Tag
The paper shows that PRM-Pruned Fragment Grafting (PPFG), a cost-minimal inference-time cross-trajectory transfer technique, is statistically indistinguishable from independent parallel CoT across three reasoning LMs and six benchmarks, with the inertness traced to poor injection targeting and selection-dominated pruning dynamics.
This paper evaluates whether structure-preserving de-identification via surrogate substitution maintains PHI detectability across multiple detectors, using equivalence testing on 57k paired spans across 7 languages and 11 detectors.