Unverified50% confidenceFactExact time
The MTP head parameter count is typically less than 1% of the main model and is jointly pre-trained with the main model.
1
Sources
50%
Confidence
Long-term
Relevance
9/10/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedWhen fine-tuning with LoRA, the actual number of parameters that need to be trained is only 0.1%-1% of the original.73% similarVerifiedFor a 7-billion-parameter model, LoRA typically requires training less than 1% of the total parameters, and GPU memory requirements can be reduced by more than 60%.71% similarUnverified即使只有5-10%的训练样本存在错位，也会导致VLM模型在下游任务上出现显著的性能退化69% similarUnverified针对有限意图集合训练的轻量级专用模型参数量通常在10M以下，其推理延迟可比通用LLM调用低2-3个数量级69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/887329API
curl https://kongchang.com/api/v1/knowledge/claims/887329MCP
get_claim(id=887329)