Abstract:Jailbreak attacks reveal critical vulnerabilities in Large Language Models (LLMs) by causing them to generate harmful or unethical content. Evaluating these threats is particularly challenging due to the evolving nature of LLMs and the sophistication required in effectively probing their vulnerabilities. Current benchmarks and evaluation methods struggle to fully address these challenges, leaving gaps in the assessment of LLM vulnerabilities. In this paper, we review existing jailbreak evaluation practices and identify three assumed desiderata for an effective jailbreak evaluation protocol. To address these challenges, we introduce GuardVal, a new evaluation protocol that dynamically generates and refines jailbreak prompts based on the defender LLM's state, providing a more accurate assessment of defender LLMs' capacity to handle safety-critical situations. Moreover, we propose a new optimization method that prevents stagnation during prompt refinement, ensuring the generation of increasingly effective jailbreak prompts that expose deeper weaknesses in the defender LLMs. We apply this protocol to a diverse set of models, from Mistral-7b to GPT-4, across 10 safety domains. Our findings highlight distinct behavioral patterns among the models, offering a comprehensive view of their robustness. Furthermore, our evaluation process deepens the understanding of LLM behavior, leading to insights that can inform future research and drive the development of more secure models.
Abstract:Large Language Models (LLMs) have demonstrated remarkable capabilities across various domains. However, their potential to generate harmful responses has raised significant societal and regulatory concerns, especially when manipulated by adversarial techniques known as "jailbreak" attacks. Existing jailbreak methods typically involve appending carefully crafted prefixes or suffixes to malicious prompts in order to bypass the built-in safety mechanisms of these models. In this work, we identify a new vulnerability in which excessive linguistic complexity can disrupt built-in safety mechanisms-without the need for any added prefixes or suffixes-allowing attackers to elicit harmful outputs directly. We refer to this phenomenon as Information Overload. To automatically exploit this vulnerability, we propose InfoFlood, a jailbreak attack that transforms malicious queries into complex, information-overloaded queries capable of bypassing built-in safety mechanisms. Specifically, InfoFlood: (1) uses linguistic transformations to rephrase malicious queries, (2) identifies the root cause of failure when an attempt is unsuccessful, and (3) refines the prompt's linguistic structure to address the failure while preserving its malicious intent. We empirically validate the effectiveness of InfoFlood on four widely used LLMs-GPT-4o, GPT-3.5-turbo, Gemini 2.0, and LLaMA 3.1-by measuring their jailbreak success rates. InfoFlood consistently outperforms baseline attacks, achieving up to 3 times higher success rates across multiple jailbreak benchmarks. Furthermore, we demonstrate that commonly adopted post-processing defenses, including OpenAI's Moderation API, Perspective API, and SmoothLLM, fail to mitigate these attacks. This highlights a critical weakness in traditional AI safety guardrails when confronted with information overload-based jailbreaks.
