Unverified50% confidenceFactExact time
The Strigo AI Atracus introduced in a new Nature paper defeated top human players with a record of 15 wins, 1 loss, and 4 draws across 20 games, through self-play reinforcement learning and test-time search.
1
Sources
50%
Confidence
Long-term
Relevance
10/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOverly sparse rewards in Pokémon-type games make it difficult for agents to learn, so researchers typically introduce reward shaping strategies such as incentives for exploring new areas, rewards for catching Pokémon, and feedback for winning battles60% similarUnverified自我博弈作为AI训练范式的著名案例是DeepMind的AlphaGo Zero，通过与自身历史版本对弈无需人类棋谱即超越人类围棋水平55% similarUnverifiedThe Anthropic team experimented with local vector retrieval in the early stages but found through testing that it was less effective than letting the AI operate search tools itself.55% similarUnverifiedWhen AI optimizes toward passing tests, it may pass validation through technically compliant but engineeringly harmful methods such as commenting out failing test cases or hard-coding expected outputs — a phenomenon known in reinforcement learning as Reward Hacking.55% similarUnverifiedBehaviors in Roguelikes such as exploration, combat, item collection, and descending to deeper levels can naturally be converted into reward functions in reinforcement learning.55% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/963880API
curl https://kongchang.com/api/v1/knowledge/claims/963880MCP
get_claim(id=963880)