leonardPKU commited on
Commit
eb089a3
1 Parent(s): 82e9bf0

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +68 -13
README.md CHANGED
@@ -121,15 +121,23 @@ language:
121
  pretty_name: PCA-Bench
122
  ---
123
 
 
124
  <h1 align="center">PCA-Bench</h1>
125
 
126
  <p align="center">
127
 
128
  <a href="https://github.com/pkunlp-icler/PCA-EVAL">
129
  <img alt="Static Badge" src="https://img.shields.io/badge/Github-Online-white">
130
- </a><a href="https://arxiv.org/abs/2310.02071">
131
- <img alt="Static Badge" src="https://img.shields.io/badge/Paper-PCAEVAL-red">
132
- <a href="https://huggingface.co/datasets/PCA-Bench/PCA-Bench-V1"><img alt="Static Badge" src="https://img.shields.io/badge/Datasets-HuggingFace-yellow">
 
 
 
 
 
 
 
133
  </a>
134
  </p>
135
 
@@ -139,28 +147,75 @@ pretty_name: PCA-Bench
139
  *PCA-Bench is an innovative benchmark for evaluating and locating errors in Multimodal LLMs when conducting embodied decision making tasks, specifically focusing on perception, cognition, and action.*
140
 
141
 
142
- ## News
143
- - PCA-Bench-V1 is released in HuggingFace Datasets (Leaderboard Coming Soon).
 
144
 
 
 
145
 
146
 
147
- ## Run Evaluation on Accuracy
148
 
 
 
 
 
 
 
 
149
  ```python
150
- #pip install datasets
151
  from datasets import load_dataset
 
 
 
 
 
 
 
 
 
 
 
 
 
152
  dataset_ad = load_dataset("PCA-Bench/PCA-Bench-V1","Autonomous Driving")
153
  dataset_dr = load_dataset("PCA-Bench/PCA-Bench-V1","Domestic Robot")
154
  dataset_og = load_dataset("PCA-Bench/PCA-Bench-V1","Open-World Game")
155
 
156
- # use your model to inference on the test_open/close split with "question_prompt" and "image" given in the datasets and extract the answers.
157
- # compute the acc regarding the action groundtruth of open track.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
158
 
 
159
  ```
160
 
161
- 📢 For close track data, please follow [this file](https://github.com/pkunlp-icler/PCA-EVAL/blob/main/pca-eval/results/chatgpt_holmes_outputs/Autonomous%20Driving.json) to organize your model output. Submit **three JSON files** from different domains, along with your **model name** and **organization** to us via [email](mailto:leo.liang.chen@stu.pku.edu.cn). Ensure you use the dataset's provided prompt as the default input for fair comparison (You should mention in the email if custom prompts were used).
162
 
163
- We will send the closed track PCA-Eval results of your model to you.
164
 
165
- ## PCA Evaluation
166
- To run PCA-Evaluation yourself, please follow the guidelines in the github repo [PCA-EVAL](https://github.com/pkunlp-icler/PCA-EVAL).
 
121
  pretty_name: PCA-Bench
122
  ---
123
 
124
+
125
  <h1 align="center">PCA-Bench</h1>
126
 
127
  <p align="center">
128
 
129
  <a href="https://github.com/pkunlp-icler/PCA-EVAL">
130
  <img alt="Static Badge" src="https://img.shields.io/badge/Github-Online-white">
131
+
132
+ <a href="https://github.com/pkunlp-icler/PCA-EVAL/blob/main/PCA_Bench_Paper.pdf">
133
+ <img alt="Static Badge" src="https://img.shields.io/badge/Paper-PCABench-red">
134
+
135
+ <a href="https://huggingface.co/datasets/PCA-Bench/PCA-Bench-V1">
136
+ <img alt="Static Badge" src="https://img.shields.io/badge/HFDataset-PCABenchV1-yellow">
137
+ </a>
138
+
139
+ <a href="https://docs.qq.com/sheet/DVUd4WUpGRHRqUnNV">
140
+ <img alt="Static Badge" src="https://img.shields.io/badge/Leaderboard-Online-blue">
141
  </a>
142
  </p>
143
 
