Publications

Parametric Stubbornness: Mechanistically Isolating the Layer Shift and Sparsity Gradient of RAG Knowledge Conflicts in Llama-3

Published in Preprint, Zenodo, 2026

A mechanistic study of hallucination dynamics in retrieval-augmented generation and LLM knowledge conflict settings.

Recommended citation: Rajabi, H. (2026). "Parametric Stubbornness: Mechanistically Isolating the Layer Shift and Sparsity Gradient of RAG Knowledge Conflicts in Llama-3." Preprint, Zenodo.
Download Paper

Identity, Toxicity, or Complexity? A Language-Specific Feature Selection Approach to Reclamatory Intent Detection

Published in Proceedings of EVALITA 2026 (Task A: MultiPRIDE), Bari, Italy, 2026

A language-specific feature selection approach for detecting reclaimed slurs across Italian, English, and Spanish.

Recommended citation: Rajabi, H., Ghavidel, F., & Ghahremani, K. (2026). "Identity, Toxicity, or Complexity? A Language-Specific Feature Selection Approach to Reclamatory Intent Detection." Proceedings of EVALITA 2026 (Task A: MultiPRIDE), Bari, Italy.
Download Paper

Reclamation-Aware Augmentation: A Community-in-the-Loop Pipeline for Generating Contrastive Slur-Reclamation Data

Published in Manuscript in preparation, 2026

A five-stage multi-model pipeline for generating contrastive minimal pairs for pragmatically sensitive classification, introducing MPEC scoring validated against community annotators.