Embodied Red Teaming for Auditing Robotic Foundation Models

• URL: http://arxiv.org/abs/2411.18676v2
• Date: Mon, 10 Feb 2025 16:32:27 GMT
• Title: Embodied Red Teaming for Auditing Robotic Foundation Models
• Abstract summary: Embodied Red Teaming (ERT) generates diverse and challenging instructions to test language-conditioned robot models.<n>ERT uses automated red teaming techniques with Vision Language Models (VLMs) to create contextually grounded, difficult instructions.<n> Experimental results show that state-of-the-art language-conditioned robot models fail or behave unsafely on ERT-generated instructions.
• Score: 39.23336328126245
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Language-conditioned robot models have the potential to enable robots to perform a wide range of tasks based on natural language instructions. However, assessing their safety and effectiveness remains challenging because it is difficult to test all the different ways a single task can be phrased. Current benchmarks have two key limitations: they rely on a limited set of human-generated instructions, missing many challenging cases, and focus only on task performance without assessing safety, such as avoiding damage. To address these gaps, we introduce Embodied Red Teaming (ERT), a new evaluation method that generates diverse and challenging instructions to test these models. ERT uses automated red teaming techniques with Vision Language Models (VLMs) to create contextually grounded, difficult instructions. Experimental results show that state-of-the-art language-conditioned robot models fail or behave unsafely on ERT-generated instructions, underscoring the shortcomings of current benchmarks in evaluating real-world performance and safety. Code and videos are available at: https://s-karnik.github.io/embodied-red-team-project-page.

Related papers

• One Word, Different Action: A Real-Robot Benchmark for Language-Conditioned Embodied Reasoning [5.503904431371255]
  We introduce One Word, Different Action, a real-robot benchmark built on physical decision states and executable actions.<n>Experiments show that modern models are near saturation on single-constraint instruction changes, yet several models degrade noticeably when multiple task constraints must be integrated into one executable decision.
  arXiv  Detail & Related papers  (2026-09-04T15:21:58Z)
• RoboWits: Unexpected Challenges for Robotic Creative Problem Solving [58.58727722988084]
  We introduce RoboWits, a bi-manual robotic benchmark designed to evaluate cognitive reasoning, creative tool use, and robustness to unexpected conditions.<n>Using the pipeline, we curated 30 diverse seed tasks and 208 tasks with mutations and graded difficulty across geometry, material, and assembly-based reasoning.<n>Our results reveal a significant performance gap: while pre-trained VLAs exhibit preliminary success on seed tasks after single-task fine-tuning, they struggle to perform on mutated tasks.
  arXiv  Detail & Related papers  (2026-05-28T17:57:15Z)
• Uncovering Linguistic Fragility in Vision-Language-Action Models via Diversity-Aware Red Teaming [64.48633529149579]
  We propose a novel framework to expose the vulnerabilities of Vision-Language-Action (VLA) models against linguistic variations.<n>Our method reduces the average task success rate from 93.33% to 5.85%, demonstrating a scalable approach to stress-testing VLA agents.
  arXiv  Detail & Related papers  (2026-04-07T08:43:36Z)
• Red-Teaming Vision-Language-Action Models via Quality Diversity Prompt Generation for Robust Robot Policies [16.134245045349036]
  We present Q-DIG, which performs red-teaming by scalably identifying diverse natural language task descriptions.<n>Q-DIG finds more diverse and meaningful failure modes compared to baseline methods.<n>Results from a user study highlight that Q-DIG generates prompts judged to be more natural and human-like than those from baselines.
  arXiv  Detail & Related papers  (2026-03-12T22:58:42Z)
• CORAL: Scalable Multi-Task Robot Learning via LoRA Experts [49.759823970016974]
  Vision-Language-Action (VLA) models in real-world robotics expose a core multi-task learning challenge.<n>Maintaining a separate full checkpoint per task is often storage- and deployment-prohibitive.<n>We present CORAL, a backbone- and-agnostic framework designed primarily to mitigate multi-task interference.
