The Spark Effect: On Engineering Creative Diversity in Multi-Agent AI Systems
Fuente:
arXiv
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917021606739968 |
|---|---|
| author | Doudkin, Alexander Voelker, Anton von Borries, Friedrich |
| author_facet | Doudkin, Alexander Voelker, Anton von Borries, Friedrich |
| contents | Creative services teams increasingly rely on large language models (LLMs) to accelerate ideation, yet production systems often converge on homogeneous outputs that fail to meet brand or artistic expectations. Art of X developed persona-conditioned LLM agents -- internally branded as "Sparks" and instantiated through a library of role-inspired system prompts -- to intentionally diversify agent behaviour within a multi-agent workflow. This white paper documents the problem framing, experimental design, and quantitative evidence behind the Spark agent programme. Using an LLM-as-a-judge protocol calibrated against human gold standards, we observe a mean diversity gain of +4.1 points (on a 1-10 scale) when persona-conditioned Spark agents replace a uniform system prompt, narrowing the gap to human experts to 1.0 point. We also surface evaluator bias and procedural considerations for future deployments. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_15568 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | The Spark Effect: On Engineering Creative Diversity in Multi-Agent AI Systems Doudkin, Alexander Voelker, Anton von Borries, Friedrich Human-Computer Interaction Artificial Intelligence I.2.11; I.2.7; J.5 Creative services teams increasingly rely on large language models (LLMs) to accelerate ideation, yet production systems often converge on homogeneous outputs that fail to meet brand or artistic expectations. Art of X developed persona-conditioned LLM agents -- internally branded as "Sparks" and instantiated through a library of role-inspired system prompts -- to intentionally diversify agent behaviour within a multi-agent workflow. This white paper documents the problem framing, experimental design, and quantitative evidence behind the Spark agent programme. Using an LLM-as-a-judge protocol calibrated against human gold standards, we observe a mean diversity gain of +4.1 points (on a 1-10 scale) when persona-conditioned Spark agents replace a uniform system prompt, narrowing the gap to human experts to 1.0 point. We also surface evaluator bias and procedural considerations for future deployments. |
| title | The Spark Effect: On Engineering Creative Diversity in Multi-Agent AI Systems |
| topic | Human-Computer Interaction Artificial Intelligence I.2.11; I.2.7; J.5 |
| url | https://arxiv.org/abs/2510.15568 |