Is S$^2$R$^2$ superseded?

S$^2$R$^2$ (Parameter-efficient fine-tuning (LoRA family)): current frontier — recent, not yet superseded in the knowledge base. 0 paper(s) critique it, 0 beat it on benchmarks — not ranked as a superseded baseline. Sub-problem: cluster led by SMART. Newer alternatives in the same sub-problem include Small Data, Big Noise, S²R².

Parameter-efficient fine-tuning (LoRA family)

S$^2$R$^2$

Where Do Prompt Perturbations Break Generation? A Segment-Level View of Robustness in LoRA-Tuned Language Models

On the frontierfirst seen May 2, 2026

Current frontier — recent, not yet superseded in the knowledge base

0 papers critique it · 0 beat it on benchmarks

What to use instead

Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.

Small Data, Big NoiseSmall Data, Big Noise: Adversarial Training for Robust Parameter-Efficient Fine-Tuning
Jun 9, 2026
S²R²Where Do Prompt Perturbations Break Generation? A Segment-Level View of Robustness in LoRA-Tuned Language Models
May 2, 2026