Unverified50% confidenceFactExact time
Although the Adapter method reduces training parameters, it introduces extra layers at inference time, increases latency, and cannot be eliminated through matrix merging.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedA drawback of the Adapter method is increased inference latency, because information flow must pass serially through the additionally inserted modules.77% similarVerifiedThe core idea of PEFT is to not update all parameters, but instead introduce a small number of trainable adapter parameters (such as LoRA's low-rank matrix decomposition), reducing VRAM and computation requirements while retaining performance close to full fine-tuning.76% similarUnverified'Slowing down' lacks an actionable definition, and no consensus has been reached on which metrics to slow (model parameter scale, training compute investment, or capability release schedules).72% similarUnverifiedThe encoder-only repair approach can reduce trainable parameters by 3× and optimizer-state memory by 3×.71% similarVerifiedDuring LoRA fine-tuning, the original model parameters are frozen and only the inserted low-rank matrices are updated, significantly reducing VRAM requirements and training costs.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1047750API
curl https://kongchang.com/api/v1/knowledge/claims/1047750MCP
get_claim(id=1047750)