DraMaze cuts repetition in long microdrama scripts
Original title: DraMaze: A Multi-Agent Pipeline with Future-Aware Memory for Diverse Microdrama Script Generation
On scripts of up to 60 episodes and 60k words, repetition fell 32.1% and Logic scored 82.5, 8.5 above the best baseline.
DraMaze is a multi-agent pipeline that searches diverse causal paths with trope-memory evolutionary tree search and mutation, then retrieves history through a Backward Narrative Dependency Tree tied to future plot goals. The authors introduce MiDramaBench, a bilingual benchmark of script logic and creative quality at different lengths, plus causal-pattern metrics. They report gains across three backbone language models.
Unconfirmed: The 32.1% reduction and 82.5 Logic score are the authors' experimental results.
- Candidate continuations concentrate on fewer narrative mechanisms as scripts lengthen.
- MiDramaBench evaluates logic and creative quality across script lengths.
- Causal-pattern metrics measure narrative repetition and diversity.
- Improvements are reported for three unnamed backbone LLMs.
From the original article
1 min read at huggingface.co
The rapid growth of microdramas has increased demand for high-quality script production. Long-form microdrama script generation with LLMs requires both diverse plot progression and long-range narrative coherence.
By Siyang He, Qiqi Wang, Fan Liu · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.