kabachuha/Qwen3-4B-Instruct-2507-SOMbliterated
TEXT GENERATIONConcurrency Cost:1Model Size:4BQuant:BF16Ctx Length:32kPublished:Feb 28, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

kabachuha/Qwen3-4B-Instruct-2507-SOMbliterated is a 4 billion parameter instruction-tuned causal language model, based on Qwen/Qwen3-4B-Instruct-2507, developed by kabachuha. This model has been 'SOMbliterated' using a multi-directional ablation method based on Self-Organizing Maps to significantly reduce refusal rates while minimizing damage to original capabilities. It is specifically designed for use cases requiring a less censored model, demonstrating superior performance in refusal reduction and lower KL divergence compared to other decensoring methods.

Loading preview...