Unverified50% confidenceFactExact time
On Opus 5.5, of the 20 skills, 12 were clearly helpful, 2 were clearly worse, and 6 had insufficient samples to judge.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/19/2026
First Seen
Valid until: 1/17/2027
Sources
Related Entities
Related Claims
UnverifiedOn Sonnet 5.5, of the 20 skills, 12 were clearly helpful, 4 were clearly worse, and 4 had insufficient samples to judge.80% similarUnverifiedIn a natural experiment, 7 of 11 skill assessment scores changed between two runs, with a maximum fluctuation of 1 point.72% similarUnverifiedThe criteria are divided into four families—foundational, premature disclosure, goal deviation, and information omission—with 19 failure modes in total, of which 18 were triggered in actual testing.69% similarUnverifiedClaude Opus 5.5 achieved a perfect 20-out-of-20 score on only 241 of the 507 tasks, for a 20-of-20 reliability of 47.53%.69% similarUnverifiedWhen the benchmark set grew from 10 tasks to 20, an agent that originally passed all 10/10 showed 10 failures under the new counting method, causing it to be misjudged as capability decay and automatically quarantined.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1065262API
curl https://kongchang.com/api/v1/knowledge/claims/1065262MCP
get_claim(id=1065262)