Robotics · 2026-10-14
CAG-WAM: Causal Action-Gain World Action Models for Robot Manipulation
Peizheng Guo, Cheng Zhang, Xingyu Zhang, Changwen Zheng, Wenwen Qiang
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
CAG-WAM analyzes action evaluation in World Action Models (WAMs) via a structural causal model that separates action-induced change from environment-driven evolution, including a backdoor path from current observation to action outcomes. For each candidate action, it predicts an action-conditioned future and an environmental future, and uses their evaluation difference as causal action gain (CAG) for training (GRPO) and inference-time action revision.
Why it matters
The authors argue that evaluating a predicted future alone can mix action effects with environment evolution, so overall scores may not reflect the causal effect of each candidate action. CAG-WAM provides a paired-future, intervention/backdoor-adjustment-inspired mechanism (CAG) to guide policy learning and to revise actions when needed.
Method
- Formulate action evaluation in WAMs with a structural causal model that includes environment-driven changes and a backdoor path via the current observation.
- For each candidate action, use PF-WM to predict both an action-conditioned future and an environmental future under the same current observation; define their evaluation difference as causal action gain (CAG).
- Train by using CAG as the advantage in group relative policy optimization (GRPO), and at inference revise actions when CAG falls below a revision threshold using predicted-future feedback.
Abstract (from arXiv)
World Action Models (WAMs) combine action generation with future prediction, allowing predicted futures to support action evaluation and robot decision making. However, the future associated with a candidate action may contain action-induced and environment-driven changes. Directly evaluating the predicted future may not fully characterize the future change associated with the candidate action. For example, a candidate action that worsens a favorable environmental future may still yield a favorable outcome, while one that improves an unfavorable environmental future may still yield an unfavorable outcome. To analyze this issue, we formulate a Structural Causal Model (SCM) for action evaluation in WAMs. The SCM formalizes how action-induced and environment-driven changes jointly shape future evaluation, and reveals a backdoor path through which environmental evolution can also affect the future evaluation associated with a candidate action. Intervention and backdoor adjustment account for this path, providing a causal basis for characterizing the future change associated with each candidate action. Based on this analysis, we propose Causal Action-Gain World Action Model (CAG-WAM). For each candidate action, CAG-WAM predicts an action-conditioned future and an environmental future under the same current observation, and defines their evaluation difference as causal action gain (CAG) for characterizing the future change associated with the candidate action. During training, we incorporate CAG into Group Relative Policy Optimization (GRPO) to guide robot policy learning. During inference, we evaluate the initial action with CAG and, when needed, revise it using predicted future information. Experiments on multiple benchmarks support the effectiveness of CAG-WAM.
Related papers
- {\Delta}WAM: Distilling Action Tangent Fields into World Action Models
- OpenWAM: An Open Framework for Composable World-Action Models
- RAP-WAM: Learning World-Action Models through Predictive Rehearsal and Action Perturbations
- Causal world models under physical parameter shifts
- Long-WAM: Scaling the Context of World-Action Models
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-10 · Indexed by Pub-AI: 2026-10-14 UTC
CAG-WAM: Causal Action-Gain World Action Models for Robot Manipulation
Preprint posted: 2026-10-10 · Original record ↗
- arxiv
- 2610.14720
Explicit source versions (1)
- v1 · 2026-10-10
- Preprint available
Source date: 2026-10-10 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
World Action Models (WAMs) combine action generation with future prediction, allowing predicted futures to support action evaluation and robot decision making. However, the future associated with a candidate action may contain action-induced and environment-driven changes. Directly evaluating the predicted future may not fully characterize the future change associated with the candidate action. For example, a candidate action that worsens a favorable environmental future may still yield a favorable outcome, while one that improves an unfavorable environmental future may still yield an unfavorable outcome. To analyze this issue, we formulate a Structural Causal Model (SCM) for action evaluation in WAMs. The SCM formalizes how action-induced and environment-driven changes jointly shape future evaluation, and reveals a backdoor path through which environmental evolution can also affect the future evaluation associated with a candidate action. Intervention and backdoor adjustment account for this path, providing a causal basis for characterizing the future change associated with each candidate action. Based on this analysis, we propose Causal Action-Gain World Action Model (CAG-WAM). For each candidate action, CAG-WAM predicts an action-conditioned future and an environmental future under the same current observation, and defines their evaluation difference as causal action gain (CAG) for characterizing the future change associated with the candidate action. During training, we incorporate CAG into Group Relative Policy Optimization (GRPO) to guide robot policy learning. During inference, we evaluate the initial action with CAG and, when needed, revise it using predicted future information. Experiments on multiple benchmarks support the effectiveness of CAG-WAM.