Unverified50% confidenceBenchmarkExact time
In the fact-updating task, the success rate of the embedding retrieval method fluctuated between 0.30 and 0.95, with a maximum gap of 0.45 across different random seeds.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified消融实验中移除搜索工具后Doubao Seed 2.0 Pro准确率从43.1%暴跌至2.4%，证明任务必须依赖外部检索71% 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一69% similarUnverified推测解码中当草稿通过率低于50%时，实际加速效果会大打折扣甚至产生负优化67% similarUnverified在歧义点的四类行为中，'先搜再问'成功率最高，平均通过率达93.4%，远高于'直接猜'的56.5%，'搜多次再猜'为51.5%66% similarUnverifiedOn dependency tasks, memory raised the success rate from a leakage-verified baseline of 0.00 to 0.55–1.00.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/895509API
curl https://kongchang.com/api/v1/knowledge/claims/895509MCP
get_claim(id=895509)