arXiv · cs.AI · Shenghui Li, Thiemo Voigt · 8 小时前模型评分38
LoRAForge：面向多 低秩适配（LoRA，Low-Rank Adaptation） 高效服务的事后压缩与执行协同设计
LoRAForge: Co-Designing Post-Hoc Compression and Execution for Efficient Multi-LoRA Serving
导读
LoRAForge 通过显式保留秩表示，把批量因子构建与秩特化执行连接起来，支持统一保留比例（LoRAForge-U）和谱能量预算分配（LoRAForge-S）两种策略。
分类置信度
中 · 待人工复核
- 只有单一媒体来源，需要人工复核。
来源级别：媒体与个人
事件类型：研究发布
证据摘录
- “We present LoRAForge, a system connecting batched factor construction to rank-specialized execution through an explicit retained-rank representation.”
- “It supports two allocation policies: uniform retention ratios (LoRAForge-U) and spectral-energy budget allocation (LoRAForge-S).”
- “Across 85 trained open-weighted Qwen and Llama adapters, the spectral policy's floored quality configuration retains about 7.7% of adapter parameters with fleet-average score drops of at most 1.2 points on a 0--100 scale.”
来源：arXiv · cs.AI · arxiv.org