.5,31.8,29.5,36.2,32,32,51.1,35.2,39.4,36.4,25,36.2,31.1,20.6,22.7,41.5,59.2,40.4,29.7,26.3,24.1,42.5,34.7,0,0,0
|
13 |
-
[mPLUG-Owl](https://github.com/X-PLUG/mPLUG-Owl),LLaMA-7B,7B,PPL,38.6,38.7,31.1,36.9,29.0,0.0,49.7,45.3,32.5,36.7,27.3,32.7,44.3,54.7,49.2,70.9,49.6,23.2,44.2,44,32.5,23.5,33.5,54.9,42,37.8,18.3,19.3,29.2,28.6,0,0,0
|
14 |
-
[Kosmos-2](https://github.com/microsoft/unilm/tree/master/kosmos-2),Decoder only 1.3B,1.3B,PPL,52.4,29.4,40.7,49.6,23.7,0.0,63.4,57.1,58.5,44,41.4,37.9,55.7,60.7,68.1,82.1,51.4,21.2,48.2,43.7,30.7,28,25.2,42.8,48.5,40.8,39.5,30,24.2,22.5,0,0,0
|
15 |
-
[Qwen-VL-Chat](https://huggingface.co/Qwen/Qwen-VL-Chat),Qwen-7B,7B,PPL,50.3,37.4,34.3,46.6,38.5,0.0,56.5,47.6,54.8,46.9,54.2,40.3,55.7,55,47.4,62.4,55.6,25.2,43.7,41.2,20.6,28.8,34.3,47.2,39.7,42.8,29.6,19.1,42.5,28.6,0,0,0
|
16 |
-
[Qwen-VL-plus](https://github.com/QwenLM/Qwen-VL/tree/master?tab=readme-ov-file#qwen-vl-plus),Qwen-LM,-,PPL for A/B/C/D,71.9,70.3,46.7,66.1,29.6,0.0,76.6,77.7,76.3,65.1,65.8,55.9,73.2,77.9,61.8,97,97.2,39.5,73.4,75.8,51.7,38.6,66.7,81.8,51.8,54.5,29.3,48,28.3,32.7,0,0,0
|
17 |
-
[LLaVA-1.5](https://github.com/haotian-liu/LLaVA),vicuna-7B,7B,PPL,58.3,39.2,36.9,53.3,34.4,0.0,63.7,62.4,66.7,51.3,60.2,38.5,47.4,59.8,69,60.6,49.8,25,45.7,56.7,31.1,24.2,35.7,50.3,46.1,39.4,29.4,28.1,39.2,22.5,0,0,0
|
18 |
-
[IDEFICS-9b-instruct](https://huggingface.co/HuggingFaceM4/idefics-9b-instruct),LLaMA-7B,7B,PPL,38.8,54.5,32.9,37.5,42.6,0.0,48.2,38.2,37.8,32.9,29,32.4,37.1,54.1,45.5,52.4,52.8,22.6,42.7,33.2,26.6,21.2,56.5,48.4,42.7,38.6,23.6,20.5,45.8,34.7,0,0,0
|
19 |
-
[InternLM-XComposer-VL](https://github.com/InternLM/InternLM-XComposer),InternLM-7B,7B,PPL,65.4,49.8,44.9,60.6,30.2,0.0,74.8,70.5,67.6,60.5,55.3,53.4,76.3,76.1,61.4,86.1,78,27.2,60.3,84.8,68.9,25.8,47.7,56.6,58.6,49.9,37.6,24.9,27.5,36.7,0,0,0
|
20 |
-
[SPHINXv1-1k](https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/SPHINX),LLaMA-2-13B,13B,Generate,68.5,37.7,40.0,61.7,32.5,0.0,75.5,72.4,75.1,63.1,67.6,50.4,64.9,76.7,60,81.5,82.4,21.8,60.3,58.5,65.5,32.6,35.9,43.4,52.4,41.2,33.9,26.1,33.3,30.6,0,0,0
|
21 |
-
[SPHINXv2-1k](https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/SPHINX),LLaMA-2-13B,13B,Generate,72.1,44.8,44.2,65.4,35.5,0.0,77.5,78.5,76.6,69,71,57.5,73.2,77.6,62.1,82.7,85.2,44.5,62.3,60.3,65.3,23.5,45.7,42.1,54.6,48.1,37.9,29.9,33.3,40.8,0,0,0
|
22 |
-
[Emu](https://github.com/baaivision/Emu),LLaMA-13B,13B,PPL,46.4,31.2,37.4,44.2,45.6,45.7,59,50,43.7,37.1,44.3,33.6,49.5,58.3,61.4,68.8,61.6,19,45.7,41.5,24.2,26.4,29.3,37.1,41.9,42.7,37.9,21.8,51.7,30.6,46.8,43.2,34.2
|
23 |
-
[Next-GPT](https://github.com/NExT-GPT/NExT-GPT),vicuna-7B,7B,PPL,31.0,27.8,30.7,31.0,40.3,42.8,36.4,35.1,25.6,29.9,36.1,30.9,39.2,41.7,31,30.9,27.4,21.2,34.2,31.8,24.4,17.4,24.2,39,35.5,33.8,25.6,24.5,46.7,24.5,45.1,19.8,36.7
|
24 |
-
[SEED-LLaMA](https://github.com/AILab-CVC/SEED),LLaMA2-Chat-13B,13B,PPL,49.9,32.4,39.0,47.3,48.0,50.6,64,55,51.3,45.4,43.3,37.9,56.7,59.2,57,55.5,52.8,18.8,49.3,44.8,28.8,24.4,29.5,41.5,46.7,39.4,43.9,20.3,54.2,32.7,50.2,40.7,65.8
|
25 |
-
[GPT-4V](https://openai.com/research/gpt-4v-system-card),\-,-,Generate,69.8,73.1,61.7,68.1,37.9,0.0,77.5,73.9,70.6,61.8,56.8,56.9,74.2,78.5,82.3,91.8,97.4,45.1,71.9,66.1,71.1,43.9,67.9,89.3,64.5,65.7,51.7,63.4,29.2,59.2,0,0,0
