# BadWAM: When World-Action Models Dream Right but Act Wrong

Page: https://stenobird.com/podcast/daily-paper-cast-7079649/badwam-when-world-action-models-dream-right-but-act-wrong
Text version: https://stenobird.com/podcast/daily-paper-cast-7079649/badwam-when-world-action-models-dream-right-but-act-wrong.md
Podcast: [Daily Paper Cast](https://stenobird.com/podcast/daily-paper-cast-7079649)
Published: 2026-07-18T03:57:58+00:00
Episode link: https://share.transistor.fm/s/abf24a34
Audio file: https://media.transistor.fm/abf24a34/cc988dbe.mp3
Processing state: not_requested
JSON: https://stenobird.com/v1/public/podcasts/daily-paper-cast-7079649/episodes/badwam-when-world-action-models-dream-right-but-act-wrong
Duration seconds: 1263

## Resource

🤗 Upvotes: 36 | cs.LG, cs.RO Authors: Qi Li, Xingyi Yang, Xinchao Wang Title: BadWAM: When World-Action Models Dream Right but Act Wrong Arxiv: http://arxiv.org/abs/2607.15207v1 Abstract: World-action models (WAMs) are emerging as a promising foundation for embodied control: rather than predicting actions alone, they learn representations that couple action generation with future world prediction. This coupling is often viewed as a source of robustness, interpretability, and safety, as a robot's action can in principle be checked against its imagined future. In this paper, we show that this assumption is fragile. We introduce BadWAM, a unified framework for modeling and evaluating World-Action Drift Attacks: a new class of WAM-specific adversarial attacks that use small visual perturbations to break the alignment between what a WAM imagines and what it executes. BadWAM characterizes this attack surface along two natural criteria: attack strength and stealthiness. When the adversary prioritizes disruption, BadWAM instantiates an action-only adversarial attack, which directly drives the model toward task-failing actions. When the adversary additionally prioritizes stealth, BadWAM instantiates an imagination-preserving adversarial attack, which seeks to induce harmful action shifts while keeping the model's predicted future close to its clean imagination. Together, these two attacks capture a spectrum of WAM-specific failures: from overt action hijacking to stealthier cases where the model appears to imagine a plausible future but executes a desynchronized action. We evaluate BadWAM across different variants of WAMs. Results show that our attacks substantially reduce task success rates under closed-loop execution. For example, our action-only attack reduces the model perf…

## Actions

- request_transcript: `POST https://stenobird.com/v1/public/podcasts/daily-paper-cast-7079649/episodes/badwam-when-world-action-models-dream-right-but-act-wrong/transcription-requests` — Idempotently request low-priority transcript generation for this episode.
- read_markdown: `GET https://stenobird.com/podcast/daily-paper-cast-7079649/badwam-when-world-action-models-dream-right-but-act-wrong.md` — Read the agent-friendly Markdown representation of this episode resource.

A page view does not enqueue transcription. Agents should invoke `request_transcript` explicitly when they need this episode processed.

## Transcript

Full transcripts are not published on public pages unless there is a clear rights basis.