Unverified50% confidenceBenchmarkExact time
Gemini 4 Carbon's answer accuracy rose from 48% to 58%, but its hallucination rate rose from 4% to 11%.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/29/2026
First Seen
Valid until: 11/12/2026
Sources
Google Releases Gemini 4 Carbon; Qwen Model Requests to Overseas Servers Spark Security Debate
bilibiliAI鼓励师_悠一10/28/2026
Related Entities
Related Claims
UnverifiedCarbon's "swap-to-refusal rate" (hallucination or erroneous generation rate) rose from 4% to 11%.69% similarUnverified移除歧义后所有模型准确率提升26.8至40.2个百分点，Gemini甚至达到81%，表明歧义是深度搜索代理最大性能瓶颈之一67% similarUnverifiedGemini 4 Carbon's unsupported claims are about half as many as Argon's, and severe behavioral deviation alerts in the internal coding environment dropped by nearly 80%.67% similarUnverifiedBoth models expressed almost identical confidence when fabricating versus when answering correctly (Gemini 93.7% vs 95.5%, Gemma 96.3% vs 99.0%), indicating that confidence scores cannot distinguish correct answers from fabrications.66% similarUnverifiedIn the experiment, unit completion accuracy increased from 72.4% to 84.9%, a gain of 12.5 percentage points.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1111561API
curl https://kongchang.com/api/v1/knowledge/claims/1111561MCP
get_claim(id=1111561)