Add Safety section + Embodied AI Safety Survey#7
Open
x-zheng16 wants to merge 1 commit into
Open
Conversation
WAMs as joint video-action generators / autoregressive predictors / diffusion planners inherit safety vulnerabilities at every embodied layer they integrate (perception fusion, cognitive planning, action control, agentic cascading). The Awesome-WAM list lacks a Safety section to surface this threat surface. Adds: - New top-level `## Safety` section between Evaluation and Citation - TOC entry `- [Safety](#safety)` - First entry: "Safety in Embodied AI: A Survey of Risks, Attacks, and Defenses" (arXiv:2605.02900) — 500+ paper survey across 5 layers / 18 subcategories, structured around Capability-Risk Duality (capability expansion = new vulnerabilities; inner-layer vulnerabilities cascade outward). Single-paragraph blurb explains relevance to WAM design (threat surface inherited by joint/cascaded WAMs at every layer they integrate). Open to PRs from the community to expand this section with additional safety-focused references.
There was a problem hiding this comment.
Pull request overview
Note
Copilot was unable to run its full agentic suite in this review.
Adds a Safety section to the repository README to highlight embodied AI safety considerations relevant to World Action Models (WAMs), and links to a recent survey.
Changes:
- Added a new Safety entry to the README table of contents.
- Added a new Safety section describing threat surfaces across the embodied AI pipeline and referencing an external survey/resource.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| @@ -67,6 +67,7 @@ This repository accompanies our survey on **World Action Models (WAMs)** — the | |||
| - [Evaluation](#evaluation) | |||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds a new
## Safetysection to surface the safety threat surface that World Action Models (joint video-action generators, autoregressive predictors, diffusion planners) inherit at every embodied layer they integrate. The list currently lacks a Safety section to organize this dimension of WAM research.What this PR adds
## Safetysection betweenEvaluationandCitation- [Safety](#safety)under## ContentsWhy this fits Awesome-WAM
The survey provides the first unified safety framework for embodied AI:
For WAM specifically: cascaded WAMs inherit perception-layer adversarial fragility; joint diffusion WAMs inherit cognitive jailbreak threats; autoregressive generators inherit planning-stage manipulation; all WAM variants inherit action-level safe-RL gaps and agentic cascading risks at deployment. The survey characterizes each of these threat layers with both attack and defense literature.
Test plan
## Safetysection renders correctly in GitHub markdown preview#safetyresolves to the new section