Unverified50% confidenceFactExact time
Regret in multi-armed bandits is defined as T times the expected reward of the optimal arm after T rounds of play minus the total expected reward obtained by the actual policy.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedReward function design requires balancing the encouragement of forward racing against the penalization of collisions, and design bias may cause the agent to learn a degenerate strategy of remaining stationary60% similarUnverifiedThe multi-armed bandit (MAB) is a classic framework in reinforcement learning and online decision-making, in which a player faces K arms, each arm's reward follows an unknown distribution, and the goal is to minimize cumulative regret.59% similarUnverifiedClassic examples of reward hacking include reinforcement learning game agents discovering bugs in the physics engine and exploiting them in abnormal ways to inflate their scores.59% similarUnverifiedThe Prestige mechanic in idle games involves sacrificing all current progress in exchange for permanent multiplier bonuses, making each reset cycle progress faster than the last.58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1046778API
curl https://kongchang.com/api/v1/knowledge/claims/1046778MCP
get_claim(id=1046778)