ReCast is a plug‑in framework that protects private inputs for fixed‑interface multimodal reasoning by locally converting them into a shared textual evidence‑query record, rewriting entities and topics with a distilled model, and mapping numerical values through an invertible, role‑aware map. A reconstruction agent then generates the required media from this protected record, allowing a remote solver to return a program whose operands are restored locally before execution. On 4,000 held‑out ChartQA and NMSQA examples, ReCast achieves 75.10% accuracy, retaining 92.43% of the unprotected remote accuracy, and flags source‑content leakage in 7.95% of solver‑bound requests, outperforming all evaluated local baselines.
Remote multimodal models offer strong numerical reasoning capabilities over charts and speech, but sending private inputs risks exposing sensitive content. Text-only sanitization cannot directly satis...
arXiv:2610. 01871v1 Announce Type: cross Abstract: Multimodal Retrieval-Augmented Generation (MRAG) has emerged as a reliable and cost-effective technique of grounding the generative capabilities of Multimodal Large Language Models (MLLMs) into relevant, up-to-date, external knowledge.
arXiv:2606. 24623v1 Announce Type: cross Abstract: Retrieval-Augmented Generation enhances large language models by incorporating external knowledge, but deploying it in sensitive scenarios risks privacy leakage via malicious prompts.
Multimodal agents for visual question answering increasingly operate as multi-step trajectories that interleave perception, retrieval, and reasoning, yet evaluation still largely reduces to final-answer accuracy. This aggregate signal cannot tell whether a correct answer was reached through grounded evidence, language priors, or accidental error cancellation.
arXiv:2607. 28374v1 Announce Type: new Abstract: Multimodal agents for visual question answering increasingly operate as multi-step trajectories that interleave perception, retrieval, and reasoning, yet evaluation still largely reduces to final-answer accuracy.
arXiv:2608. 11691v1 Announce Type: new Abstract: Reinforcement-learning (RL) post-training equips multimodal large reasoning models (MLRMs) with exploratory chains of thought (CoT), substantially improving visual reasoning.
UnifiedAttack introduces a benchmark for testing the safety of large multimodal models (LMMs) in tasks that combine text and image to produce harmful content. The benchmark focuses on the additional harm that arises from cross‑modal synergy and includes filtered multimodal samples and synthesized disinformation queries. A synergistic hijacking framework—comprising In‑Context Reskinning (ICR) and Cognitive Planning Injection (CPI)—is proposed to expose vulnerabilities, and extensive evaluations show that UnifiedAttack consistently bypasses current alignment defenses in state‑of‑the‑art architectures.
arXiv:2602. 24210v3 Announce Type: replace-cross Abstract: Large reasoning models (LRMs) produce reasoning traces (RTs) that often contain sensitive information.
ReFrame is a training‑free framework that enhances safety alignment for multimodal large language models at test time. It uses two lightweight agents: one generates risk and utility evidence, and the other rewrites prompts and routes images to create a safe proxy before invoking the deployed MLLM. Experiments show that ReFrame improves jailbreak defense, safety awareness, and reduces over‑sensitivity while maintaining multimodal utility.
arXiv:2606. 04067v1 Announce Type: cross Abstract: As LLMs become increasingly woven into everyday workflows, user queries sent to cloud hosted LLMs routinely mix task-essential content with task non-essential sensitive disclosures, yet type based PII redaction is context agnostic and may raise two issues: over disclosing untyped sensitive context and over removing answer bearing spans.
arXiv:2607.28225v2 Announce Type: replace Abstract: Agentic vision-language models (VLMs), which interleave textual reasoning with explicit tool calls such as cropping and code-based image manipulati...
The paper introduces OverThink, a slowdown attack that forces reasoning language models (RLMs) to produce many more reasoning tokens while still giving correct answers. By injecting decoy reasoning problems—such as Markov decision processes, language translation, or graphic comprehension—into the model’s context, attackers can dramatically increase token generation (up to 46× on SQuAD and 17× on coding agents). The study evaluates the attack on both proprietary and open-source RLMs across multiple datasets, explores multimodal and coding‑agent variants, and tests several defenses, concluding that defending against OverThink is challenging and that newer RLMs are even more vulnerable due to higher per‑token costs and increased reasoning token usage.