Research

Which and When to Admit: Gradient Admission for Data-Centric Small Language Model Finetuning

Single source1 source7 Oct 2026

arXiv:2610.07553v1 Announce Type: cross Abstract: LoRA fine-tuning adapts small language models (SLMs) to heterogeneous instruction data within a low-rank update subspace, making it vulnerable to three structural problems: conflicting gradients that cancel, st

Reported by 1 source

arxiv.orghttps://arxiv.org/abs/2610.07553

← Back to all developments · Signals are a development reported by more than one source. See the methodology.