Unverified50% confidenceCautionExact time
Benchmark results may be affected by prompt wording, model version, inference settings, and label parsing rules, and aggregate scores may obscure differences across categories.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
Unverified分词特征差异发生在推理链路底层，难以通过提示工程掩盖，可作为模型识别手段79% similarUnverifiedThe method responds to controlled modifications of the declared alternative categories, meaning that when category definitions in the goal model are adjusted, the classification results change accordingly.75% similarUnverifiedBy constructing semantically equivalent but differently phrased prompt variants and observing the output distribution, one can distinguish whether sensitivity originates from the model or the evaluator.75% similarUnverified多项实验表明，模型的表面推理与其实际决策机制之间存在显著偏差，通过激活修补可发现真正影响输出的内部路径与书面推理不吻合74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1003403API
curl https://kongchang.com/api/v1/knowledge/claims/1003403MCP
get_claim(id=1003403)