Unverified50% confidenceFactExact time
In Event2Vec, norm clipping, whenever the history vector is updated, if its L2 norm exceeds the threshold τ, shrinks it back to the sphere according to h_t ← h_t · τ / max(τ, ‖h_t‖).
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAfter truncating a vector, L2 normalization must be reapplied; otherwise, similarity ranking will silently produce incorrect results without raising any error.57% similarUnverifiedRobustness measures whether two LLMs can still be distinguished after progressively removing the most discriminative tokens.55% similarUnverifiedGradient clipping works by proportionally scaling down the entire gradient vector when its norm exceeds a preset threshold, so that its norm equals exactly that threshold.55% similarUnverified只有在LLM推理比embedding加向量检索慢数十倍以上时，语义缓存才能同时实现降本与提速55% similarUnverifiedWhen LLMs perform mental arithmetic on long-chain exponentiation (e.g., (1+r)^300), small rounding errors at each step are amplified along the exponential chain, causing EMI monthly payment deviations of tens of rupees, whereas errors in linear proportional calculations (area conversion) do not propagate.55% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1028349API
curl https://kongchang.com/api/v1/knowledge/claims/1028349MCP
get_claim(id=1028349)