Yuxin Liu

The paper analyzes information-sharing mechanisms in oligopolies, finding that privacy protection alone is insufficient to incentivize suppliers to share data; successful sharing requires combining privacy safeguards with a sufficiently informative external signal.

This paper proposes an efficient human intervention mechanism, Deep Interaction, for correcting reasoning errors in large language models, achieving over 25% improvement in correction success rate and reducing token usage by approximately 40%.