Abstract:Large language models (LLMs) are increasingly used for tasks that require complex reasoning. Most benchmarks focus on final outcomes but overlook the intermediate reasoning steps - such as planning, revision, and decision making under resource constraints. We argue that measuring these internal processes is essential for understanding model behavior and improving reliability. We propose using strategic games as a natural evaluation environment: closed, rule-based systems with clear states, limited resources, and automatic feedback. We introduce a framework that evaluates LLMs along three core dimensions: planning, revision, and resource-constrained decision making. To operationalize this, we define metrics beyond win rate, including overcorrection risk rate, correction success rate, improvement slope, and over-budget ratio. In 4320 adversarial rounds across 12 leading models, ChatGPT-o3-mini achieves the top composite score, with a win rate of 74.7 percent, a correction success rate of 78.6 percent, and an improvement slope of 0.041. By contrast, Qwen-Plus, despite an overcorrection risk rate of 81.6 percent, wins only 25.6 percent of its matches - primarily due to excessive resource use. We also observe a negative correlation between overcorrection risk rate and correction success rate (Pearson r = -0.51, p = 0.093), suggesting that more frequent edits do not always improve outcomes. Our findings highlight the value of assessing not only what LLMs decide but how they arrive at those decisions
Abstract:Large reasoning models (LRMs) such as Claude 3.7 Sonnet and OpenAI o1 achieve strong performance on mathematical benchmarks using lengthy chain-of-thought (CoT) reasoning, but the resulting traces are often unnecessarily verbose. This inflates token usage and cost, limiting deployment in latency-sensitive or API-constrained settings. We introduce PREMISE (PRompt-based Efficient Mathematical Inference with Strategic Evaluation), a prompt-only framework that reduces reasoning overhead without modifying model weights. PREMISE combines trace-level diagnostics with gradient-inspired prompt optimization to minimize redundant computation while preserving answer accuracy. The approach jointly optimizes brevity and correctness through a multi-objective textual search that balances token length and answer validity. Unlike prior work, PREMISE runs in a single-pass black-box interface, so it can be applied directly to commercial LLMs. On GSM8K, SVAMP, and Math500 we match or exceed baseline accuracy ($96\%\rightarrow96\%$ with Claude, $91\%\rightarrow92\%$ with Gemini) while reducing reasoning tokens by up to $87.5\%$ and cutting dollar cost by $69$--$82\%$. These results show that prompt-level optimization is a practical and scalable path to efficient LRM inference without compromising reasoning quality.
Abstract:Large Language Models (LLMs) have shown remarkable progress across domains, yet their ability to perform inductive reasoning - inferring latent rules from sparse examples - remains limited. It is often assumed that chain-of-thought (CoT) prompting, as used in Large Reasoning Models (LRMs), enhances such reasoning. We investigate this assumption with creating four controlled, diagnostic game-based tasks - chess, Texas Hold'em, dice games, and blackjack - with hidden human-defined rules. We find that CoT reasoning can degrade inductive performance, with LRMs often underperforming their non-reasoning counterparts. To explain this, we present a theoretical framework that reveals how reasoning steps can amplify error through three failure modes: incorrect sub-task decomposition, incorrect sub-task solving, and incorrect final answer summarization. Based on our theoretical and empirical analysis, we introduce structured interventions that adapt CoT generation according to our identified failure types. These interventions improve inductive accuracy without retraining. Our findings suggest that effective (CoT) reasoning depends not only on taking more steps but also on ensuring those steps are well-structured.
Abstract:Large foundation models (LFMs) are susceptible to two distinct vulnerabilities: hallucinations and jailbreak attacks. While typically studied in isolation, we observe that defenses targeting one often affect the other, hinting at a deeper connection. We propose a unified theoretical framework that models jailbreaks as token-level optimization and hallucinations as attention-level optimization. Within this framework, we establish two key propositions: (1) \textit{Similar Loss Convergence} - the loss functions for both vulnerabilities converge similarly when optimizing for target-specific outputs; and (2) \textit{Gradient Consistency in Attention Redistribution} - both exhibit consistent gradient behavior driven by shared attention dynamics. We validate these propositions empirically on LLaVA-1.5 and MiniGPT-4, showing consistent optimization trends and aligned gradients. Leveraging this connection, we demonstrate that mitigation techniques for hallucinations can reduce jailbreak success rates, and vice versa. Our findings reveal a shared failure mode in LFMs and suggest that robustness strategies should jointly address both vulnerabilities.
Abstract:Prompt quality plays a critical role in the performance of large language models (LLMs), motivating a growing body of work on prompt optimization. Most existing methods optimize prompts over a fixed dataset, assuming static input distributions and offering limited support for iterative improvement. We introduce SIPDO (Self-Improving Prompts through Data-Augmented Optimization), a closed-loop framework for prompt learning that integrates synthetic data generation into the optimization process. SIPDO couples a synthetic data generator with a prompt optimizer, where the generator produces new examples that reveal current prompt weaknesses and the optimizer incrementally refines the prompt in response. This feedback-driven loop enables systematic improvement of prompt performance without assuming access to external supervision or new tasks. Experiments across question answering and reasoning benchmarks show that SIPDO outperforms standard prompt tuning methods, highlighting the value of integrating data synthesis into prompt learning workflows.
