Congratulations to one of our students, Dharunish Yugeswardeenoo, for his acceptance to ACL SRW 2024! This is an absolutely incredible achievement, let alone as a high school student. ACL SRW is intended for junior and senior PhD students, and open to advanced masters/undergraduates. For context, ACL, or the Association for Computational Linguistics, is the world's top NLP conference.
In our research, we propose a novel prompting method, Question-Analysis-Prompting (QAP), to improve performance for LLMs on reasoning tasks. We evaluate our method on GPT-class models on arithmetic datasets GSM8K, AQuA, and SAT, and commonsense dataset StrategyQA. We compare QAP with other state-of-the-art prompts including Chain-of-Thought (CoT), Plan and Solve Prompting (PS+), and Take A Deep Breath (TADB). QAP outperforms all state-of-the-art prompts on AQuA and SAT datasets on both GPT3.5 and GPT4. QAP consistently ranks among the top-2 prompts on 75% of the tests. We conduct a thorough analysis on our findings.
For more details, check out our paper: Read the paper here.
We are excited about the possibility of Dharunish presenting his work at the 2024 conference in August, which will be held in Bangkok, Thailand. Whether or not he decides to present, we couldn't be more proud of his achievements and look forward to his continued contributions to the field of AI and NLP!