bedderautomation/

refusal-geometry-qwen25-3b

Direct measurement of refusal direction geometry extracted via OBLITERATUS analysis pipeline. Model: Qwen/Qwen2.5-3B-Instruct (3.09B params, 36 layers, hidden_dim=2048) Probing: 512 harmful + 512 harmless prompt pairs with chat template wrapping Date: 2026-03-10

	Key Findings




	


	1. Refusal Layer Distribution

Refusal concentrates in the final 11 layers (25-35). Layer 35 is the decision point.

Layer Magnitude… See the full description on the dataset page: https://huggingface.co/datasets/bedderautomation/refusal-geometry-qwen25-3b.