askalgore/Qwen3-4B-Instruct-2507-heretic: Decensored Qwen3-4B-Instruct

This model is a 4 billion parameter instruction-tuned causal language model, a decensored version of the original Qwen/Qwen3-4B-Instruct-2507. It was created using the Heretic v1.0.1 tool, specifically targeting the reduction of content refusals.

Key Capabilities

• Decensored Output: Achieves 0/100 refusals on a test set, a significant reduction from the original model's 99/100 refusals.
• Enhanced General Abilities: Demonstrates significant improvements in instruction following, logical reasoning, text comprehension, mathematics, science, coding, and tool usage.
• Long-Context Understanding: Supports a native context length of 262,144 tokens, with a recommended output length of 16,384 tokens for most queries.
• Improved Alignment: Shows markedly better alignment with user preferences in subjective and open-ended tasks, leading to more helpful responses and higher-quality text generation.
• Multilingual Knowledge: Offers substantial gains in long-tail knowledge coverage across multiple languages.

Good for

• Applications requiring a less restrictive content policy.
• Complex tasks demanding strong logical reasoning and instruction following.
• Scenarios benefiting from extensive context understanding, such as document analysis or long-form content generation.
• Creative writing and open-ended conversational agents where subjective responses are desired.
• Tool-calling and agentic workflows, especially when integrated with frameworks like Qwen-Agent.