Altered Thoughts, Altered Actions: Probing Chain-of-Thought Vulnerabilities in VLA Robotic Manipulation

• URL: http://arxiv.org/abs/2603.12717v1
• Date: Fri, 13 Mar 2026 07:02:51 GMT
• Title: Altered Thoughts, Altered Actions: Probing Chain-of-Thought Vulnerabilities in VLA Robotic Manipulation
• Abstract summary: Recent Vision-Language-Action (VLA) models increasingly adopt chain-of-thought (CoT) reasoning, generating a natural-language plan before decoding motor commands.<n>We ask: which properties of this intermediate plan does the action decoder actually rely on, and can targeted corruption of the reasoning trace alone degrade a robot's physical task performance?<n>We design a taxonomy of seven text corruptions organized into three attacker tiers (blind noise, mechanical-semantic, and LLM-adaptive) and apply them to a state-of-the-art reasoning VLA across 40 tabletop manipulation tasks.
• Score: 27.2709460237741
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Recent Vision-Language-Action (VLA) models increasingly adopt chain-of-thought (CoT) reasoning, generating a natural-language plan before decoding motor commands. This internal text channel between the reasoning module and the action decoder has received no adversarial scrutiny. We ask: which properties of this intermediate plan does the action decoder actually rely on, and can targeted corruption of the reasoning trace alone -- with all inputs left intact -- degrade a robot's physical task performance? We design a taxonomy of seven text corruptions organized into three attacker tiers (blind noise, mechanical-semantic, and LLM-adaptive) and apply them to a state-of-the-art reasoning VLA across 40 LIBERO tabletop manipulation tasks. Our results reveal a striking asymmetry: substituting object names in the reasoning trace reduces overall success rate by 8.3~percentage points (pp) -- reaching $-$19.3~pp on goal-conditioned tasks and $-$45~pp on individual tasks -- whereas sentence reordering, spatial-direction reversal, token noise, and even a 70B-parameter LLM crafting plausible-but-wrong plans all have negligible impact (within $\pm$4~pp). This asymmetry indicates that the action decoder depends on entity-reference integrity rather than reasoning quality or sequential structure. Notably, a sophisticated LLM-based attacker underperforms simple mechanical object-name substitution, because preserving plausibility inadvertently retains the entity-grounding structure the decoder needs. A cross-architecture control using a non-reasoning VLA confirms the vulnerability is exclusive to reasoning-augmented models, while instruction-level attacks degrade both architectures -- establishing that the internal reasoning trace is a distinct and stealthy threat vector invisible to input-validation defenses.

Related papers

• Grounding SWE-Agent Decisions in Architecture-0 Design: Navigating Unknown Unknowns through Physical Mapping [4.754027236572029]
  We investigate how Autonomous Software Engineering Agents (SWE-Agents) navigate Unknown Unknowns (UUs)<n>UUs are the nascent system design phase plagued by implicit engineering constraints, or Unknown Unknowns (UUs)<n>We propose the Physical Mapping Guard (PMG) to eradicate physical-layer and validation-layer gaming.
  arXiv  Detail & Related papers  (2026-09-15T14:04:33Z)
• TrapVLA: Trapping Vision-Language-Action Models in Configured Failure Modes [65.10128145059127]
  This work introduces Configured Failure Trapping, a novel backdoor attack task against Vision-Language-Action (VLA) models.<n>We propose an effective data engine for synthesizing high-quality target trajectories and an automated suite for measuring configured-failure fidelity.<n>Experiments show that TrapVLA effectively injects configured failure modes into VLA models while largely preserving performance on clean data.
  arXiv  Detail & Related papers  (2026-08-27T03:44:49Z)
• VASO: Formally Verifiable Self-Evolving Skills for Physical AI Agents [57.240036084348354]
  We introduce VASO, a framework for verification-guided self-evolution of robot skill contracts.<n>VASO verifies logically inconsistent skill contracts, then verifies plans induced by the skill against global and local temporal specifications.<n>On Clearpath Jackal and PX4 quadcopter tasks, VASO reaches 97.2% formal-specification compliance using fewer than 100 optimization samples.
