LLM Digest
Subscribe

Story

arxiv_cs_lg · Nov 13, 2026 · paper

Source brief

Direction or Magnitude? Learning Scalars over Task-Calibrated Directions

arxiv.orgNov 13, 2026
original source linked

In brief

Parameter-efficient fine-tuning (PEFT) aims to preserve adaptation quality while reducing training costs. We propose SCALAR, which achieves performance comparable to LoRA with substantially fewer trainable parameters, lower peak training memory, and higher throughput. SCALAR cons

Continue reading

Read the original at arxiv.org →Open in live feedRead that day’s brief

Earlier in this thread 4 items