Unverified50% confidenceTradeoffExact time
若单一模型漏检率为 p，两个训练背景不同的模型同时漏检同一错误的概率近似降至 p²，20%漏检率理论上可降至 4%
1
Sources
50%
Confidence
Long-term
Relevance
7/12/2026
First Seen
Sources
Related Claims
VerifiedReAct's errors amplify in a chain: if the accuracy per step is p, the overall success rate after n steps is approximately p to the power of n; with a per-step accuracy of 95%, the overall success rate drops to approximately 59.9% after 10 steps and only 35.8% after 20 steps.73% similarUnverified若单帧误检率为p，连续8帧独立误检的概率为p^8，即使单帧误检率高达10%，8帧连续误检概率也只有10^-871% similarUnverifiedExperiments conducted by the research team on models such as GPT-2 showed that after only 5–9 generations of recursive training, the quality of model outputs degrades significantly.68% similarUnverified将FP32权重降至INT8后，典型Transformer模型基准测试分数下降通常不超过1%，推理速度提升2至4倍，内存占用减少75%68% similarUnverifiedIn 2023, researchers demonstrated that contaminating less than 1% of training data is sufficient to alter model behavior under specific trigger conditions68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/489025API
curl https://kongchang.com/api/v1/knowledge/claims/489025MCP
get_claim(id=489025)