Unverified50% confidenceFactExact time
Token control is typically implemented through system prompt injection rules, output filter regex matching, or RLHF fine-tuning.
1
Sources
50%
Confidence
Long-term
Relevance
10/9/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedStructured output applies constraints to the model's token sampling via constrained decoding, typically based on finite-state automata (FSA) or context-free grammars (CFG).67% similarUnverifiedGBNF 通过在每个 Token 生成步骤对 logits 进行掩码过滤，强制模型只能生成符合预定义语法规则的 Token 序列66% similarUnverifiedImplementations of structured output generally fall into two categories: dynamically filtering tokens via grammar constraints during the inference phase, and guiding output through prompt engineering followed by parsing and validation.61% similarUnverifiedConstrained decoding allows only tokens conforming to grammar rules to be sampled by setting the logits of illegal tokens to negative infinity at each decoding step; FSM is applicable to regular languages, while CFG can handle nested structures such as JSON and XML.59% similarUnverifiedOutlines constructs a Finite State Machine (FSM) that dynamically masks tokens at each generation step that would render the format invalid, allowing the model to sample only from the subset of legal tokens.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/992286API
curl https://kongchang.com/api/v1/knowledge/claims/992286MCP
get_claim(id=992286)