Unverified50% confidenceSolutionExact time
Perturbation testing can refer to NLP robustness evaluation benchmarks (such as CheckList and AdvGLUE), which categorize perturbations into three types: character-level, lexical-level, and semantic-level.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedEach test item consists of three parts: a target review, a reconstructed context, and plausible misreading options.73% similarUnverifiedThe evaluation adopts progressively stricter nested matching criteria, encompassing three dimensions: concept hit, evidence snippet localization, and correctness judgment.71% similarUnverifiedThe scorer design is divided into three layers: L1 is deterministic code assertions for Side Effects, L2 is LLM-as-a-Judge plus gold-standard review for root-cause analysis quality, and L3 is usage statistics.70% similarUnverified三种自动化评测方案为代码断言、底层环境校验、LLM大模型裁判，可信度依次递减70% similarUnverifiedCommon detail-testing methods used by developers include jailbreak robustness testing, instruction-following consistency testing, and ambiguity resolution testing70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/995603API
curl https://kongchang.com/api/v1/knowledge/claims/995603MCP
get_claim(id=995603)