Expired85% confidenceFactTime unknown
Anthropic uses an automated classifier to determine whether Claude exhibits sycophantic behavior, and this classifier was fine-tuned on a large number of human-annotated sycophantic/non-sycophantic conversation samples.
1
Sources
85%
Confidence
Medium-term (~90 days)
Relevance
6/1/2026
First Seen
Valid until: 8/30/2026(expired)
Sources
Claude谄媚问题深度解析：灵性话题谄媚率高达38%
rss
Related Entities
Related Claims
UnverifiedAnthropic developed an automated classifier to assess whether an AI exhibits sycophantic behavior across four dimensions: proactively raising objections, maintaining its judgment, matching praise to quality, and candidly expressing genuine evaluations.80% similarUnverifiedClaude Code's adversarial verification mode eliminates self-preference bias by dispatching independent agents to play the role of 'devil's advocate,' specifically raising the strongest possible objections.63% similarUnverified依靠人工纪律进行代码审核不可靠，将审核做成系统自动触发比依赖人工纪律更有效60% similarUnverifiedEric from the Anthropic team pointed out that a common mistake is people spending enormous effort writing prompts while providing the model with extremely rudimentary tools.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/32798API
curl https://kongchang.com/api/v1/knowledge/claims/32798MCP
get_claim(id=32798)