|
26 |
-
[VideoChat](https://github.com/OpenGVLab/Ask-Anything),Vicuna-7B,7B,PPL,36.7,35.4,34.2,36.2,37.3,0.0,44.3,40.7,32.2,36.9,32.9,32.6,42.3,51.1,45.7,35.2,46.8,20.6,43.2,39.4,34.3,19.7,30.3,51.6,41.5,34,30.6,27.4,40,30.6,0,0,0
|
27 |
-
[Video-ChatGPT](https://github.com/mbzuai-oryx/Video-ChatGPT),LLaMA-7B,7B,PPL,38.3,49.8,31.6,36.9,33.7,0.0,44.1,37,35.8,30.7,44.2,31.1,29.9,49.9,39.8,49.7,40.6,22,33.2,37.2,22.4,25,46.1,61.4,42.6,32.2,27,19,37.5,24.5,0,0,0
|
28 |
-
[Valley](https://github.com/RupertLuo/Valley),LLaMA-13B,13B,PPL,35.3,40.7,28.5,33.9,33.7,0.0,45.3,36.4,33.7,30.6,27.1,31.5,35.1,52,35.2,44.9,43.4,23.8,33.2,37.2,26,22.7,37.1,52.2,31.5,32.1,21.9,26.5,35.8,28.6,0,0,0
|
|
|
1 |
Model,Language Model,Model Size,Evaluation Method,Avg. Single,Avg. Multi,Avg. Video,Avg. P1,Avg. P2,Avg. P3,Scene Understanding,Instance Identity,Instance Attribute,Instance Location,Instance Counting,Spatial Relation,Instance Interaction,Visual Reasoning,Text Recognition,Celebrity Recognition,Landmark Recognition,Chart Understanding,Visual Referring Expression,Science Knowledge,Emotion Recognition,Visual Mathematics,Difference Spotting,Meme Comprehension,Global Video Understanding,Action Recognition,Action Predicion,Procedure Understanding,In-Context Captioning,Interleaved Image-Text Analysis,Text-to-Image Generation,Next Image Prediction,Text-Image Creation
|
2 |
+
[BLIP-2](https://github.com/salesforce/LAVIS),Flan-T5-XL,3B,PPL,46.8,22.8,36.0,44.2,37.3,0.0,58.5,48.6,49.0,39.1,43.4,36.2,48.5,52.9,60.7,51.8,51.4,19.2,43.2,52.4,29.3,22.0,17.8,38.6,42.5,37.7,36.2,22.9,40.0,30.6,0.0,0.0,0.0
|
3 |
+
[InstructBLIP](https://github.com/salesforce/LAVIS),Flan-T5-XL,3B,PPL,52.4,25.8,36.5,48.6,36.1,0.0,58.9,49.7,61.7,35.1,58.1,34.9,47.4,55.9,61.4,48.5,45.4,26.4,41.7,47.7,34.5,21.2,22.8,35.2,41.5,36.1,40.5,24.5,36.7,34.7,0.0,0.0,0.0
|
4 |
+
[InstructBLIP-Vicuna](https://github.com/salesforce/LAVIS),Vicuna-7B,7B,PPL,47.5,41.1,33.0,44.2,28.4,0.0,53.6,43.9,49.0,37.8,56.5,35.8,43.3,56.2,57.2,60.3,44.4,27.9,39.2,39.4,23.0,26.5,36.5,55.4,40.4,38.6,31.2,15.6,26.7,32.7,0.0,0.0,0.0
|
5 |
+
[LLaVA](https://github.com/haotian-liu/LLaVA),LLaMA-7B,7B,PPL,42.4,32.5,32.6,40.2,34.3,0.0,53.8,47.5,38.3,34.2,42.0,34.7,40.2,52.9,46.4,51.8,45.6,30.3,40.2,37.6,34.3,20.5,27.0,50.0,44.1,36.2,25.1,18.6,40.0,20.4,0.0,0.0,0.0
|
6 |
+
[MiniGPT-4](https://github.com/Vision-CAIR/MiniGPT-4),Vicuna-7B,7B,PPL,45.0,25.7,34.3,42.5,39.0,0.0,56.3,49.2,45.8,37.9,45.3,32.6,47.4,57.1,41.8,55.2,45.2,20.2,41.2,43.3,24.2,25.0,19.0,46.7,39.0,38.7,27.4,28.6,45.8,22.5,0.0,0.0,0.0
|
7 |
+
[VPGTrans](https://github.com/VPGTrans/VPGTrans),LLaMA-7B,7B,PPL,36.6,29.3,33.2,35.8,21.9,0.0,46.9,38.6,33.6,35.6,27.5,34.4,33.0,50.8,47.6,52.4,38.2,30.1,34.7,36.1,31.5,27.3,24.6,44.0,37.8,38.2,20.9,33.5,19.2,28.6,0.0,0.0,0.0
|
8 |
+
[MultiModal-GPT](https://github.com/open-mmlab/Multimodal-GPT),LLaMA-7B,7B,PPL,36.7,44.1,32.6,35.9,36.7,0.0,46.9,42.5,32.0,32.3,27.7,29.7,29.9,48.3,35.2,60.9,50.4,24.2,42.2,37.6,32.1,27.3,40.1,56.5,37.6,38.7,25.3,24.4,39.2,30.6,0.0,0.0,0.0
|
9 |
+
