Unverified50% confidenceBenchmarkExact time
在GPT Researcher的实验中，传统嵌入检索仅46%的上下文相关
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/30/2026
First Seen
Valid until: 12/29/2026
Sources
Related Entities
Related Claims
Unverified在96%的关卡里，GPT-6找到答案所需的尝试次数比普通人还少68% similarUnverifiedGPT Researcher实验中决策式方法相比传统嵌入检索将有效上下文比例提升了约27个百分点68% similarUnverifiedGPT模型在回忆能力单独测试中表现糟糕，在空间信息、转录内容、物理实验等所有类别上仅约50%事实被正确回忆，开源模型更差66% similarUnverified该73% vs 46%的数据来自单一来源实验分享，尚未见到大规模多场景复现验证，实验设置未充分披露62% similarUnverifiedJev与Neural-MCTS的首选动作只有约63%的时间一致62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/960483API
curl https://kongchang.com/api/v1/knowledge/claims/960483MCP
get_claim(id=960483)