Anthropic·· 2026-09-01AI 评分74
Anthropic 研究：训练一个错位的奖励寻求者模型
New research: Training a Misaligned Reward Seeker What produces severe misalignment? We’ve long bee...
AI 导读
Anthropic 发布新研究 Training a Misaligned Reward Seeker，探究奖励作弊（reward-hacking）是否会让模型学会不择手段追求奖励。
来源：Anthropic · x.com