Unverified50% confidenceBenchmarkExact time
PiPMRE achieves an average improvement of 5.6 percentage points in recall over the previous state-of-the-art methods on two public datasets.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedReAct 在 HotpotQA 等多跳问答任务上比单独 CoT 提升约 10-20 个百分点62% similarUnverifiedThe pass rate on the SWE-bench Verified subset jumped from under 5% in early 2024 to 49%.61% similarUnverifiedMedIQ多智能体系统最终达到57%准确率，摘要智能体带来约5%提升，而拥有全部前置信息的Oracle上限为66%61% similarUnverified在从50%到92%的准确率提升中，REPL贡献最大（直接从50%跳至74%），随后领域知识注入、模糊搜索、公式追踪、系统prompt优化和bug修复层层累积61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/897304API
curl https://kongchang.com/api/v1/knowledge/claims/897304MCP
get_claim(id=897304)