Abstract:Intent detection, a core component of natural language understanding, has considerably evolved as a crucial mechanism in safeguarding large language models (LLMs). While prior work has applied intent detection to enhance LLMs' moderation guardrails, showing a significant success against content-level jailbreaks, the robustness of these intent-aware guardrails under malicious manipulations remains under-explored. In this work, we investigate the vulnerability of intent-aware guardrails and demonstrate that LLMs exhibit implicit intent detection capabilities. We propose a two-stage intent-based prompt-refinement framework, IntentPrompt, that first transforms harmful inquiries into structured outlines and further reframes them into declarative-style narratives by iteratively optimizing prompts via feedback loops to enhance jailbreak success for red-teaming purposes. Extensive experiments across four public benchmarks and various black-box LLMs indicate that our framework consistently outperforms several cutting-edge jailbreak methods and evades even advanced Intent Analysis (IA) and Chain-of-Thought (CoT)-based defenses. Specifically, our "FSTR+SPIN" variant achieves attack success rates ranging from 88.25% to 96.54% against CoT-based defenses on the o1 model, and from 86.75% to 97.12% on the GPT-4o model under IA-based defenses. These findings highlight a critical weakness in LLMs' safety mechanisms and suggest that intent manipulation poses a growing challenge to content moderation guardrails.
Abstract:Channel knowledge map (CKM) has emerged as a crucial technology for next-generation communication, enabling the construction of high-fidelity mappings between spatial environments and channel parameters via electromagnetic information analysis. Traditional CKM construction methods like ray tracing are computationally intensive. Recent studies utilizing neural networks (NNs) have achieved efficient CKM generation with reduced computational complexity and real-time processing capabilities. Nevertheless, existing research predominantly focuses on single-antenna systems, failing to address the beamforming requirements inherent to MIMO configurations. Given that appropriate precoding vector selection in MIMO systems can substantially enhance user communication rates, this paper presents a TransUNet-based framework for constructing CKM, which effectively incorporates discrete Fourier transform (DFT) precoding vectors. The proposed architecture combines a UNet backbone for multiscale feature extraction with a Transformer module to capture global dependencies among encoded linear vectors. Experimental results demonstrate that the proposed method outperforms state-of-the-art (SOTA) deep learning (DL) approaches, yielding a 17\% improvement in RMSE compared to RadioWNet. The code is publicly accessible at https://github.com/github-whh/TransUNet.
Abstract:Multimodal Large Language Models (MLLMs) have shown remarkable versatility but face challenges in demonstrating true visual understanding, particularly in chart reasoning tasks. Existing benchmarks like ChartQA reveal significant reliance on text-based shortcuts and probabilistic pattern-matching rather than genuine visual reasoning. To rigorously evaluate visual reasoning, we introduce a more challenging test scenario by removing textual labels and introducing chart perturbations in the ChartQA dataset. Under these conditions, models like GPT-4o and Gemini-2.0 Pro experience up to a 30% performance drop, underscoring their limitations. To address these challenges, we propose Socratic Chart, a new framework that transforms chart images into Scalable Vector Graphics (SVG) representations, enabling MLLMs to integrate textual and visual modalities for enhanced chart understanding. Socratic Chart employs a multi-agent pipeline with specialized agent-generators to extract primitive chart attributes (e.g., bar heights, line coordinates) and an agent-critic to validate results, ensuring high-fidelity symbolic representations. Our framework surpasses state-of-the-art models in accurately capturing chart primitives and improving reasoning performance, establishing a robust pathway for advancing MLLM visual understanding.