diff --git a/literature_review/papers/SQL_PALM.md b/literature_review/papers/SQL_PALM.md new file mode 100644 index 0000000..70261c0 --- /dev/null +++ b/literature_review/papers/SQL_PALM.md @@ -0,0 +1,17 @@ +To contribute kindly follow the following framework: +1. Name of the paper +* [SQL-PALM: IMPROVED LARGE LANGUAGE +MODELADAPTATION FOR TEXT-TO-SQL](https://arxiv.org/pdf/2306.00739.pdf) +2. The datatset Used +* [Spider](https://yale-lily.github.io/spider) +3. The model Used +* PaLM-2 and it's variations +4. Novelty introduced +* Fine-tuned SQL-PaLM +5. How it can be integrated in the project +* (To be reviewed and discussed) +6. Additional information you believe will be helpful +* Fine-tuned SQL-PaLM adapts large-sized LLMs (PaLM-2) to domain-specific Text-to-SQL data through fine-tuning on the Spider train split, The fine tuned model further acheives better result than the few-shot model +* Section 3 provides a good summary of the different kinds of datasets and models employed to create a baseline testing enviornmnet +* **The fine-tuned SQL-PaLM model achieves a test-suite accuracy of 78.2% on the Spider dataset. On the other hand, ChatGPT, evaluated with the OpenAI-default prompt, achieves a test-suite accuracy of 60.1%.** +* The further sections talk about the real life usage of the said models and the future scope they provide and what more work can be done on them \ No newline at end of file