Undergraduate Research
Feni University · Sep 2025 – Mar 2026
Thesis: An Empirical Analysis of Hallucination Evaluation Metrics in Bangla using LLMs
- Designed and executed an empirical evaluation framework to assess the logical and semantic stability of Generative AI models including Gemini 2.5 Flash, Gemma 3, and LLaMA 3.2. Investigated hallucination behavior and model reliability using Natural Language Inference (NLI), Named Entity Recognition (NER), and consistency-based evaluation on the BanglaQA dataset. Focused on trustworthiness in low-resource language settings and contributed toward building more reliable and interpretable LLM evaluation baselines. Led a 3-member undergraduate thesis team, coordinated methodology design, experimental workflow, and technical reporting with faculty supervision.