  arXiv  Detail & Related papers  (2026-03-10T07:28:41Z)
• Metamorphic Testing of Vision-Language Action-Enabled Robots [12.232630336816463]
  We propose Metamorphic Testing (MT) to alleviate the test oracle problem in Vision-Language-Action (VLA) models.<n>An empirical study involving five VLA models, two simulated robots, and four robotic tasks shows that MT can effectively alleviate the test oracle problem.
  arXiv  Detail & Related papers  (2026-02-26T03:32:43Z)
• Heterogeneous Robot Collaboration in Unstructured Environments with Grounded Generative Intelligence [54.91177026001217]
  Large language model (LLM)-enabled teaming methods typically assume well-structured and known environments.<n>We present SPINE-HT, a framework that addresses these limitations by grounding the reasoning abilities of LLMs in the context of a heterogeneous robot team.<n>Our framework achieves nearly twice the success rate compared to prior LLM-enabled heterogeneous teaming approaches.
  arXiv  Detail & Related papers  (2025-10-30T18:24:38Z)
• Automated Red Teaming with GOAT: the Generative Offensive Agent Tester [8.947465706080523]
  Red teaming assesses how large language models can produce content that violates norms, policies, and rules set during their safety training. Most existing automated methods in the literature are not representative of the way humans tend to interact with AI models. We introduce Generative Offensive Agent Tester (GOAT), an automated agentic red teaming system that simulates plain language adversarial conversations.
  arXiv  Detail & Related papers  (2024-10-02T14:47:05Z)
• RACER: Rich Language-Guided Failure Recovery Policies for Imitation Learning [19.023560632891467]
  We propose a scalable data generation pipeline that augments expert demonstrations with failure recovery trajectories. We then introduce Rich languAge-guided failure reCovERy (RACER), a supervisor-actor framework. Our experimental results show that RACER outperforms the state-of-the-art Robotic View Transformer on RLbench.
  arXiv  Detail & Related papers  (2024-09-23T02:50:33Z)
• Alice in Wonderland: Simple Tasks Showing Complete Reasoning Breakdown in State-Of-the-Art Large Language Models [13.532180752491954]
  We demonstrate a dramatic breakdown of function and reasoning capabilities of state-of-the-art models trained at the largest available scales. The breakdown is dramatic, as models show strong fluctuations across even slight problem variations that should not affect problem solving. We take these initial observations to stimulate urgent re-assessment of the claimed capabilities of current generation of Large Language Models.
  arXiv  Detail & Related papers  (2024-06-04T07:43:33Z)
• Large Language Models for Orchestrating Bimanual Robots [19.60907949776435]
  We present LAnguage-model-based Bimanual ORchestration (LABOR) to analyze task configurations and devise coordination control policies. We evaluate our method through simulated experiments involving two classes of long-horizon tasks using the NICOL humanoid robot.
  arXiv  Detail & Related papers  (2024-04-02T15:08:35Z)
• Yell At Your Robot: Improving On-the-Fly from Language Corrections [84.09578841663195]
  We show that high-level policies can be readily supervised with human feedback in the form of language corrections. This framework enables robots not only to rapidly adapt to real-time language feedback, but also incorporate this feedback into an iterative training scheme.
  arXiv  Detail & Related papers  (2024-03-19T17:08:24Z)
• Interactive Planning Using Large Language Models for Partially Observable Robotics Tasks [54.60571399091711]
  Large Language Models (LLMs) have achieved impressive results in creating robotic agents for performing open vocabulary tasks. We present an interactive planning technique for partially observable tasks using LLMs.
  arXiv  Detail & Related papers  (2023-12-11T22:54:44Z)
• FLIRT: Feedback Loop In-context Red Teaming [79.63896510559357]
  We propose an automatic red teaming framework that evaluates a given black-box model and exposes its vulnerabilities. Our framework uses in-context learning in a feedback loop to red team models and trigger them into unsafe content generation.
  arXiv  Detail & Related papers  (2023-08-08T14:03:08Z)
• Instruct2Act: Mapping Multi-modality Instructions to Robotic Actions with Large Language Model [63.66204449776262]
  Instruct2Act is a framework that maps multi-modal instructions to sequential actions for robotic manipulation tasks. Our approach is adjustable and flexible in accommodating various instruction modalities and input types. Our zero-shot method outperformed many state-of-the-art learning-based policies in several tasks.
