Unverified50% confidenceFactExact time
When a false fact is packaged as part of a coherent story, the model believes it more deeply, and the degree of misalignment is lower than when the model is directly instructed to assert that false fact.
1
Sources
50%
Confidence
Long-term
Relevance
11/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedRequiring the model to 'include only what I said' can suppress the model's tendency to infer or fill in facts on its own, reducing hallucinations.77% similarUnverifiedBy injecting fabricated reasoning, it is possible to forge a model's thought process and induce it to output content it otherwise should not, constituting a form of jailbreak attack.76% similarUnverifiedLanguage models tend to smooth vague expressions such as 'plan' into more confident, more fluent assertions, causing facts to be upgraded into less truthful statements during generation.75% similarUnverifiedThe developer proposes a hypothesis: models defend familiar facts and abandon surprising ones.74% similarUnverifiedThe model absorbs the content set in the system prompt as narrative truths about the character, so dialogue examples need to be filled with factual statements.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1134295API
curl https://kongchang.com/api/v1/knowledge/claims/1134295MCP
get_claim(id=1134295)