LLPF connects independently trained models by moving toward the target, rescaling each layer's weights back to a fixed variance sphere, and retraining; it works on modern architectures and across training hyperparameters.
16 Published as a conference paper at ICLR 2026 Table 3: Hyperparameters used for applying Algorithm 2 to find the low-loss path toward origin
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Connecting Independently Trained Modes via Layer-Wise Connectivity
LLPF connects independently trained models by moving toward the target, rescaling each layer's weights back to a fixed variance sphere, and retraining; it works on modern architectures and across training hyperparameters.