qwen35-27b IPD Str-TFT Deontological Native-Tool (agentic-moral-alignment)

An open-weight fine-tuned checkpoint of Qwen3.5-27B, adapted for agentic behavior under deontological moral-alignment objectives and native tool use. The model is trained around Iterated Prisoner's Dilemma-style strategy signals (Str-TFT) to encourage cooperative, rule-following agent behavior rather than purely capability-maximizing outputs. Source: supabase.