launch/MET-D-Qwen3-8B-en-only

TEXT GENERATIONConcurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 8, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

MET-D-Qwen3-8B-en-only is an 8 billion parameter English-only moral reasoning model fine-tuned from Qwen3-8B. Developed by launch, this model is designed to judge actions from a specific character's perspective within moral dilemmas, providing explicit chain-of-thought reasoning. It excels at generating nuanced moral judgments and explanations, making it suitable for applications requiring ethical AI reasoning and perspective-taking in English.

Loading preview...

Overview

MET-D-Qwen3-8B-en-only is an English-only moral reasoning model, fine-tuned from the Qwen3-8B base model. Its core function is to analyze moral dilemmas, character descriptions, and candidate actions, then judge the action from that character's perspective. The model provides a detailed chain-of-thought explanation before delivering its judgment.

Key Capabilities

  • Perspective-based Moral Reasoning: Judges actions based on a specified character's viewpoint, rather than a universal moral code.
  • Explicit Chain-of-Thought: Generates detailed reasoning traces to explain its judgments, enhancing transparency and interpretability.
  • Specific Question Answering: Answers two structured questions for each scenario:
    • Is the action acceptable? (Yes / No / Ambiguous)
    • Would doing/not doing the action cause emotional/mental discomfort? (Yes / No)
  • English-Only Focus: This specific checkpoint is optimized and trained exclusively on English data for moral reasoning tasks.

Training and Uniqueness

The model was trained using self-generated reasoning traces, which were then rejection-sampled against ground-truth answers derived from specific character perspectives. This method addresses the challenge of verifying reasoning in moral dilemmas by providing a clear ground truth for evaluation. It is part of the broader MET collection which includes multilingual and other single-language variants across different base models.