 
147
  *PCA-Bench is an innovative benchmark for evaluating and locating errors in Multimodal LLMs when conducting embodied decision making tasks, specifically focusing on perception, cognition, and action.*
148
 
149
 
150
+ ## Release
151
+ - [2024.02.15] [PCA-Bench-V1](https://github.com/pkunlp-icler/PCA-EVAL) is released. We release the open and closed track data in [huggingface](https://huggingface.co/datasets/PCA-Bench/PCA-Bench-V1). We also set an online [leaderboard ](https://docs.qq.com/sheet/DVUd4WUpGRHRqUnNV) accepting users' submission.
152
+ - [2023.12.15] [PCA-EVAL](https://arxiv.org/abs/2310.02071) is accepted to Foundation Model for Decision Making Workshop @NeurIPS 2023. PCA-Evaluation tool is released in github.
153
 
154
+ ## Leaderboard
155
+ [Leaderboard with Full Metrics](https://docs.qq.com/sheet/DVUd4WUpGRHRqUnNV)
156
 
157
 
 
158
 
159
+ ## Submit Results
160
+
161
+ 📢 For close track evaluaiton and PCA-Evaluation, please follow [this file](https://github.com/pkunlp-icler/PCA-EVAL/blob/main/pca-eval/results/chatgpt_holmes_outputs/Autonomous%20Driving.json) to organize your model output. Submit **Six JSON files** from different domains and different tracks, along with your **model name** and **organization** to us via [email](mailto:leo.liang.chen@stu.pku.edu.cn). Ensure you use the dataset's provided prompt as the default input for fair comparison.
162
+
163
+ We will send the PCA-Eval results of your model to you and update the leaderboard.
164
+
165
+ We provide sample code to get the six json files. User only needs to add your model inference code:
166
  ```python
167
+ # Sample code for PCA-Eval
168
  from datasets import load_dataset
169
+ from tqdm import tqdm
170
+ import json
171
+ import os
172
+
173
+ def YOUR_INFERENCE_CODE(prompt,image):
174
+ """Simple single round multimodal conversation call.
175
+ """
176
+ response = YOUR_MODEL.inference(prompt,image)
177
+ return response
178
+
179
+ output_path = "./Results-DIR-PATH/"
180
+ os.mkdir(output_path)
181
+
182
  dataset_ad = load_dataset("PCA-Bench/PCA-Bench-V1","Autonomous Driving")
183
  dataset_dr = load_dataset("PCA-Bench/PCA-Bench-V1","Domestic Robot")
184
  dataset_og = load_dataset("PCA-Bench/PCA-Bench-V1","Open-World Game")
185
 
186
+ test_dataset_dict = {"Autonomous-Driving":dataset_ad,"Domestic-Robot":dataset_dr,"Open-World-Game":dataset_og}
187
+ test_split = ["test_closed","test_open"]
188
+ test_domain = list(test_dataset_dict.keys())
189
+
190
+ for domain in test_domain:
191
+ for split in test_split:
192
+ print("testing on %s:%s"%(domain,split))
193
+
194
+ prediction_results = []
195
+ output_filename = output_path+"%s-%s.json"%(domain,split)
196
+ prompts = test_dataset_dict[domain][split]['question_prompt']
197
+ images = test_dataset_dict[domain][split]['image']
198
+
199
+ for prompt_id in tqdm(range(len(prompts))):
200
+ user_inputs = prompts[prompt_id] # do not change the prompts for fair comparison
201
+ index = prompt_id
202
+ image = images[prompt_id]
203
+
204
+ outputs = YOUR_INFERENCE_CODE(user_inputs,image)
205
+
206
+ prediction_results.append({
207
+ 'prompt': user_inputs,
208
+ 'model_output': outputs,
209
+ 'index': index,
210
+ })
211
+
212
+ with open(output_filename, 'w') as f:
213
+ json.dump(prediction_results, f, indent=4)
214
 
215
+ # submit the 6 json files in the output_path to our email
216
  ```
217
 
218
+ You could also simply compute the multiple-choice accuracy locally as a comparison metric in your own experiments. However, in the online leaderboard, we only consider the average action score and Genuine PCA score when ranking models.
219
 
 
220
 
221
+ For more information, refer to the offical [github repo](https://github.com/pkunlp-icler/PCA-EVAL)