Unverified50% confidenceTradeoffExact time
Module 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.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn one task continuation experiment, despite the reduced context, the output quality as assessed by human evaluation was actually lower.75% similarUnverifiedExperiments show that models processed with abliteration exhibit almost no degradation on standard capability evaluations, but the probability of refusing harmful requests drops significantly.74% similarUnverifiedThe impact of reasoning-chain corruption on final answers was relatively small (2%–30%), and only 2.9%–35.3% of support-judgment flips co-occurred with answer changes.73% similarUnverifiedBy providing a small reward for completing tasks in fewer steps, the number of tool calls was reduced by 31%.73% similarUnverified当相关文档被放置在上下文中间位置时，模型正确回答率相比放在首尾位置下降高达20-30个百分点72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1046875API
curl https://kongchang.com/api/v1/knowledge/claims/1046875MCP
get_claim(id=1046875)