[Otter](https://github.com/Luodian/Otter),LLaMA-7B,7B,PPL,36.0,32.0,32.3,35.2,39.0,0.0,45.9,39.7,31.9,31.6,26.4,32.0,33.0,49.2,39.3,59.7,53.0,23.6,41.2,36.1,37.3,22.0,27.4,46.7,36.6,37.9,26.0,24.8,42.5,30.6,0.0,0.0,0.0
|
10 |
+
[OpenFlamingo](https://github.com/mlfoundations/open_flamingo),LLaMA-7B,7B,PPL,36.6,43.5,32.4,35.8,36.7,0.0,46.7,42.3,31.7,33.4,27.4,29.8,29.9,47.7,35.6,60.3,49.8,24.2,42.2,39.0,32.1,27.3,39.9,54.9,37.6,38.4,25.2,24.1,38.3,32.7,0.0,0.0,0.0
|
11 |
+
[LLaMA-AdapterV2](https://github.com/OpenGVLab/LLaMA-Adapter),LLaMA-7B,7B,PPL,36.0,34.7,31.4,35.1,0.0,0.0,45.2,38.5,29.3,33.0,29.7,35.5,39.2,52.0,48.7,58.5,46.4,24.2,41.2,40.1,39.7,23.5,29.1,52.2,41.9,38.2,18.8,20.3,0.0,0.0,0.0,0.0,0.0
|
12 |
+
[GVT](https://github.com/TencentARC/GVT),Vicuna-7B,7B,PPL,34.9,45.8,31.0,34.2,40.2,0.0,41.7,35.5,31.8,29.5,36.2,32.0,32.0,51.1,35.2,39.4,36.4,25.0,36.2,31.1,20.6,22.7,41.5,59.2,40.4,29.7,26.3,24.1,42.5,34.7,0.0,0.0,0.0
|
13 |
+
[mPLUG-Owl](https://github.com/X-PLUG/mPLUG-Owl),LLaMA-7B,7B,PPL,38.6,38.7,31.1,36.9,29.0,0.0,49.7,45.3,32.5,36.7,27.3,32.7,44.3,54.7,49.2,70.9,49.6,23.2,44.2,44.0,32.5,23.5,33.5,54.9,42.0,37.8,18.3,19.3,29.2,28.6,0.0,0.0,0.0
|
14 |
+
[Kosmos-2](https://github.com/microsoft/unilm/tree/master/kosmos-2),Decoder only 1.3B,1.3B,PPL,52.4,29.4,40.7,49.6,23.7,0.0,63.4,57.1,58.5,44.0,41.4,37.9,55.7,60.7,68.1,82.1,51.4,21.2,48.2,43.7,30.7,28.0,25.2,42.8,48.5,40.8,39.5,30.0,24.2,22.5,0.0,0.0,0.0
|
15 |
+
[Qwen-VL-Chat](https://huggingface.co/Qwen/Qwen-VL-Chat),Qwen-7B,7B,PPL,50.3,37.4,34.3,46.6,38.5,0.0,56.5,47.6,54.8,46.9,54.2,40.3,55.7,55.0,47.4,62.4,55.6,25.2,43.7,41.2,20.6,28.8,34.3,47.2,39.7,42.8,29.6,19.1,42.5,28.6,0.0,0.0,0.0
|
16 |
+
[Qwen-VL-plus](https://github.com/QwenLM/Qwen-VL/tree/master?tab=readme-ov-file#qwen-vl-plus),Qwen-LM,-,PPL for A/B/C/D,71.9,70.3,46.7,66.1,29.6,0.0,76.6,77.7,76.3,65.1,65.8,55.9,73.2,77.9,61.8,97.0,97.2,39.5,73.4,75.8,51.7,38.6,66.7,81.8,51.8,54.5,29.3,48.0,28.3,32.7,0.0,0.0,0.0
|
17 |
+
[LLaVA-1.5](https://github.com/haotian-liu/LLaVA),vicuna-7B,7B,PPL,58.3,39.2,36.9,53.3,34.4,0.0,63.7,62.4,66.7,51.3,60.2,38.5,47.4,59.8,69.0,60.6,49.8,25.0,45.7,56.7,31.1,24.2,35.7,50.3,46.1,39.4,29.4,28.1,39.2,22.5,0.0,0.0,0.0
|
18 |
+
[IDEFICS-9b-instruct](https://huggingface.co/HuggingFaceM4/idefics-9b-instruct),LLaMA-7B,7B,PPL,38.8,54.5,32.9,37.5,42.6,0.0,48.2,38.2,37.8,32.9,29.0,32.4,37.1,54.1,45.5,52.4,52.8,22.6,42.7,33.2,26.6,21.2,56.5,48.4,42.7,38.6,23.6,20.5,45.8,34.7,0.0,0.0,0.0
|
19 |
+
[InternLM-XComposer-VL](https://github.com/InternLM/InternLM-XComposer),InternLM-7B,7B,PPL,65.4,49.8,44.9,60.6,30.2,0.0,74.8,70.5,67.6,60.5,55.3,53.4,76.3,76.1,61.4,86.1,78.0,27.2,60.3,84.8,68.9,25.8,47.7,56.6,58.6,49.9,37.6,24.9,27.5,36.7,0.0,0.0,0.0
|
20 |
+
[SPHINXv1-1k](https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/SPHINX),LLaMA-2-13B,13B,Generate,68.5,37.7,40.0,61.7,32.5,0.0,75.5,72.4,75.1,63.1,67.6,50.4,64.9,76.7,60.0,81.5,82.4,21.8,60.3,58.5,65.5,32.6,35.9,43.4,52.4,41.2,33.9,26.1,33.3,30.6,0.0,0.0,0.0
|
21 |
+
[SPHINXv2-1k](https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/SPHINX),LLaMA-2-13B,13B,Generate,72.1,44.8,44.2,65.4,35.5,0.0,77.5,78.5,76.6,69.0,71.0,57.5,73.2,77.6,62.1,82.7,85.2,44.5,62.3,60.3,65.3,23.5,45.7,42.1,54.6,48.1,37.9,29.9,33.3,40.8,0.0,0.0,0.0
