Unverified50% confidenceFactExact time
Metrics based on pretrained embeddings, such as BERTScore, attempt to introduce semantic similarity but still operate at the sentence level and struggle to track whether specific factual claims in an answer are consistent with the context.
1
Sources
50%
Confidence
Long-term
Relevance
9/28/2026
First Seen
Sources
Related Entities
Related Claims
Unverified当输入存在歧义时LLM不会报错，而是沿训练数据中最高频的语义路径继续生成，用统计上最常见的答案填补需求空白68% similarUnverifiedThe study points out that aggregated accuracy metrics systematically obscure asymmetric capabilities in multilingual factual reasoning, which is a manifestation of Simpson's paradox.68% similarUnverifiedAutoregressive models suffer from the exposure bias problem, where the model sees real prefixes during training but faces its own erroneous generated prefixes during inference, leading to error accumulation.67% similarUnverifiedAfter adopting a frequency-aware training schedule (sampling rare tokens at a rate higher than uniform sampling), the model matches the parameter-matched BERT and TinyBERT baselines on the rarest token-frequency bin.67% similarUnverifiedWhen measurement instruments have systematic biases, the conclusions drawn may be misleading no matter how sophisticated the statistical methods are.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/955563API
curl https://kongchang.com/api/v1/knowledge/claims/955563MCP
get_claim(id=955563)