  arXiv  Detail & Related papers  (2023-05-18T17:59:49Z)
• ICE-Score: Instructing Large Language Models to Evaluate Code [7.556444391696562]
  We propose textttICE-Score, a new evaluation metric via instructing large language models for code assessments. Our metric addresses the limitations of existing approaches by achieving superior correlations with functional correctness and human preferences. Our results demonstrate that our metric surpasses state-of-the-art metrics for code generation.
  arXiv  Detail & Related papers  (2023-04-27T16:38:17Z)
• ARNOLD: A Benchmark for Language-Grounded Task Learning With Continuous States in Realistic 3D Scenes [72.83187997344406]
  ARNOLD is a benchmark that evaluates language-grounded task learning with continuous states in realistic 3D scenes. ARNOLD is comprised of 8 language-conditioned tasks that involve understanding object states and learning policies for continuous goals.
  arXiv  Detail & Related papers  (2023-04-09T21:42:57Z)
• CodeLMSec Benchmark: Systematically Evaluating and Finding Security Vulnerabilities in Black-Box Code Language Models [58.27254444280376]
  Large language models (LLMs) for automatic code generation have achieved breakthroughs in several programming tasks. Training data for these models is usually collected from the Internet (e.g., from open-source repositories) and is likely to contain faults and security vulnerabilities. This unsanitized training data can cause the language models to learn these vulnerabilities and propagate them during the code generation procedure.
  arXiv  Detail & Related papers  (2023-02-08T11:54:07Z)
• VIMA: General Robot Manipulation with Multimodal Prompts [82.01214865117637]
  We show that a wide spectrum of robot manipulation tasks can be expressed with multimodal prompts. We develop a new simulation benchmark that consists of thousands of procedurally-generated tabletop tasks. We design a transformer-based robot agent, VIMA, that processes these prompts and outputs motor actions autoregressively.
  arXiv  Detail & Related papers  (2022-10-06T17:50:11Z)
• Grounding Language with Visual Affordances over Unstructured Data [26.92329260907805]
  We propose a novel approach to efficiently learn language-conditioned robot skills from unstructured, offline and reset-free data. We exploit a self-supervised visuo-lingual affordance model, which requires as little as 1% of the total data with language. We find that our method is capable of completing long-horizon, multi-tier tasks in the real world, while requiring an order of magnitude less data than previous approaches.
  arXiv  Detail & Related papers  (2022-10-04T21:16:48Z)
• What Matters in Language Conditioned Robotic Imitation Learning [26.92329260907805]
  We study the most critical challenges in learning language conditioned policies from offline free-form imitation datasets. We present a novel approach that significantly outperforms the state of the art on the challenging language conditioned long-horizon robot manipulation CALVIN benchmark.
  arXiv  Detail & Related papers  (2022-04-13T08:45:32Z)
• Adversarial GLUE: A Multi-Task Benchmark for Robustness Evaluation of Language Models [86.02610674750345]
  Adversarial GLUE (AdvGLUE) is a new multi-task benchmark to explore and evaluate the vulnerabilities of modern large-scale language models under various types of adversarial attacks. We apply 14 adversarial attack methods to GLUE tasks to construct AdvGLUE, which is further validated by humans for reliable annotations. All the language models and robust training methods we tested perform poorly on AdvGLUE, with scores lagging far behind the benign accuracy.
  arXiv  Detail & Related papers  (2021-11-04T12:59:55Z)
• Learning Language-Conditioned Robot Behavior from Offline Data and Crowd-Sourced Annotation [80.29069988090912]
  We study the problem of learning a range of vision-based manipulation tasks from a large offline dataset of robot interaction. We propose to leverage offline robot datasets with crowd-sourced natural language labels. We find that our approach outperforms both goal-image specifications and language conditioned imitation techniques by more than 25%.
  arXiv  Detail & Related papers  (2021-09-02T17:42:13Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.