Unverified50% confidenceOpinionExact time
端到端模型改变的主要是架构，而许多失败模式（噪声、口音、语码转换）的根源在于数据分布而非架构
1
Sources
50%
Confidence
Long-term
Relevance
8/11/2026
First Seen
Sources
Related Claims
Unverified合成数据的反馈循环会导致模型崩溃，表现为词汇单一化、句子结构趋同、长尾内容消失69% similarUnverified用新数据更新模型时可能破坏原有能力，即灾难性遗忘，工业界通过失败案例库、数据增强与混合回放技术来整合新数据68% similarUnverified端到端模型在代表性充足与代表性不足口音之间的WER差距并未按比例缩小，属于数据分布问题而非架构问题67% similarUnverifiedInference methods based on model internal features face a trade-off between false positives and false negatives65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/731905API
curl https://kongchang.com/api/v1/knowledge/claims/731905MCP
get_claim(id=731905)