search

xlang-ai / UnifiedSKG

[EMNLP 2022] Unifying and multi-tasking structured knowledge grounding with language models
https://arxiv.org/abs/2201.05966
Apache License 2.0
550 stars 58 forks source link

NQ-Tables test set evaluation #39

Open yhifny opened 1 year ago

yhifny commented 1 year ago

Hi,

What is the best SKG-trained model to evaluate the natural questions tables (NQ-Tables) test set? Is it "from_all_T5_large_prefix_wikitq2"?

Thanks,

Yasser

Timothyxxx commented 1 year ago

Hi, sorry for the delay, I guess yes, the from_all_T5_large_prefix_wikitq2 is trained on all SKG tasks data and further trained on the wikitablequestions dataset, but if you can further train it if you have more data

Timothyxxx commented 1 year ago

For now, I think it is better to use bigger model, the study was done a few years ago, there are lots of better models for you to choose now, also optimized for structured knowledge grounding tasks