|
22 |
+
[Emu](https://github.com/baaivision/Emu),LLaMA-13B,13B,PPL,46.4,31.2,37.4,44.2,45.6,45.7,59.0,50.0,43.7,37.1,44.3,33.6,49.5,58.3,61.4,68.8,61.6,19.0,45.7,41.5,24.2,26.4,29.3,37.1,41.9,42.7,37.9,21.8,51.7,30.6,46.8,43.2,34.2
|
23 |
+
[Next-GPT](https://github.com/NExT-GPT/NExT-GPT),vicuna-7B,7B,PPL,31.0,27.8,30.7,31.0,40.3,42.8,36.4,35.1,25.6,29.9,36.1,30.9,39.2,41.7,31.0,30.9,27.4,21.2,34.2,31.8,24.4,17.4,24.2,39.0,35.5,33.8,25.6,24.5,46.7,24.5,45.1,19.8,36.7
|
24 |
+
[SEED-LLaMA](https://github.com/AILab-CVC/SEED),LLaMA2-Chat-13B,13B,PPL,49.9,32.4,39.0,47.3,48.0,50.6,64.0,55.0,51.3,45.4,43.3,37.9,56.7,59.2,57.0,55.5,52.8,18.8,49.3,44.8,28.8,24.4,29.5,41.5,46.7,39.4,43.9,20.3,54.2,32.7,50.2,40.7,65.8
|
25 |
+
[GPT-4V](https://openai.com/research/gpt-4v-system-card),\-,-,Generate,69.8,73.1,61.7,68.1,37.9,0.0,77.5,73.9,70.6,61.8,56.8,56.9,74.2,78.5,82.3,91.8,97.4,45.1,71.9,66.1,71.1,43.9,67.9,89.3,64.5,65.7,51.7,63.4,29.2,59.2,0.0,0.0,0.0
|
26 |
+
[VideoChat](https://github.com/OpenGVLab/Ask-Anything),Vicuna-7B,7B,PPL,36.7,35.4,34.2,36.2,37.3,0.0,44.3,40.7,32.2,36.9,32.9,32.6,42.3,51.1,45.7,35.2,46.8,20.6,43.2,39.4,34.3,19.7,30.3,51.6,41.5,34.0,30.6,27.4,40.0,30.6,0.0,0.0,0.0
|
27 |
+
[Video-ChatGPT](https://github.com/mbzuai-oryx/Video-ChatGPT),LLaMA-7B,7B,PPL,38.3,49.8,31.6,36.9,33.7,0.0,44.1,37.0,35.8,30.7,44.2,31.1,29.9,49.9,39.8,49.7,40.6,22.0,33.2,37.2,22.4,25.0,46.1,61.4,42.6,32.2,27.0,19.0,37.5,24.5,0.0,0.0,0.0
|
28 |
+
[Valley](https://github.com/RupertLuo/Valley),LLaMA-13B,13B,PPL,35.3,40.7,28.5,33.9,33.7,0.0,45.3,36.4,33.7,30.6,27.1,31.5,35.1,52.0,35.2,44.9,43.4,23.8,33.2,37.2,26.0,22.7,37.1,52.2,31.5,32.1,21.9,26.5,35.8,28.6,0.0,0.0,0.0
|
file/result_v2_task.csv
CHANGED
@@ -1,28 +1,28 @@
|
|
1 |
Model,Language Model,Model Size,Evaluation Method,Avg. Single,Avg. Multi,Avg. Video,Avg. P1,Avg. P2,Avg. P3,Scene Understanding,Instance Identity,Instance Attribute,Instance Location,Instance Counting,Spatial Relation,Instance Interaction,Visual Reasoning,Text Recognition,Celebrity Recognition,Landmark Recognition,Chart Understanding,Visual Referring Expression,Science Knowledge,Emotion Recognition,Visual Mathematics,Difference Spotting,Meme Comprehension,Global Video Understanding,Action Recognition,Action Predicion,Procedure Understanding,In-Context Captioning,Interleaved Image-Text Analysis,Text-to-Image Generation,Next Image Prediction,Text-Image Creation
|
2 |
-
[BLIP-2](https://github.com/salesforce/LAVIS),Flan-T5-XL,3B,PPL,44.1,28.2,34.8,41,35.3,0,58.5,48.6,49,39.1,43.4,36.2,48.5,52.9,60.7,51.8,51.4,19.2,43.2,52.4,29.3,22,17.8,38.6,42.5,37.7,36.2,22.9,40,30.6,0,0,0
|
3 |
-
[InstructBLIP](https://github.com/salesforce/LAVIS),Flan-T5-XL,3B,PPL,45.5,29,35.7,42.2,35.7,0,58.9,49.7,61.7,35.1,58.1,34.9,47.4,55.9,61.4,48.5,45.4,26.4,41.7,47.7,34.5,21.2,22.8,35.2,41.5,36.1,40.5,24.5,36.7,34.7,0,0,0
|
4 |
-
[InstructBLIP-Vicuna](https://github.com/salesforce/LAVIS),Vicuna-7B,7B,PPL,43.4,46,31.5,41.4,29.7,0,53.6,43.9,49,37.8,56.5,35.8,43.3,56.2,57.2,60.3,44.4,27.9,39.2,39.4,23,26.5,36.5,55.4,40.4,38.6,31.2,15.6,26.7,32.7,0,0,0
|
5 |
-
