Unverified50% confidenceConceptExact time
Keyword filtering is string-level blacklist matching, which can be bypassed by swapping synonyms, translating names, or describing gameplay; semantic recognition relies on the model's understanding of the overall meaning of the input, and can be judged as a replication request even if no game name appears.
1
Sources
50%
Confidence
Long-term
Relevance
11/3/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedContent filtering refers to the interception of text at the keyword, semantic, or classifier level before or after model output74% similarUnverifiedA real-time content filtering layer typically comprises three components: rule-based filtering using keywords and regular expressions, semantic filtering based on classifiers, and secondary review by another LLM.67% similarUnverifiedLeet spelling variants refer to writing that replaces letters with numbers or symbols (such as writing ignore as 1gn0r3), which is inherently unfriendly to filters based on keyword or string matching, and defense usually requires normalization preprocessing or semantic-level judgment.66% similarUnverifiedIn spam identification, senders may generate adversarial examples by inserting special characters, making homophone substitutions, splitting keywords, or adjusting word order to evade model detection66% similarUnverifiedA controlled vocabulary is a manual indexing tool in information retrieval, in which domain experts predefine a set of standardized terms and establish hierarchical, equivalence and associative relationships among them.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1131264API
curl https://kongchang.com/api/v1/knowledge/claims/1131264MCP
get_claim(id=1131264)