Fine-Tuning Improves Information Conveyance in Language Models 事件

Fine-Tuning Improves Information Conveyance in Language Models arXiv:2605.30844v1 Announce Type: new Abstract: Fine-tuning is often believed to reduce uncertainty and diversity in large language models, but existing analyses overlook output length, a key confounder, and therefore fail to capture how uncertainty is distributed across an entire generation rollout. To address this, we propose Canopy Entropy ($\mathrm{CE}^\star$), a measure that views language generation from a tree perspective, wh