  arXiv  Detail & Related papers  (2026-06-03T20:02:35Z)
• Imbuing Large Language Models with Bidirectional Logic for Robust Chain Repair [44.80087038178069]
  We introduce Teleological Reasoning Infilling (TRI), a training framework that endows decoder-only transformers with a native emphgoal-conditioned bridging capability.<n>At inference, TRI operates as a surgical repair module within a dual-system loop.<n>Experiments on three benchmarks demonstrate that TRI achieves state-of-the-art performance across all tasks, while reducing per-problem token expenditure by 31.2%.
  arXiv  Detail & Related papers  (2026-06-03T15:58:48Z)
• Provably Secure Agent Guardrail [89.79561918065122]
  Existing defense architectures rely on empirical semantic guardrails and probabilistic large model adjudicators.<n>This paper proposes a new security paradigm for agents based on the fundamental limitations of logical reasoning.
  arXiv  Detail & Related papers  (2026-05-28T02:12:41Z)
• PRISM: : Planning and Reasoning with Intent in Simulated Embodied Environments [59.07829883257003]
  Built on five multi-room apartments, PRISM structures 300 human-verified tasks into three capability tiers.<n>PRISM exposes an agent-agnostic executable action API that allows arbitrary agents to be evaluated end-to-end.
  arXiv  Detail & Related papers  (2026-05-12T04:59:47Z)
• TRAP: Hijacking VLA CoT-Reasoning via Adversarial Patches [16.543720979135397]
  Chain-of-Thought(CoT) reasoning has demonstrated strong capabilities in robotic manipulation.<n>We show that CoT reasoning introduces a novel attack vector for targeted control hijacking.<n>We propose TRAP, the first targeted adversarial attack framework for CoT-reasoning Vision-Language-Action models.
  arXiv  Detail & Related papers  (2026-03-24T12:14:12Z)
• Dynamic analysis enhances issue resolution [53.50448142467294]
  DAIRA (Dynamic Analysis-enhanced Issue Resolution Agent) is an automated repair framework that embeds dynamic analysis into the agent's reasoning cycle.<n> Driven by a Test Tracing-Driven methodology, DAIRA utilizes lightweight monitors to extract critical runtime data.<n>When powered by Gemini 3 Flash Preview, DAIRA establishes a new state-of-the-art (SOTA) performance, achieving a 79.4% resolution rate on the SWE-bench Verified dataset.
  arXiv  Detail & Related papers  (2026-03-23T14:48:54Z)
• Lying to Win: Assessing LLM Deception through Human-AI Games and Parallel-World Probing [16.419761149171215]
  We introduce a logically grounded framework to elicit and quantify deceptive behavior by embedding Large Language Models.<n>Our method employs a conversational forking mechanism: at the point of object identification, the dialogue state is duplicated into multiple parallel worlds.<n>We evaluate GPT-4o, Gemini-2.5-Flash, and Qwen-3-235B across three incentive levels: neutral, loss-based, and existential.
  arXiv  Detail & Related papers  (2026-03-07T13:21:53Z)
• TraceGuard: Process-Guided Firewall against Reasoning Backdoors in Large Language Models [19.148124494194317]
  We propose TraceGuard, a process-guided security framework that transforms small-scale models into robust reasoning firewalls.<n>Our approach treats the reasoning trace as an untrusted payload and establishes a defense-in-depth strategy.<n>We demonstrate robustness against adaptive adversaries in a grey-box setting, establishing TraceGuard as a viable, low-latency security primitive.
  arXiv  Detail & Related papers  (2026-03-02T22:19:13Z)
• Sponge Tool Attack: Stealthy Denial-of-Efficiency against Tool-Augmented Agentic Reasoning [58.432996881401415]
  Recent work augments large language models (LLMs) with external tools to enable agentic reasoning.<n>We propose Sponge Tool Attack (STA), which disrupts agentic reasoning solely by rewriting the input prompt.<n>STA generates benign-looking prompt rewrites from the original one with high semantic fidelity.
