Unverified50% confidenceFactExact time
The SWORD benchmark covers eight widely used languages to systematically examine models' cross-lingual fact-checking consistency.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe study tested six mainstream large language models.71% similarUnverifiedThe article recommends that when deploying multilingual AI systems, one should build dedicated distortion test sets for target languages, introduce multi-level fact verification pipelines, and set differentiated confidence thresholds for different languages.67% similarUnverifiedEleutherAI's Language Model Evaluation Harness focuses on capability benchmark testing65% similarUnverifiedSome large language models are capable of reproducing verbatim the questions and answers from popular benchmarks.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/893863API
curl https://kongchang.com/api/v1/knowledge/claims/893863MCP
get_claim(id=893863)