Unverified50% confidenceBenchmarkExact time
在GPQA V2版本测试中,Glimmer的幻觉率高达82%
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/18/2026
First Seen
Valid until: 11/16/2026
Sources
Related Claims
UnverifiedMuse Glimmer在Artificial Analysis第二版测试中幻觉率高达82%,知识工作和幻觉是其最大弱点72% similarUnverifiedO4-mini在GPQA博士级问答评测中准确率超过80%60% similarUnverified在 GPQA 上,领域博士生平均正确率约65%,非专家用户约34%60% similarUnverified人类领域专家在 GPQA 上平均约得 65%59% similarUnverifiedPagedAttention可将GPU显存利用率从通常的20%-40%提升至接近理论上限,吞吐量相比原生HuggingFace推理提升数倍至十余倍59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/767848API
curl https://kongchang.com/api/v1/knowledge/claims/767848MCP
get_claim(id=767848)