Unverified50% confidenceFactExact time
A harmful-prompt indicator variable plus other framing effects together explain on average only 15% of the score variance within checkpoints; 85% of the variation comes from the checkpoint itself.
1
Sources
50%
Confidence
Long-term
Relevance
10/23/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedCausal probes achieved a 17%–118% improvement in steering capability across different models, with only a 3% decrease in concept detection capability.75% similarUnverifiedIn a standalone test of the "view incomplete" prompt on 100 questions, the wrong-answer rate was 16% without the prompt and dropped to 5% with it.74% similarUnverifiedCorrupted reasoning chains shifted support judgments by 10%–22% under the blind-verifier prompt.73% similarUnverifiedThe standard practice aggregates all heterogeneous harmful prompts into a single uniform mean vector, which may conflate distinct refusal-related phenomena.73% similarUnverifiedOn Track B, the flat RAG baseline can detect 76%–97% of true contradictions, but produces false positives on 16%–43% of surface-similar safe drafts (the specific ratio depends on the backend).71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1088122API
curl https://kongchang.com/api/v1/knowledge/claims/1088122MCP
get_claim(id=1088122)