研究显示：世界模型的回报误差集中在事件转折点而非像素误差
Events, Not Pixels: Where World Models Get the Return Wrong
这篇论文发现世界模型出错的地方很反直觉：不是画面糊的步骤，而是拿奖励的那几步。做规划或 RL 的人可以看看这个诊断思路。
arXiv 论文提出一个 transition 级诊断方法，用真实转移替换想象转移来定位世界模型规划中的回报误差来源。在 44 组通过可靠性检查的 Atari-100k 模型-环境对（含 3 个已发布模型家族）上，修复事件转移可消除中位数 70% 的回报误差，而按 MSE、SSIM 或 LPIPS 最大误差修复的效果还不如随机选择。事件步骤的像素渲染质量与其他步骤相当（中位像素误差比 0.85），说明视觉保真度并不能代表回报判断的准确性。事件补丁在 39/44 组中优于随机补丁（sign test p=1.4×10⁻⁷），中位增益 +44 个百分点。
Events, Not Pixels: Where World Models Get the Return Wrong
World models are increasingly used for planning, where a planner ranks candidate actions by returns imagined by the model. However, these models are judged mainly by how well they render frames, even though a visually faithful rollout can still assign the wrong return. To identify where this mismatch arises, we introduce a transition-level diagnostic that replaces imagined transitions with real ones and measures the reduction in return error against a budget-matched random control. Our analysis yields two findings. (1) Event transitions, the few steps where a reward is earned or an episode ends, are rendered as well as other steps (median pixel-error ratio 0.85), yet restoring them removes a median of 70% of the return error. In contrast, restoring transitions with the largest MSE, SSIM or LPIPS error removes less than a random choice. Across the 44 Atari-100k model--collection pairs that pass our reliability checks, including three released families, event patches outperform random patches in 39 (sign test p=1.4 x 10^{-7}), with a median gain of +44 percentage points. (2) The remaining error is structured: a complementary patch separates errors at true events from those elsewhere, while an eight-step timing tolerance recovers few events missed by our deterministic models. This structure also guides repair: a rule recorded before four repair outcomes were known predicts when a diffusion model repairs these failures. Event patches improve plan ranking, and the same concentration appears on control tasks with rare reward events.
