Unverified50% confidenceFactExact time
Adapters insert small trainable modules between model layers, freezing the original weights and achieving capability extension through external modules.
1
Sources
50%
Confidence
Long-term
Relevance
11/2/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedBy fine-tuning with LoRA over a shared, frozen backbone and maintaining only two lightweight adapters, one for the student and one for the teacher, memory usage can be reduced from two full models to one backbone plus two small adapters.71% similarUnverifiedPartial freezing means freezing the bottom several layers of the encoder and unfreezing only the top N layers, while keeping the decoder, joint network and prompt embeddings trainable.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1127969API
curl https://kongchang.com/api/v1/knowledge/claims/1127969MCP
get_claim(id=1127969)