SOURCE-LINKED INTELLIGENCE

Routing Is Not Enough: Diagnosing Intra-Adapter Subspace Contention in MoE+LoRA Fine-Tuning

arXiv · AI, language, vision and robotics · article · Sep 2, 2026 · UTC

Multi-domain fine-tuning often combines MoE routing with LoRA, assuming that token-level routing separates domain-specific updates. We test this assumption in MoE+LoRA using Python code paired with biomedical text and mathematical reasoning. Although these domains show near-disjoint expert routing, adding biomedical data substantially increases code perplexity, indicating that routing separation alone may not prevent negative transfer. To localize the failure, we introduce Jaccard routing overlap and adapter-gradient cosine similarity, which measure expert sharing and update compatibility, res

Read original source ↗ Open in workspace

recordType

paper

region

Global