任务格式如何影响适配：LoRA 与全量微调在提示词安全分类上的受控对比

View PDF

Abstract:Fine-tuning a language model for classification involves many design decisions. Comparisons of low-rank adaptation (LoRA) with full fine-tuning are common but often confounded, and task format is rarely studied. We compare both regimes on prompt-safety classification with a matched 400M encoder-decoder pair (Ettin) across three task formats, with up to 15 seeds per cell and pre-specified analyses. At the pre-specified learning rates, full fine-tuning leads LoRA by 2-3 pp in every architecture and format trained on the label alone. On the encoder, the gap closes once both rates are chosen by the same search. Training LoRA to write a reason after its verdict adds 4.0 pp and closes the gap, but a content-free sentence of equal length does not. Trained models score above a prompted 8B in domain but fall behind it on two of three out-of-distribution corpora, trailing by 8-23 pp on adversarial jailbreaks and flagging 27-66% of benign but harmful-looking prompts as unsafe, against the 8B's 2-3%. Together, the results suggest that the adaptation regime matters less than how each model is trained, its learning rate, and training target, and that in-domain accuracy alone is an incomplete guide to robustness.

Comments:	33 pages (14 pages main text, plus references and appendices), 1 figure, 18 tables
Subjects:	Computation and Language (cs.CL)
MSC classes:	I.2.7, I.2.6
ACM classes:	I.2.7; I.2.6
Cite as:	arXiv:2610.33813 [cs.CL]
	(or arXiv:2610.33813v1 [cs.CL] for this version)
	https://doi.org/10.48550/arXiv.2610.33813 arXiv-issued DOI via DataCite (pending registration)

Submission history

From: John Alderete [view email]
[v1] Sat, 24 Oct 2026 17:24:54 UTC (269 KB)