Unverified50% confidenceSolutionExact time
上下文偏置（Contextual biasing）在解码阶段介入，通过向模型提供期望词表提高特定词的输出概率，实现方式包括浅层融合、深层偏置和 TCPGen 等架构
1
Sources
50%
Confidence
Long-term
Relevance
8/21/2026
First Seen
Sources
Related Claims
Unverified上下文偏置/热词偏置在解码阶段介入，通过提升特定词汇在 beam search 中的输出概率让模型倾向输出英文原名，实现方式包括浅层融合和深层偏置79% similarUnverifiedReranking uses Cross-Encoder models to perform fine-grained scoring and reordering of candidate documents against the query, improving the quality of context sent to the LLM.71% similarUnverified为模型设定角色标签会激活对应的语料分布，使输出概率分布向该领域专业表达倾斜，从而提升输出质量71% similarUnverifiedThe ability of a large model to actively infer design specifications from platform context technically falls under Contextual Reasoning, which internalizes industry common sense from massive amounts of data rather than relying on hard-coded rules.70% similarUnverifiedStructured prompts can significantly improve the output quality of large models compared to natural language descriptions, because the attention mechanism of LLMs is more sensitive to explicit constraints69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/783430API
curl https://kongchang.com/api/v1/knowledge/claims/783430MCP
get_claim(id=783430)