Megadose AI progress, ranked and analyzed.

Not All Ranks Are Equal: Budget-Aware LoRA Merging Across Tasks

· HF Daily Papers ·
The paper says smarter rank allocation can narrow the gap between merged LoRAs and task-specific adapters.

The authors argue that current LoRA merging methods waste budget by giving every layer, and sometimes every task, the same rank. They introduce Net Utility, a data-free SVD-based scoring method that selects singular directions by estimated task value and cross-task interference. Applied across five merging methods and three merging spaces, it improves average performance by 2.1% on vision tasks and 2.2% on language tasks. HF Daily Papers' note

score 4

Categories: Research