Taleef Tamsal

I am a Master’s student in Computer Science at Purdue University Fort Wayne, working on trustworthy and robust NLP under the supervision of Dr. Jonathan Rusert. My research focuses on LLM safety and alignment, adversarial robustness for efficiency-tuned models (e.g., PEFT + quantization), and evaluation methods that surface brittle behavior in realistic settings.

Research Snapshot

Broadly, I study how to make language systems reliable when the constraints are real: limited compute, imperfect data, fast-changing deployments, and adversarial pressure. I care about research that transfers into practice, where evaluation is not a final checkbox, but part of the system design.

Current Work

• Master’s Thesis: Robustness–efficiency trade-offs in instruction-tuned LLMs, focusing on how LoRA/QLoRA and quantization can change jailbreak susceptibility under modern black-box attack families.

• SemEval 2026: Participating in Task 6 (CLARITY - political question evasion detection) and Task 8 (MTRAGEval - multi-turn RAG evaluation).