Unverified50% confidenceOpinionExact time
context-mode's 98% savings is an upper bound measured under ideal conditions and does not count the tokens the model consumes to generate the summary script itself, which is also part of the context.
1
Sources
50%
Confidence
Long-term
Relevance
10/17/2026
First Seen
Sources
Related Entities
Related Claims
Unverified框架层的抽象会对底层LLM调用的原始prompt进行二次加工，导致框架日志中看到的用户输入与模型实际接收的完整prompt存在差异68% similarUnverifiedThe author points out that about 70% of LLM calls are simple classification, entity extraction, or short-text summarization, which small models can handle with lower latency.67% similarUnverifiedTraditional embedding models map text fragments to fixed vectors without considering the position of those fragments within a larger document.66% similarVerifiedContextual Retrieval was proposed by Anthropic; it prepends an LLM-generated contextual summary to each chunk so that the chunk retains the necessary background information even when separated from the original text.64% similarUnverifiedWhen a model is responsible for both generating a summary and reading that summary to continue its work, the compression step becomes an attack surface that may be self-exploited.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1054914API
curl https://kongchang.com/api/v1/knowledge/claims/1054914MCP
get_claim(id=1054914)