Driving · 2026-11-13
Bind-WAM: Binding-Uncertainty-Guided Selective Future Imagination in World Action Models
Chenhao Qiu, Shuyi Chen, Songen Gu, Shufeng Nan, Tianyu Wang, Siyu Lin, Yifan Li, Bingwen Zhu, Yanwei Fu
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
Bind-WAM is a world action model that imagines future only when needed. It uses language-object binding uncertainty—supervised by an offline labeling pipeline that detects reference mismatch, invisibility, and identity drift—to route between direct action prediction and joint future-action denoising. Simulation and real-world experiments show binding uncertainty as an interpretable signal for selective future imagination, with reduced task completion time.
Why it matters
Generating a visual future at every replanning step is expensive and sometimes unnecessary. The authors frame selective future imagination in World Action Models as a state-dependent routing problem and propose language-object binding uncertainty as an interpretable signal to decide when joint future-action generation is useful. This aims to reduce computation while preserving task success, and the paper reports improvements in success and completion time on simulation and real-world manipulation tasks.
Method
- Offline construct pseudo-labels for binding uncertainty from demonstrations using Qwen3-VL-32B grounding, SAM 3.1 instance masks, and DINOv3 identity consistency; labels capture reference mismatch, invisibility, and identity drift.
- Train a lightweight ViT-like predictor to output frame-level binding uncertainty U_t from the current RGB observation and language instruction.
- At inference, route a shared WAM between direct action prediction and joint future-action denoising based on U_t and its change ΔU_t between consecutive replanning steps.
Limitation
The authors plan to explore incorporating task-outcome feedback into router training to better identify when future imagination improves execution.
Abstract (from arXiv)
World Action Models (WAMs) can improve robot learning by jointly predicting future visual states and actions, but generating a visual future at every replanning step is expensive and sometimes unnecessary. This raises a critical question: when should a robot policy imagine the future? In this paper, we propose Bind-WAM, a binding-uncertainty-guided WAM for selective future imagination. Bind-WAM introduces language-object binding uncertainty, selectively invoking joint future-action generation when the current observation does not support stable task-relevant action generation. An offline labeling pipeline uses complete demonstrations and action context to identify task-relevant objects and detect reference mismatch, invisibility, and identity drift. The resulting pseudo-labels supervise a lightweight router that predicts binding uncertainty from the current observation and language instruction. At inference, a shared WAM policy invokes selective future imagination according to the predicted uncertainty and its change between consecutive replanning steps. Simulation and real-world experiments demonstrate that Bind-WAM can establish binding uncertainty as a strong, interpretable signal for selective future imagination.
Related papers
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First announced by arXiv: 2026-11-13 · Indexed by Pub-AI: 2026-11-13 UTC
Bind-WAM: Binding-Uncertainty-Guided Selective Future Imagination in World Action Models
First announced by arXiv: 2026-11-13 · Original record ↗
- arxiv
- 2611.15368
Explicit source versions (1)
- v1 · Date not supplied
- Preprint available
Source date: 2026-11-13 · First detected: 2026-11-13 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · RSS guid=oai:arXiv.org:2611.15368v1; arxiv:announce_type=new; item pubDate (official metadata)
Abstract · arxiv
World Action Models (WAMs) can improve robot learning by jointly predicting future visual states and actions, but generating a visual future at every replanning step is expensive and sometimes unnecessary. This raises a critical question: when should a robot policy imagine the future? In this paper, we propose Bind-WAM, a binding-uncertainty-guided WAM for selective future imagination. Bind-WAM introduces language-object binding uncertainty, selectively invoking joint future-action generation when the current observation does not support stable task-relevant action generation. An offline labeling pipeline uses complete demonstrations and action context to identify task-relevant objects and detect reference mismatch, invisibility, and identity drift. The resulting pseudo-labels supervise a lightweight router that predicts binding uncertainty from the current observation and language instruction. At inference, a shared WAM policy invokes selective future imagination according to the predicted uncertainty and its change between consecutive replanning steps. Simulation and real-world experiments demonstrate that Bind-WAM can establish binding uncertainty as a strong, interpretable signal for selective future imagination.
