Search papers, labs, and topics across Lattice.
This study investigates the limitations of combining Mixture of Experts (MoE) routing with Low-Rank Adaptation (LoRA) in multi-domain fine-tuning, revealing that token-level routing does not sufficiently prevent negative transfer when domains share similar gradients. By employing metrics such as Jaccard routing overlap and adapter-gradient cosine similarity, the authors diagnose that interference primarily stems from competing gradients within the same low-rank adapter subspace. The introduction of SpawnLoRA, which adds gated sub-adapters to mitigate this contention, shows significant improvements in reducing negative transfer across various model configurations compared to traditional methods.
Nearly orthogonal domain gradients can still wreak havoc in MoE+LoRA setups, but a new approach effectively curbs this interference.
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, respectively. These diagnostics indicate that interference arises mostly from nearly orthogonal domain gradients competing within the same low-rank adapter subspace. We address this issue with SpawnLoRA, which dynamically adds gated sub-adapters inside MoE experts when adapter-level contention is detected, while keeping the router fixed. We evaluate SpawnLoRA on Phi-tiny-MoE-instruct and OLMoE-1B-7B across multiple mixture settings and find that it effectively reduces negative transfer compared with standard and rank-adaptive LoRA. These results demonstrate that structural separation inside experts provides benefits beyond routing or rank expansion alone.
