Unverified50% confidenceBenchmarkExact time
The algorithm's regret upper bound is O(√(L* log N) + √(T log N)/ε), where ε is the privacy parameter.
1
Sources
50%
Confidence
Long-term
Relevance
10/22/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedFor adaptively generated input (where the input may depend on the algorithm's previous outputs), even with L*=0, any differentially private algorithm cannot avoid Ω̃(√T) regret.76% similarUnverifiedThe previously best differentially private online expert prediction algorithm had a regret bound of Õ(√T); when L* is much smaller than T, this bound is far higher than the non-private lower bound Õ(√L*).71% similarUnverifiedThe Littlestone dimension is finite if and only if a no-regret online learning algorithm exists, with a mistake bound of O(d log T).67% similarUnverifiedConstrained decoding can only constrain the legality of the token sequence but cannot constrain semantics; models can still output incorrect enum values or out-of-range numbers.58% similarUnverifiedThe core formula of Reddit's Hot sorting algorithm is approximately: score = log10(max(|ups-downs|, 1)) + sign(ups-downs) × epoch_seconds / 45000.57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1084204API
curl https://kongchang.com/api/v1/knowledge/claims/1084204MCP
get_claim(id=1084204)