[LLaVA](https://github.com/haotian-liu/LLaVA),LLaMA-7B,7B,PPL,40.6,38.5,31,38.7,30.2,0,53.8,47.5,38.3,34.2,42,34.7,40.2,52.9,46.4,51.8,45.6,30.3,40.2,37.6,34.3,20.5,27,50,44.1,36.2,25.1,18.6,40,20.4,0,0,0
|
6 |
-
[MiniGPT-4](https://github.com/Vision-CAIR/MiniGPT-4),Vicuna-7B,7B,PPL,41.7,32.9,33.4,39.4,34.1,0,56.3,49.2,45.8,37.9,45.3,32.6,47.4,57.1,41.8,55.2,45.2,20.2,41.2,43.3,24.2,25,19,46.7,39,38.7,27.4,28.6,45.8,22.5,0,0,0
|
7 |
-
[VPGTrans](https://github.com/VPGTrans/VPGTrans),LLaMA-7B,7B,PPL,37.4,34.3,32.6,36.2,23.9,0,46.9,38.6,33.6,35.6,27.5,34.4,33,50.8,47.6,52.4,38.2,30.1,34.7,36.1,31.5,27.3,24.6,44,37.8,38.2,20.9,33.5,19.2,28.6,0,0,0
|
8 |
-
[MultiModal-GPT](https://github.com/open-mmlab/Multimodal-GPT),LLaMA-7B,7B,PPL,37.5,48.3,31.5,37.4,34.9,0,46.9,42.5,32,32.3,27.7,29.7,29.9,48.3,35.2,60.9,50.4,24.2,42.2,37.6,32.1,27.3,40.1,56.5,37.6,38.7,25.3,24.4,39.2,30.6,0,0,0
|
9 |
-
[Otter](https://github.com/Luodian/Otter),LLaMA-7B,7B,PPL,37.6,37.1,31.3,36.4,36.6,0,45.9,39.7,31.9,31.6,26.4,32,33,49.2,39.3,59.7,53,23.6,41.2,36.1,37.3,22,27.4,46.7,36.6,37.9,26,24.8,42.5,30.6,0,0,0
|
10 |
-
[OpenFlamingo](https://github.com/mlfoundations/open_flamingo),LLaMA-7B,7B,PPL,37.5,47.4,31.3,37.3,35.5,0,46.7,42.3,31.7,33.4,27.4,29.8,29.9,47.7,35.6,60.3,49.8,24.2,42.2,39,32.1,27.3,39.9,54.9,37.6,38.4,25.2,24.1,38.3,32.7,0,0,0
|
11 |
-
[LLaMA-AdapterV2](https://github.com/OpenGVLab/LLaMA-Adapter),LLaMA-7B,7B,PPL,39,40.7,29.8,37.5,0,0,45.2,38.5,29.3,33,29.7,35.5,39.2,52,48.7,58.5,46.4,24.2,41.2,40.1,39.7,23.5,29.1,52.2,41.9,38.2,18.8,20.3,0,0,0,0,0
|
12 |
-
[GVT](https://github.com/TencentARC/GVT),Vicuna-7B,7B,PPL,33.5,50.4,30.1,34.4,38.6,0,41.7,35.5,31.8,29.5,36.2,32,32,51.1,35.2,39.4,36.4,25,36.2,31.1,20.6,22.7,41.5,59.2,40.4,29.7,26.3,24.1,42.5,34.7,0,0,0
|
13 |
-
[mPLUG-Owl](https://github.com/X-PLUG/mPLUG-Owl),LLaMA-7B,7B,PPL,41.3,44.2,29.4,39.4,28.9,0,49.7,45.3,32.5,36.7,27.3,32.7,44.3,54.7,49.2,70.9,49.6,23.2,44.2,44,32.5,23.5,33.5,54.9,42,37.8,18.3,19.3,29.2,28.6,0,0,0
|
14 |
-
[Kosmos-2](https://github.com/microsoft/unilm/tree/master/kosmos-2),Decoder only 1.3B,1.3B,PPL,49.5,34,39.7,46.3,23.3,0,63.4,57.1,58.5,44,41.4,37.9,55.7,60.7,68.1,82.1,51.4,21.2,48.2,43.7,30.7,28,25.2,42.8,48.5,40.8,39.5,30,24.2,22.5,0,0,0
|
15 |
-
[Qwen-VL-Chat](https://huggingface.co/Qwen/Qwen-VL-Chat),Qwen-7B,7B,PPL,46,40.8,32.8,43.1,35.5,0,56.5,47.6,54.8,46.9,54.2,40.3,55.7,55,47.4,62.4,55.6,25.2,43.7,41.2,20.6,28.8,34.3,47.2,39.7,42.8,29.6,19.1,42.5,28.6,0,0,0
|
16 |
-
[Qwen-VL-plus](https://github.com/QwenLM/Qwen-VL/tree/master?tab=readme-ov-file#qwen-vl-plus),Qwen-LM,-,PPL for A/B/C/D,69,74.3,45.9,65.3,30.5,0,76.6,77.7,76.3,65.1,65.8,55.9,73.2,77.9,61.8,97,97.2,39.5,73.4,75.8,51.7,38.6,66.7,81.8,51.8,54.5,29.3,48,28.3,32.7,0,0,0
|
17 |
-
[LLaVA-1.5](https://github.com/haotian-liu/LLaVA),vicuna-7B,7B,PPL,50.8,43,35.8,47.3,30.8,0,63.7,62.4,66.7,51.3,60.2,38.5,47.4,59.8,69,60.6,49.8,25,45.7,56.7,31.1,24.2,35.7,50.3,46.1,39.4,29.4,28.1,39.2,22.5,0,0,0
|
18 |
-
[IDEFICS-9b-instruct](https://huggingface.co/HuggingFaceM4/idefics-9b-instruct),LLaMA-7B,7B,PPL,37.9,52.5,31.4,38,40.3,0,48.2,38.2,37.8,32.9,29,32.4,37.1,54.1,45.5,52.4,52.8,22.6,42.7,33.2,26.6,21.2,56.5,48.4,42.7,38.6,23.6,20.5,45.8,34.7,0,0,0
|
19 |
-
