README.md

metadata

title: Dialectic Reasoning
emoji: 😻
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 6.10.0
python_version: '3.12'
app_file: app.py
pinned: false

Dialectic Reasoning

Interactive demo for the dialectic LoRA model family, fine-tuned to identify genuine tensions, make conditional commitments, and reach integrative resolutions instead of hedging.

Current Best: 4B v3

The strongest model in the family is the Qwen3-4B v3 LoRA:

• Trained on 507 examples (408 original + 99 domain-diverse traces from 3 model families)
• Rubric avg: 9.8/10 — all 14 held-out prompts score "strong"
• generic_hedge: 0.00 (eliminated)

The earlier 8B model (6.6/10 on 408 traces) demonstrated that data diversity matters more than model size.

What This Demo Shows

• Crux identification — finding the real decision point
• Conditional commitment — "if X, then Y; if Z, then W"
• Integrative resolution — not "both sides have merit" but concrete synthesis

This is not a balanced conversation bot. It is a demo of a specific trained capability.

Evidence

For methodology and evaluation:

• Best model: hikewa/dialectic-qwen3-4b-v3-lora
• 8B model: hikewa/dialectic-qwen3-8b-lora
• Dataset + eval artifacts: hikewa/dialectic-reasoning-traces

Limitations

• The Space is a demo wrapper, not a research paper
• Training data is synthetic (multi-model generated)
• English-only
• Stronger evidence comes from held-out evaluation, not from chat impressions