add qqmm (#33)

- add qqmm (aa65c27b94fd6688e121c037a24bc77a56487954)
- fixed (9198f1f4c54b6c885c3cec4de0a25f82ce27866b)
- merged main (9ee47573376122eedde0144b17153ad7367573de)

results.csv

@@ -26,4 +26,5 @@ LLaVE-2B,1.95,Self-Reported,65.2,62.1,60.2,65.2,84.9
 LLaVE-0.5B,0.894,Self-Reported,59.1,57.4,50.3,59.8,82.9
 UniME(LLaVA-OneVision-7B-LoRA-Res336),8.03,Self-Reported,70.7,66.8,66.6,70.5,90.9
 UniME(LLaVA-1.6-7B-LoRA-LowRes),7.57,Self-Reported,66.6,60.6,52.9,67.9,85.1
-UniME(Phi-3.5-V-LoRA),4.2,Self-Reported,64.2,54.8,55.9,64.5,81.8
+UniME(Phi-3.5-V-LoRA),4.2,Self-Reported,64.2,54.8,55.9,64.5,81.8
+QQMM-embed,8.297,Self-Reported,72.175,70.07,69.52,71.175,87.075

utils.py

@@ -40,9 +40,9 @@ All tasks are reformulated as ranking tasks, where the model follows instruction
 or a combination of both. MMEB is divided into 20 in-distribution datasets, which can be used for
 training, and 16 out-of-distribution datasets, reserved for evaluation.
 
-The detailed explanation of the benchmark and datasets can be found in our paper: https://doi.org/10.48550/arXiv.2410.05160.
-Github link: https://github.com/TIGER-AI-Lab/VLM2Vec
-Overview: https://tiger-ai-lab.github.io/VLM2Vec/
+The detailed explanation of the benchmark and datasets can be found in our paper: https://doi.org/10.48550/arXiv.2410.05160. \n
+Github link: https://github.com/TIGER-AI-Lab/VLM2Vec. \n
+Overview: https://tiger-ai-lab.github.io/VLM2Vec/. \n
 """
 
 TABLE_INTRODUCTION = """"""
@@ -97,7 +97,7 @@ SUBMIT_INTRODUCTION = """# Submit on MMEB Leaderboard Introduction
 ]
 ```
 You may refer to the Github page for instructions about evaluating your model. 
-Github link: https://github.com/TIGER-AI-Lab/VLM2Vec
+Github link: https://github.com/TIGER-AI-Lab/VLM2Vec. \n
 Please send us an email at [email protected], attaching the JSON file. We will review your submission and update the leaderboard accordingly.
 """