Unverified50% confidenceFactExact time
Claude Opus 5 induced separate scoring rules for absolute difficulty judgments and pairwise comparison difficulty judgments, respectively.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/30/2026
First Seen
Valid until: 1/28/2027
Sources
Related Entities
Related Claims
UnverifiedOn Opus 5.5, asking Claude to write out its full reasoning process before answering may be flagged and rejected, with the error reason shown as 'reasoning extraction,' intended to prevent other companies from distilling Opus 5.5.68% similarUnverifiedThe paper identifies five physical constraints, including partial observability, irreversibility, and unverifiable outcomes, which together lead to fewer pre-fixed choices and less room for error correction.65% similarUnverifiedThe most noticeable writing characteristic in Opus 5.5 model output is an abnormal preference for the word 'dependable'.63% similarUnverifiedThe experimental findings concluded that conventional behavioral alignment evaluations alone would have difficulty identifying the alignment problems in Hacker Opus, and that alignment auditing is becoming increasingly difficult.61% similarUnverifiedIn refusing, Opus 5.5 did three things: clearly identify the problem, explain the risk, and return the final decision to the user.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1113947API
curl https://kongchang.com/api/v1/knowledge/claims/1113947MCP
get_claim(id=1113947)