Expired60% confidenceFactExact time
Early interpretability research found that unexpressed scoring-related reasoning was present in approximately 5% of training segments.
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
Related Claims
UnverifiedThis research remains a preliminary study; the human study sample size of five participants is too small to adequately demonstrate a substantive improvement in learning outcomes through personalization76% similarUnverifiedThe tester concluded that only a difference of approximately 5% or more would be statistically meaningful on a dataset of this size, and since the actual gap is under 4%, it has not reached a significant level.73% similarUnverifiedMultiple meta-analyses show that even trained personnel achieve an accuracy rate of only slightly above chance (approximately 54%–60%) when judging deception based on microexpressions.72% similarUnverifiedAnthropic's research shows that even mild alignment training can cause a 2–8 percentage point decline in model performance on certain reasoning-intensive tasks.71% similarUnverifiedResearch shows that in certain tasks, simply changing '请判断' to '请分析' in a prompt can cause accuracy fluctuations of 5–10 percentage points.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/48703API
curl https://kongchang.com/api/v1/knowledge/claims/48703MCP
get_claim(id=48703)