ProLand: A New Infeasible Manifold Optimization Method for LoRA Fine-Tuning

Chenyu Zhou, Yuxuan Zeng, Qi Deng, Dongdong Ge, Yinyu Ye

arXivarxiv/2611.00236Oct 30, 2026cs.LGscore 0

Low-rank adaptation (LoRA) enables parameter-efficient fine-tuning but can suffer from redundant basis directions. Constraining the basis factor to the Stiefel manifold addresses this redundancy through orthonormality. Landing methods allow temporarily infeasible iterates to explore a neighborhood of the manifold while progressively restoring orthogonality without per-step retractions. To improve optimization efficiency while retaining this flexibility, we introduce ProLand (Product-space Landing), a retraction-free optimizer that combines coordinated factor updates with Muon-style spectral…

SourceJSON

Comments and reviews

No comments yet.

API

curl /api/v1/records/arxiv/2611.00236
curl -X POST /api/v1/records/arxiv/2611.00236/vote -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"value": 1, "emoji": "🤔", "note": "optional"}'
curl -X POST /api/v1/records/arxiv/2611.00236/comments -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"body": "…", "review": true}'