Unverified50% confidenceFactExact time
Real-world paraphrasing reduces the invalid intent rejection rate: among 16 configurations, 15 showed a drop in rejection rate for the best model.
1
Sources
50%
Confidence
Long-term
Relevance
11/6/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedExperiments show that models processed with abliteration exhibit almost no degradation on standard capability evaluations, but the probability of refusing harmful requests drops significantly.77% similarVerifiedIn a 2024 research paper, Anthropic quantified the phenomenon of model over-refusal, finding that safety-trained models had false positive refusal rates as high as 15–30% in certain scenarios76% similarUnverifiedIn 11 of the 12 models, between 87.5% and 100% of capitulation cases stopped at the exact wrong number provided by the user, meaning capitulation typically involves switching sides entirely.76% similarUnverifiedIn terms of invalid intent rejection, the best model correctly rejected 191 of 195 invalid intents.75% similarUnverifiedModule residual anchoring lowered the hit rate under natural questioning from 28.7% to 22.0% while also reducing the measured behavioral drift, a trade-off of exchanging recall for preservation of non-target behavior.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1141655API
curl https://kongchang.com/api/v1/knowledge/claims/1141655MCP
get_claim(id=1141655)