[InternLM-XComposer-VL](https://github.com/InternLM/InternLM-XComposer),InternLM-7B,7B,PPL,64.2,52.2,42.8,59.2,32.1,0,74.8,70.5,67.6,60.5,55.3,53.4,76.3,76.1,61.4,86.1,78,27.2,60.3,84.8,68.9,25.8,47.7,56.6,58.6,49.9,37.6,24.9,27.5,36.7,0,0,0
|
20 |
-
[SPHINXv1-1k](https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/SPHINX),LLaMA-2-13B,13B,Generate,63,39.7,38.4,56.4,31.9,0,75.5,72.4,75.1,63.1,67.6,50.4,64.9,76.7,60,81.5,82.4,21.8,60.3,58.5,65.5,32.6,35.9,43.4,52.4,41.2,33.9,26.1,33.3,30.6,0,0,0
|
21 |
-
[SPHINXv2-1k](https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/SPHINX),LLaMA-2-13B,13B,Generate,66.7,43.9,42.6,60.2,37,0,77.5,78.5,76.6,69,71,57.5,73.2,77.6,62.1,82.7,85.2,44.5,62.3,60.3,65.3,23.5,45.7,42.1,54.6,48.1,37.9,29.9,33.3,40.8,0,0,0
|
22 |
-
[Emu](https://github.com/baaivision/Emu),LLaMA-13B,13B,PPL,45.3,33.2,36.1,42.5,41.1,41.4,59,50,43.7,37.1,44.3,33.6,49.5,58.3,61.4,68.8,61.6,19,45.7,41.5,24.2,26.4,29.3,37.1,41.9,42.7,37.9,21.8,51.7,30.6,46.8,43.2,34.2
|
23 |
-
[Next-GPT](https://github.com/NExT-GPT/NExT-GPT),vicuna-7B,7B,PPL,30.8,31.6,29.9,30.7,35.6,33.9,36.4,35.1,25.6,29.9,36.1,30.9,39.2,41.7,31,30.9,27.4,21.2,34.2,31.8,24.4,17.4,24.2,39,35.5,33.8,25.6,24.5,46.7,24.5,45.1,19.8,36.7
|
24 |
-
[SEED-LLaMA](https://github.com/AILab-CVC/SEED),LLaMA2-Chat-13B,13B,PPL,46.5,35.5,37.6,43.9,43.4,52.3,64,55,51.3,45.4,43.3,37.9,56.7,59.2,57,55.5,52.8,18.8,49.3,44.8,28.8,24.4,29.5,41.5,46.7,39.4,43.9,20.3,54.2,32.7,50.2,40.7,65.8
|
25 |
-
[GPT-4V](https://openai.com/research/gpt-4v-system-card),\-,-,Generate,70,78.6,61.3,69.2,44.2,0,77.5,73.9,70.6,61.8,56.8,56.9,74.2,78.5,82.3,91.8,97.4,45.1,71.9,66.1,71.1,43.9,67.9,89.3,64.5,65.7,51.7,63.4,29.2,59.2,0,0,0
|
26 |
-
[VideoChat](https://github.com/OpenGVLab/Ask-Anything),Vicuna-7B,7B,PPL,37.4,41,33.4,37,35.3,0,44.3,40.7,32.2,36.9,32.9,32.6,42.3,51.1,45.7,35.2,46.8,20.6,43.2,39.4,34.3,19.7,30.3,51.6,41.5,34,30.6,27.4,40,30.6,0,0,0
|
27 |
-
[Video-ChatGPT](https://github.com/mbzuai-oryx/Video-ChatGPT),LLaMA-7B,7B,PPL,35.8,53.8,30.2,36.4,31,0,44.1,37,35.8,30.7,44.2,31.1,29.9,49.9,39.8,49.7,40.6,22,33.2,37.2,22.4,25,46.1,61.4,42.6,32.2,27,19,37.5,24.5,0,0,0
|
28 |
-
[Valley](https://github.com/RupertLuo/Valley),LLaMA-13B,13B,PPL,34.9,44.7,28,34.5,32.2,0,45.3,36.4,33.7,30.6,27.1,31.5,35.1,52,35.2,44.9,43.4,23.8,33.2,37.2,26,22.7,37.1,52.2,31.5,32.1,21.9,26.5,35.8,28.6,0,0,0
|
|
|
1 |
Model,Language Model,Model Size,Evaluation Method,Avg. Single,Avg. Multi,Avg. Video,Avg. P1,Avg. P2,Avg. P3,Scene Understanding,Instance Identity,Instance Attribute,Instance Location,Instance Counting,Spatial Relation,Instance Interaction,Visual Reasoning,Text Recognition,Celebrity Recognition,Landmark Recognition,Chart Understanding,Visual Referring Expression,Science Knowledge,Emotion Recognition,Visual Mathematics,Difference Spotting,Meme Comprehension,Global Video Understanding,Action Recognition,Action Predicion,Procedure Understanding,In-Context Captioning,Interleaved Image-Text Analysis,Text-to-Image Generation,Next Image Prediction,Text-Image Creation
|
2 |
+
[BLIP-2](https://github.com/salesforce/LAVIS),Flan-T5-XL,3B,PPL,44.1,28.2,34.8,41.0,35.3,0.0,58.5,48.6,49.0,39.1,43.4,36.2,48.5,52.9,60.7,51.8,51.4,19.2,43.2,52.4,29.3,22.0,17.8,38.6,42.5,37.7,36.2,22.9,40.0,30.6,0.0,0.0,0.0
