Unverified50% confidenceFactExact time
After switching embedding models, the ranking results are preserved but the scale of similarity scores changes completely, causing a fixed threshold of 0.8 to break down.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified评估阶段忘记调用 model.eval() 会导致模型仍使用当前批次统计量，在批次较小时严重影响推理结果稳定性69% similarUnverifiedBenchmark saturation refers to the point at which, after model performance improves to a certain level, benchmark tests can no longer effectively differentiate capability differences between models.68% similarUnverifiedAgent链式调用多步之后，每次LLM输出的细微差异导致误差逐步放大，最终结果偏离预期67% similarUnverifiedThe distribution shift problem in behavior cloning causes error accumulation, with the expected total loss after T steps being O(εT²) rather than the linear O(εT).66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/998008API
curl https://kongchang.com/api/v1/knowledge/claims/998008MCP
get_claim(id=998008)