Answers are grounded in this paper and cite their sources.
Investigates LoRA fine-tuning effects on prostate cancer grading accuracy, suggesting a reliable benchmarking method.
This record is the archived pre-registration protocol for the low-rank adaptation (LoRA) fine-tuning branch of the CLAM-TR M.Sc. thesis (Karadeniz Technical University, 2026). It fixes, ahead of execution, the pre-registered question (whether the six-aggregator QWK rank ordering is preserved under LoRA-adapted versus frozen UNI features), the decision rule on the 5-fold mean QWK, and the statistical-analysis plan (paired-t and Wilcoxon tests, Cohen's d, bootstrap 95% CIs, Benjamini-Hochberg correction); it reports the completed Stage 3-4 outcome against that decision rule (best LoRA-UNI 5-fold mean QWK 0.9529 +/- 0.0023, within the frozen-feature noise envelope; TOST equivalence at a +/-0.005 QWK margin). Deposited to provide an external timestamp ahead of the thesis defence.
No takes yet. Share an insight, caveat, or question.
Hassan et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: