Unverified50% confidenceFactExact time
On unseen words and sentences, category purity increased, and this improvement generalized to samples not seen during training.
1
Sources
50%
Confidence
Long-term
Relevance
11/17/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedA bag-of-words classifier trained on the same calls performed close to the activation probe, indicating that most of the gain comes from labeled training rather than from the activations themselves.73% similarUnverifiedEmpirical results show that stronger context and category–feature co-occurrence likelihood (PMI) improve learning in larger models, whereas verbatim rule-style statements weaken flexible category learning.71% similarUnverified更宽泛的分词器扩大了词表规模或覆盖范围,会增加嵌入层参数量和内存占用并提高训练成本70% similarUnverifiedUsing nonsense words as prompts can guide the model into regions of the latent space that are less frequently activated in training data, thereby producing more diverse results.70% similarUnverifiedIncremental pre-training uses unlabeled plain-text corpora, and the training objective is still predicting the next token; in essence it is continued pre-training.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1179413API
curl https://kongchang.com/api/v1/knowledge/claims/1179413MCP
get_claim(id=1179413)