Unverified50% confidenceFactExact time
VLM能够根据自然语言描述识别训练数据中从未出现的目标类别，这种能力被称为零样本（zero-shot）或开放词汇（open-vocabulary）识别
1
Sources
50%
Confidence
Long-term
Relevance
8/22/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn LoRA, matrix A is initialized with random Gaussian values and matrix B is initialized to zero, ensuring that the added quantity is zero at the start of training so as not to damage the original model's capabilities, and introducing no additional latency during inference.63% similarUnverifiedFew-shot Learning相比全量微调的优势在于零训练成本和即时切换能力，只需更换几个示例即可调整品牌语气62% similarUnverifiedTeams training label-based zero-shot models should actively shuffle candidate order during evaluation as a routine robustness check.62% similarUnverifiedAn open-weights strategy means that model parameters can be downloaded and used for free, but training data and the complete training process are generally not made public, and it has become the mainstream choice for companies such as Meta Llama and Mistral.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/787393API
curl https://kongchang.com/api/v1/knowledge/claims/787393MCP
get_claim(id=787393)