Unverified50% confidenceDecision RuleExact time
The loss function should ignore padding tokens (using ignore_index).
1
Sources
50%
Confidence
Long-term
Relevance
10/23/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSetting ignore_index in the loss function to point to the padding index can mask the contribution of padding positions to the loss; PyTorch's CrossEntropyLoss supports this parameter.77% similarUnverifiedIf cached tokens are always zero, first check and fix the prompt prefix, because any dynamic content in the prefix invalidates implicit caching; one tester saw a broken prefix inflate costs more than sixfold.64% similarUnverifiedWhen the data corresponding to a cursor is deleted, some implementations return an error rather than an empty result.63% similarUnverifiedWeak identifiers (such as cookies and browsing history) may become invalid due to clearing the cache or changing devices, whereas strong identifiers will not.58% similarUnverifiedToken metrics should be based on the authoritative usage field returned in the API response, with the application side only doing fallback estimation when that value is missing, and should distinguish "actual usage" from "estimated usage" in metric naming.57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1089780API
curl https://kongchang.com/api/v1/knowledge/claims/1089780MCP
get_claim(id=1089780)