<aside>
📌
Summary
During my ETRI internship, I enhanced AI-driven fact-checking methodologies by optimizing prompt engineering, refining fact validation pipelines, and developing dual-layer evaluation metrics. These improvements boosted AI fact-checking accuracy from 84% to 94%, demonstrating the impact of structured reasoning and NLP-driven optimizations. 🚀
</aside>
📅 Internship Period: 2024.07 - 2024.08
🏢 Affiliation: Language Intelligence Lab, ETRI
💼 Role: Undergraduate Research Intern – Fact Verification & Prompt Optimization
🎯 Key Responsibilities & Achievements
$\color{gray}\rule{361px}{1.5px}$
1️⃣ Atomic Fact Validation for AI-Based Fact-Checking
- Developed an AI fact verification pipeline to assess the factual correctness of segmented sentences.
- Curated and refined datasets, ensuring the removal of uninformative or incomplete data.
- Implemented ChatGPT-based AI fact-checking workflows, improving factual validation accuracy.
📌 Key Impact:
- Improved AI fact-checking accuracy by ensuring data integrity and structured fact verification.
- Enhanced AI reasoning models, reducing dependence on irrelevant context.
2️⃣ Prompt Engineering for AI Performance Optimization
- Designed & tested multiple prompt structures (e.g., question-based vs. explanation-based formats).
- Conducted iterative tuning experiments to improve AI’s ability to differentiate true vs. false claims.