CyberJurors: Multi-Agent AI for E-Commerce Dispute Resolution
A recent study has unveiled CyberJurors, a framework designed for multi-agent simulations that addresses disputes in e-commerce transactions via crowdsourced jury decisions. Published on arXiv, the research introduces the E-commerce Dispute Verdicts (EDV) task alongside VerdictBench, a multimodal benchmark comprising 6,000 actual cases that showcase crowdsourced jury outcomes. CyberJurors employs an Individual Verdict Chain-of-Thought, breaking down the EDV task into four organized reasoning phases, which enhances the perception of subtle clues and clarifies the causal relationships among evidence. This framework seeks to overcome the challenges posed by existing approaches in managing redundant, multi-round, multimodal evidence according to diverse platform-specific standards. The findings emphasize AI's increasing influence in automating intricate decision-making in digital marketplaces.
Key facts
- CyberJurors is a multi-agent framework for e-commerce dispute verdicts.
- VerdictBench includes 6,000 real-world e-commerce dispute cases.
- The EDV task requires grounding clues from multimodal evidence.
- Individual Verdict Chain-of-Thought has four reasoning stages.
- The research is published on arXiv with ID 2605.28369.
- E-commerce platforms use crowdsourced jurors for dispute resolution.
- Existing methods are insufficient for e-commerce dispute scenarios.
- The framework clarifies causal logic between evidence pieces.
Entities
Institutions
- arXiv