Unverified50% confidenceFactExact time
Faithfulness, sometimes also called groundedness, measures whether every claim in a generated answer can be found in the retrieved context, focusing on whether the model fabricated content outside the context.
1
Sources
50%
Confidence
Long-term
Relevance
10/17/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedLanham等人2023年的研究《Measuring Faithfulness in Chain-of-Thought Reasoning》使用了CoT扰动测试76% similarUnverifiedThe paper redefines faithfulness evaluation as a context-aware "alignment-then-labeling" problem: first identify phrase-level correspondences between informal and formal statements, then assign a faithfulness label to each pair under explicitly given contextual interpretations and assumptions.74% similarUnverifiedAnswer accuracy and retrieval faithfulness are two mutually independent evaluation dimensions; faithfulness evaluation requires verifying whether each key claim in the answer can be directly traced back to the retrieved documents.72% similarUnverifiedFaithfulness corresponds to the concept of factual consistency (Faithfulness / Factual Consistency) in natural language generation evaluation, and differs subtly from truthfulness: truthfulness means whether the answer matches facts in the objective world, while faithfulness only requires that the answer does not go beyond the given context; therefore faithfulness is a more measurable and more controllable metric, whereas truthfulness also depends on the reliability of the retrieval source itself.71% similarUnverifiedAt the current stage, efforts should be made to maintain the readability and faithfulness of the chain of thought, using it as a diagnostic tool rather than directly suppressing it.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1054183API
curl https://kongchang.com/api/v1/knowledge/claims/1054183MCP
get_claim(id=1054183)