Dharunish Yugeswardeenoo, a high school student, has achieved acceptance to ACL SRW 2024, a premier venue typically reserved for advanced doctoral and master's candidates. This accomplishment represents exceptional recognition within the computational linguistics field.
The accepted work introduces Question-Analysis-Prompting (QAP), a novel method designed to enhance large language model performance on reasoning tasks. The team evaluated QAP across: - Arithmetic datasets: GSM8K, AQuA, and SAT - Commonsense dataset: StrategyQA - Models tested: GPT-3.5 and GPT-4
The research demonstrates that QAP outperforms all state-of-the-art prompts on AQuA and SAT datasets on both GPT-3.5 and GPT-4. Comparative analysis included Chain-of-Thought, Plan and Solve Prompting, and Take A Deep Breath methodologies.
Paper: https://arxiv.org/pdf/2407.03624
The presentation took place in August 2024 in Bangkok, Thailand.
