Unverified50% confidenceOpinionExact time
When validating ASR post-editing methods on public benchmarks, one needs to be wary of inflated performance caused by pre-training data overlap.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
Unverified训练数据在输入模型前会经过严格的去重、过滤和质量筛选，低质量内容、重复文本、有害内容会被剔除73% similarUnverified在GPU训练前进行前置的数据质检可以节省训练成本并排除因数据问题导致的实验结论偏差72% similarUnverifiedDuring fine-tuning, the prompt format in the training phase must strictly match the inference phase; otherwise, performance degradation will occur without any error being reported.72% similarUnverifiedGlobal classification is not suitable for local tampering detection, and in small-data scenarios, self-supervised pre-training and domain priors are often more effective than larger models.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1028271API
curl https://kongchang.com/api/v1/knowledge/claims/1028271MCP
get_claim(id=1028271)