|
3 |
+
[InstructBLIP](https://github.com/salesforce/LAVIS),Flan-T5-XL,3B,PPL,45.5,29.0,35.7,42.2,35.7,0.0,58.9,49.7,61.7,35.1,58.1,34.9,47.4,55.9,61.4,48.5,45.4,26.4,41.7,47.7,34.5,21.2,22.8,35.2,41.5,36.1,40.5,24.5,36.7,34.7,0.0,0.0,0.0
|
4 |
+
[InstructBLIP-Vicuna](https://github.com/salesforce/LAVIS),Vicuna-7B,7B,PPL,43.4,46.0,31.5,41.4,29.7,0.0,53.6,43.9,49.0,37.8,56.5,35.8,43.3,56.2,57.2,60.3,44.4,27.9,39.2,39.4,23.0,26.5,36.5,55.4,40.4,38.6,31.2,15.6,26.7,32.7,0.0,0.0,0.0
|
5 |
+
[LLaVA](https://github.com/haotian-liu/LLaVA),LLaMA-7B,7B,PPL,40.6,38.5,31.0,38.7,30.2,0.0,53.8,47.5,38.3,34.2,42.0,34.7,40.2,52.9,46.4,51.8,45.6,30.3,40.2,37.6,34.3,20.5,27.0,50.0,44.1,36.2,25.1,18.6,40.0,20.4,0.0,0.0,0.0
|
6 |
+
[MiniGPT-4](https://github.com/Vision-CAIR/MiniGPT-4),Vicuna-7B,7B,PPL,41.7,32.9,33.4,39.4,34.1,0.0,56.3,49.2,45.8,37.9,45.3,32.6,47.4,57.1,41.8,55.2,45.2,20.2,41.2,43.3,24.2,25.0,19.0,46.7,39.0,38.7,27.4,28.6,45.8,22.5,0.0,0.0,0.0
|
7 |
+
[VPGTrans](https://github.com/VPGTrans/VPGTrans),LLaMA-7B,7B,PPL,37.4,34.3,32.6,36.2,23.9,0.0,46.9,38.6,33.6,35.6,27.5,34.4,33.0,50.8,47.6,52.4,38.2,30.1,34.7,36.1,31.5,27.3,24.6,44.0,37.8,38.2,20.9,33.5,19.2,28.6,0.0,0.0,0.0
|
8 |
+
[MultiModal-GPT](https://github.com/open-mmlab/Multimodal-GPT),LLaMA-7B,7B,PPL,37.5,48.3,31.5,37.4,34.9,0.0,46.9,42.5,32.0,32.3,27.7,29.7,29.9,48.3,35.2,60.9,50.4,24.2,42.2,37.6,32.1,27.3,40.1,56.5,37.6,38.7,25.3,24.4,39.2,30.6,0.0,0.0,0.0
|
9 |
+
[Otter](https://github.com/Luodian/Otter),LLaMA-7B,7B,PPL,37.6,37.1,31.3,36.4,36.6,0.0,45.9,39.7,31.9,31.6,26.4,32.0,33.0,49.2,39.3,59.7,53.0,23.6,41.2,36.1,37.3,22.0,27.4,46.7,36.6,37.9,26.0,24.8,42.5,30.6,0.0,0.0,0.0
|
10 |
+
[OpenFlamingo](https://github.com/mlfoundations/open_flamingo),LLaMA-7B,7B,PPL,37.5,47.4,31.3,37.3,35.5,0.0,46.7,42.3,31.7,33.4,27.4,29.8,29.9,47.7,35.6,60.3,49.8,24.2,42.2,39.0,32.1,27.3,39.9,54.9,37.6,38.4,25.2,24.1,38.3,32.7,0.0,0.0,0.0
|
11 |
+
[LLaMA-AdapterV2](https://github.com/OpenGVLab/LLaMA-Adapter),LLaMA-7B,7B,PPL,39.0,40.7,29.8,37.5,0.0,0.0,45.2,38.5,29.3,33.0,29.7,35.5,39.2,52.0,48.7,58.5,46.4,24.2,41.2,40.1,39.7,23.5,29.1,52.2,41.9,38.2,18.8,20.3,0.0,0.0,0.0,0.0,0.0
|
12 |
+
[GVT](https://github.com/TencentARC/GVT),Vicuna-7B,7B,PPL,33.5,50.4,30.1,34.4,38.6,0.0,41.7,35.5,31.8,29.5,36.2,32.0,32.0,51.1,35.2,39.4,36.4,25.0,36.2,31.1,20.6,22.7,41.5,59.2,40.4,29.7,26.3,24.1,42.5,34.7,0.0,0.0,0.0
|
13 |
+
[mPLUG-Owl](https://github.com/X-PLUG/mPLUG-Owl),LLaMA-7B,7B,PPL,41.3,44.2,29.4,39.4,28.9,0.0,49.7,45.3,32.5,36.7,27.3,32.7,44.3,54.7,49.2,70.9,49.6,23.2,44.2,44.0,32.5,23.5,33.5,54.9,42.0,37.8,18.3,19.3,29.2,28.6,0.0,0.0,0.0
|
14 |
+
[Kosmos-2](https://github.com/microsoft/unilm/tree/master/kosmos-2),Decoder only 1.3B,1.3B,PPL,49.5,34.0,39.7,46.3,23.3,0.0,63.4,57.1,58.5,44.0,41.4,37.9,55.7,60.7,68.1,82.1,51.4,21.2,48.2,43.7,30.7,28.0,25.2,42.8,48.5,40.8,39.5,30.0,24.2,22.5,0.0,0.0,0.0
|
15 |
+
[Qwen-VL-Chat](https://huggingface.co/Qwen/Qwen-VL-Chat),Qwen-7B,7B,PPL,46.0,40.8,32.8,43.1,35.5,0.0,56.5,47.6,54.8,46.9,54.2,40.3,55.7,55.0,47.4,62.4,55.6,25.2,43.7,41.2,20.6,28.8,34.3,47.2,39.7,42.8,29.6,19.1,42.5,28.6,0.0,0.0,0.0
|
16 |
+