  arXiv  Detail & Related papers  (2026-01-24T19:36:51Z)
• VIRO: Robust and Efficient Neuro-Symbolic Reasoning with Verification for Referring Expression Comprehension [51.76841625486355]
  Referring Expression (REC) aims to localize the image region corresponding to a natural-language query.<n>Recent neuro-symbolic REC approaches leverage large language models (LLMs) and vision-language models (VLMs) to perform compositional reasoning.<n>We introduce VIRO, a neuro-symbolic framework that embeds lightweight operator-level verifiers within reasoning steps.
  arXiv  Detail & Related papers  (2026-01-19T07:21:19Z)
• DeepThinkVLA: Enhancing Reasoning Capability of Vision-Language-Action Models [51.76664843721462]
  DeepThinkVLA is a new architecture for Vision-Language-Action models.<n>It generates sequential CoT with causal attention and switches to bidirectional attention for fast decoding of action vectors.<n>It achieves a 97.0% success rate on the LIBERO benchmark.
  arXiv  Detail & Related papers  (2025-10-31T05:26:16Z)
• Model-agnostic Adversarial Attack and Defense for Vision-Language-Action Models [25.45513133247862]
  Vision-Language-Action (VLA) models have achieved revolutionary progress in robot learning.<n>Despite this progress, their adversarial robustness remains underexplored.<n>We propose both adversarial patch attack and corresponding defense strategies for VLA models.
  arXiv  Detail & Related papers  (2025-10-15T07:42:44Z)
• Countermind: A Multi-Layered Security Architecture for Large Language Models [0.0]
  This paper proposes Countermind, a multi-layered security architecture intended to shift defenses from a reactive, post hoc posture to a proactive, pre-inference, and intra-inference enforcement model.<n>The architecture proposes a fortified perimeter designed to structurally validate and transform all inputs, and an internal governance mechanism intended to constrain the model's semantic processing pathways before an output is generated.
  arXiv  Detail & Related papers  (2025-10-13T18:41:18Z)
• Chain-of-Code Collapse: Reasoning Failures in LLMs via Adversarial Prompting in Code Generation [0.3495246564946556]
  Large Language Models (LLMs) have achieved remarkable success in tasks requiring complex reasoning.<n>Do these models truly reason, or do they merely exploit shallow statistical patterns?<n>We introduce Chain-of-Code Collapse, where we investigate the robustness of reasoning LLMs by introducing a suite of semantically faithful yet adversarially structured prompt perturbations.
  arXiv  Detail & Related papers  (2025-06-08T02:43:46Z)
• ShadowCode: Towards (Automatic) External Prompt Injection Attack against Code LLMs [56.46702494338318]
  This paper introduces a new attack paradigm: (automatic) external prompt injection against code-oriented large language models.<n>We propose ShadowCode, a simple yet effective method that automatically generates induced perturbations based on code simulation.<n>We evaluate our method across 13 distinct malicious objectives, generating 31 threat cases spanning three popular programming languages.
  arXiv  Detail & Related papers  (2024-07-12T10:59:32Z)
• Advancing Generalized Transfer Attack with Initialization Derived Bilevel Optimization and Dynamic Sequence Truncation [49.480978190805125]
  Transfer attacks generate significant interest for black-box applications. Existing works essentially directly optimize the single-level objective w.r.t. surrogate model. We propose a bilevel optimization paradigm, which explicitly reforms the nested relationship between the Upper-Level (UL) pseudo-victim attacker and the Lower-Level (LL) surrogate attacker.
  arXiv  Detail & Related papers  (2024-06-04T07:45:27Z)
• Alternating Objectives Generates Stronger PGD-Based Adversarial Attacks [78.2700757742992]
  Projected Gradient Descent (PGD) is one of the most effective and conceptually simple algorithms to generate such adversaries. We experimentally verify this assertion on a synthetic-data example and by evaluating our proposed method across 25 different $ell_infty$-robust models and 3 datasets. Our strongest adversarial attack outperforms all of the white-box components of AutoAttack ensemble.
  arXiv  Detail & Related papers  (2022-12-15T17:44:31Z)

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.