---
title: "pico-WAM: Evaluating Counterfactual Prediction and Planning Error in a Compact World Action Model"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2611.01039v1"
permalink: "https://feed.astralor.com/items/itm_65c3d00e930b430ce152e211"
published: "2026-10-31T15:28:50.000Z"
score: 54
channel: "tech"
tags: ["RAG","WAM","Evaluating Counterfactual Prediction","Planning Error","Compact World Action","Model When","World Action Model","Usin"]
---

# pico-WAM: Evaluating Counterfactual Prediction and Planning Error in a Compact World Action Model

## 速览

When a World Action Model fails at control, is the bottleneck proposal generation or trajectory scoring? We introduce pico-WAM, a compact model with supplied physical state for tracing policy-dynamics interactions. Usin…

## 关键点

- When a World Action Model fails at control, is the bottleneck proposal generation or trajectory scoring?
- We introduce pico-WAM, a compact model with supplied physical state for tracing policy-dynamics interactions.
- Using 206 Push-T demonstrations and 15 training seeds, we compare data regimes and search configurations.

## 全文

When a World Action Model fails at control, is the bottleneck proposal generation or trajectory scoring? We introduce pico-WAM, a compact model with supplied physical state for tracing policy-dynamics interactions. Using 206 Push-T demonstrations and 15 training seeds, we compare data regimes and search configurations. Counterfactual branch training improves short-horizon cost ranking and enlarges the coverage advantage of policy-restarted over warm-started search by 4.8 percentage points; restart wins under both data regimes. Replacing learned trajectory scores with simulator costs improves coverage by 12.2 points while retaining learned proposals. Substituting simulator states in proposal generation alone shows no detected gain. Component diagnostics and a separate door-opening study show why better ranking need not yield better control or outperform imitation. These controlled interventions identify scoring as an effective improvement target in the tested setting and provide a reproducible protocol for diagnosing compact world models.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_65c3d00e930b430ce152e211