Tag
This paper investigates perturbation-based data selection methods (RSP and RSP-Drop) for adapting medical language models, showing improved Macro-F1 scores over random selection when evaluated with MedGemma-4B on five medical QA datasets.
PROOF is a profile-oriented benchmark that evaluates factual reliability in large language models using Wikidata-based multiple-choice questions with controlled interventions, revealing domain-specific coverage and stability under perturbations.
MAWILE is a developer workbench for auditing the sensitivity of LLM judges to perturbations across prompts, rubrics, inputs, and outputs, ensuring robust and meaningful evaluations.
This paper conducts a multi-level analysis of how input perturbations propagate through decoder-only language models, assessing robustness via output behavior, hidden-state geometry, and attention-head function across models like GPT-2 and Qwen2.5.