danishnaseer00

Reproducibility artifact for a study of how post-training quantization changes small language models' responses to misleading contextual cues. Frozen trial set, 15 generation files, 3,078 automated judgments, human annotation audit, and full analysis pipeline.