Analogical reasoning is fundamental to human cognition and holds an important place in various fields. Thus, we introduce the multimodal analogical reasoning task, which can be formulate as link prediction without explicitly providing relations. To support it, we collect an Multimodal Analogical Reasoning dataSet (MARS) and a multimodal knowledge graph (MarKG). Among them, MARS has 10,685 training, 1,228 validation and 1,415 test instances. MarKG contains 11,292 entities, 192 relations and 76,424 images, include 2,063 analogy entities and 27 analogy relations.
For more tails about the task and the datasets, please refer to our ICLR 2023 paper:
MarKG and MARS are distributed under CC BY-SA 4.0 license , download the textual data of them by following links:
The image data (18.72G) of MARS and MarKG can be be downloaded through the Google Drive or the Baidu Yun(TeraBox)(code:7hoc) .
We provide the best checkpoints of our transformer-based models during the fine-tuning and pre-training phrases at this link.
For more introduction and guidence, please check out our Github repository.
If you have any question about the multimodal analogical reasoning task and the datasets MARS and MarKG, or you want to submit your results to update the leaderboard, feel free to emtail to us:
If you use or extend our work, please cite the paper as follows:
@inproceedings{zhang2023multimodal, title={Multimodal Analogical Reasoning over Knowledge Graphs}, author={Ningyu Zhang and Lei Li and Xiang Chen and Xiaozhuan Liang and Shumin Deng and Huajun Chen}, booktitle={The Eleventh International Conference on Learning Representations }, year={2023}, url={https://openreview.net/forum?id=NRHajbzg8y0P} }
Rank | Model | Hit@1 | Hit@3 | Hit@10 | MRR |
---|---|---|---|---|---|
1 Oct 01, 2022 |
MKGformer (MART) (SIGIR '22) |
0.301 | 0.367 | 0.408 | 0.341 |
2 Oct 01, 2022 |
ViLBERT (MART) (NeurIPS '19) |
0.256 | 0.312 | 0.347 | 0.292 |
3 Oct 01, 2022 |
FLAVA (MART) (CVPR '22) |
0.264 | 0.303 | 0.319 | 0.288 |
4 Oct 01, 2022 |
RSME (ANALOGY) (ACM MM '21) |
0.266 | 0.298 | 0.311 | 0.285 |
5 Oct 01, 2022 |
VisualBERT (MART) (arxiv '19) |
0.261 | 0.292 | 0.321 | 0.284 |
6 Oct 01, 2022 |
IKRL (ANALOGY) (IJCAI '17) |
0.266 | 0.294 | 0.310 | 0.283 |
7 Oct 01, 2022 |
TransAE (ANALOGY) (IJCNN '19') |
0.261 | 0.285 | 0.293 | 0.276 |
8 Oct 01, 2022 |
ViLT (MART) (ICML '21) |
0.245 | 0.275 | 0.303 | 0.266 |