nabin2004/AOS-qwen3-8b-narrated-dpo

AOS-qwen3-8b-narrated-dpo

Direct Preference Optimization (DPO) aligned LoRA adapter for Manim Community Edition mathematical and educational animation synthesis.

Alignment Objective

Aligns the model to strongly prefer generating synchronized `manim-voiceover` code examples:

• —Chosen: Clean VoiceoverScene scripts with self.set_speech_service(GTTSService()), animation duration tracking (run_time=tracker.duration), and phonetic mathematical explanations.
• —Rejected: Silent, un-narrated standard Scene code.

Lineage

• —Base LLM: Qwen/Qwen3-8B
• —SFT Prior: nabin2004/AOS-qwen3-8b-narrated-adapter