Unverified50% confidenceBenchmarkExact time
The researchers tested ten hosted models under two calling interfaces and recorded a total of 322 real hallucination cases.
1
Sources
50%
Confidence
Long-term
Relevance
9/18/2026
First Seen
Sources
Related Entities
Related Claims
Unverified幻觉（Hallucination）分为内在幻觉和外在幻觉两类，检测技术路线包括基于知识库的事实核查、基于模型不确定性的检测和基于多次采样一致性的方法70% similarUnverifiedThe research team used the Llama-3.1-8B-Instruct model as the experimental subject, examining probe generalization performance on 3 held-out deception detection datasets.67% similarVerifiedCommonly used evaluation benchmarks in the field of hallucination detection include TruthfulQA, HaluEval, and FActScore.67% similarUnverifiedThe research team tested seven Transformer models on 4,800 TranSGrid instances.66% similarUnverifiedThe study employs four complementary causal diagnostic methods, covering four decoder-only architecture models and eight distinct prompt families.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/931266API
curl https://kongchang.com/api/v1/knowledge/claims/931266MCP
get_claim(id=931266)