[Qwen-VL-plus](https://github.com/QwenLM/Qwen-VL/tree/master?tab=readme-ov-file#qwen-vl-plus),Qwen-LM,-,PPL for A/B/C/D,69.0,74.3,45.9,65.3,30.5,0.0,76.6,77.7,76.3,65.1,65.8,55.9,73.2,77.9,61.8,97.0,97.2,39.5,73.4,75.8,51.7,38.6,66.7,81.8,51.8,54.5,29.3,48.0,28.3,32.7,0.0,0.0,0.0
|
17 |
+
[LLaVA-1.5](https://github.com/haotian-liu/LLaVA),vicuna-7B,7B,PPL,50.8,43.0,35.8,47.3,30.8,0.0,63.7,62.4,66.7,51.3,60.2,38.5,47.4,59.8,69.0,60.6,49.8,25.0,45.7,56.7,31.1,24.2,35.7,50.3,46.1,39.4,29.4,28.1,39.2,22.5,0.0,0.0,0.0
|
18 |
+
[IDEFICS-9b-instruct](https://huggingface.co/HuggingFaceM4/idefics-9b-instruct),LLaMA-7B,7B,PPL,37.9,52.5,31.4,38.0,40.3,0.0,48.2,38.2,37.8,32.9,29.0,32.4,37.1,54.1,45.5,52.4,52.8,22.6,42.7,33.2,26.6,21.2,56.5,48.4,42.7,38.6,23.6,20.5,45.8,34.7,0.0,0.0,0.0
|
19 |
+
[InternLM-XComposer-VL](https://github.com/InternLM/InternLM-XComposer),InternLM-7B,7B,PPL,64.2,52.2,42.8,59.2,32.1,0.0,74.8,70.5,67.6,60.5,55.3,53.4,76.3,76.1,61.4,86.1,78.0,27.2,60.3,84.8,68.9,25.8,47.7,56.6,58.6,49.9,37.6,24.9,27.5,36.7,0.0,0.0,0.0
|
20 |
+
[SPHINXv1-1k](https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/SPHINX),LLaMA-2-13B,13B,Generate,63.0,39.7,38.4,56.4,31.9,0.0,75.5,72.4,75.1,63.1,67.6,50.4,64.9,76.7,60.0,81.5,82.4,21.8,60.3,58.5,65.5,32.6,35.9,43.4,52.4,41.2,33.9,26.1,33.3,30.6,0.0,0.0,0.0
|
21 |
+
[SPHINXv2-1k](https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/SPHINX),LLaMA-2-13B,13B,Generate,66.7,43.9,42.6,60.2,37.0,0.0,77.5,78.5,76.6,69.0,71.0,57.5,73.2,77.6,62.1,82.7,85.2,44.5,62.3,60.3,65.3,23.5,45.7,42.1,54.6,48.1,37.9,29.9,33.3,40.8,0.0,0.0,0.0
|
22 |
+
[Emu](https://github.com/baaivision/Emu),LLaMA-13B,13B,PPL,45.3,33.2,36.1,42.5,41.1,41.4,59.0,50.0,43.7,37.1,44.3,33.6,49.5,58.3,61.4,68.8,61.6,19.0,45.7,41.5,24.2,26.4,29.3,37.1,41.9,42.7,37.9,21.8,51.7,30.6,46.8,43.2,34.2
|
23 |
+
[Next-GPT](https://github.com/NExT-GPT/NExT-GPT),vicuna-7B,7B,PPL,30.8,31.6,29.9,30.7,35.6,33.9,36.4,35.1,25.6,29.9,36.1,30.9,39.2,41.7,31.0,30.9,27.4,21.2,34.2,31.8,24.4,17.4,24.2,39.0,35.5,33.8,25.6,24.5,46.7,24.5,45.1,19.8,36.7
|
24 |
+
[SEED-LLaMA](https://github.com/AILab-CVC/SEED),LLaMA2-Chat-13B,13B,PPL,46.5,35.5,37.6,43.9,43.4,52.3,64.0,55.0,51.3,45.4,43.3,37.9,56.7,59.2,57.0,55.5,52.8,18.8,49.3,44.8,28.8,24.4,29.5,41.5,46.7,39.4,43.9,20.3,54.2,32.7,50.2,40.7,65.8
|
25 |
+
[GPT-4V](https://openai.com/research/gpt-4v-system-card),\-,-,Generate,70.0,78.6,61.3,69.2,44.2,0.0,77.5,73.9,70.6,61.8,56.8,56.9,74.2,78.5,82.3,91.8,97.4,45.1,71.9,66.1,71.1,43.9,67.9,89.3,64.5,65.7,51.7,63.4,29.2,59.2,0.0,0.0,0.0
|
26 |
+
[VideoChat](https://github.com/OpenGVLab/Ask-Anything),Vicuna-7B,7B,PPL,37.4,41.0,33.4,37.0,35.3,0.0,44.3,40.7,32.2,36.9,32.9,32.6,42.3,51.1,45.7,35.2,46.8,20.6,43.2,39.4,34.3,19.7,30.3,51.6,41.5,34.0,30.6,27.4,40.0,30.6,0.0,0.0,0.0
|
27 |
+
[Video-ChatGPT](https://github.com/mbzuai-oryx/Video-ChatGPT),LLaMA-7B,7B,PPL,35.8,53.8,30.2,36.4,31.0,0.0,44.1,37.0,35.8,30.7,44.2,31.1,29.9,49.9,39.8,49.7,40.6,22.0,33.2,37.2,22.4,25.0,46.1,61.4,42.6,32.2,27.0,19.0,37.5,24.5,0.0,0.0,0.0
|
28 |
+
[Valley](https://github.com/RupertLuo/Valley),LLaMA-13B,13B,PPL,34.9,44.7,28.0,34.5,32.2,0.0,45.3,36.4,33.7,30.6,27.1,31.5,35.1,52.0,35.2,44.9,43.4,23.8,33.2,37.2,26.0,22.7,37.1,52.2,31.5,32.1,21.9,26.5,35.8,28.6,0.0,0.0,0.0
|