Singapore

In hard-label black-box adversarial attacks, where only the top-1 predicted label is accessible, the prohibitive query complexity poses a major obstacle to practical deployment. In this paper, we focus on optimizing a representative class of attacks that search for the optimal ray direction yielding the minimal $\ell_p$-norm perturbation required to push a benign image into the adversarial region. Inspired by Nesterov’s Accelerated Gradient (NAG), we propose a momentum-based algorithm, `ARS-OPT`, which proactively estimates the gradient at a future position inferred from accumulated momentum. We provide a theoretical analysis of its convergence behavior, showing that `ARS-OPT` enables more accurate directional updates and achieves faster, more stable optimization. Furthermore, we derive the step size analytically, eliminating the need for expensive line-search procedures. To further accelerate convergence, we incorporate surrogate-model-based priors into `ARS-OPT`&#39;s gradient estimation, resulting in `PARS-OPT` with enhanced performance. The superiority of our approach is supported by rigorous theoretical analysis under mild assumptions. Extensive experiments on ImageNet and CIFAR-10 demonstrate that our method surpasses thirteen state-of-the-art approaches in query efficiency. The source code will be released online.

AAAI 2026

Improving the Convergence Rate of Ray Search Optimization for Query-Efficient Hard-Label Attacks

peai: privacy & security

ml: optimization

cv: adversarial attacks & robustness

EMNLP 2021 is planned to be a hybrid event in Punta Cana, Dominican Republic, with both on-site and fully virtual participation possible. The experience for on-site participants would closely approximate a normal pre-COVID *ACL conference, with 5-6 thematically organized parallel sessions and live Q/A and interactive discussion immediately after the talks. Presentations by virtual participants will be equitably interleaved with those of on-site participants, projected on the auditorium screens as if on-site, and also followed immediately by live Q/A and interactive discussion at a time during reasonable waking hours for the virtual presenter. For all participants, on-site and virtual, who are unable to attend a session due to either time-zone issues or because they are participating in another session live, talk recordings and slides will be available online at a minimum after the live presentation (and in many cases before as well), and questions may be submitted in advance on session-specific discussion boards and answered live in session with the usual visual aids if desired.

<iframe style="width:700px;height:400px" src="https://online.fliphtml5.com/ebtyf/ceby/"  seamless="seamless" scrolling="no" frameborder="0" allowtransparency="true" allowfullscreen="true" ></iframe>

Please Note: The EMNLP registration system is not currently connected to the underline site as we are still in the process of building out EMNLP 2021. You will receive access instructions from underline the week of November 1st. 

Access is given only to EMNLP upon registration, if you have not registered please do so [here](https://2021.emnlp.org/registration).

Registered attendees will receive access the week of November 1st.

EMNLP 2021

EMNLP 2021 is planned to be a hybrid event in Punta Cana, Dominican Republic, with both on-site and fully virtual participation possible.

In hard-label black-box adversarial attacks, where only the top-1 predicted label is accessible, the prohibitive query complexity poses a major obstacle to practical deployment. In this paper, we focus on optimizing a representative class of attacks that search for the optimal ray direction yielding the minimal $\ell_p$-norm perturbation required to push a benign image into the adversarial region. Inspired by Nesterov’s Accelerated Gradient (NAG), we propose a momentum-based algorithm, `ARS-OPT`, which proactively estimates the gradient at a future position inferred from accumulated momentum. We provide a theoretical analysis of its convergence behavior, showing that `ARS-OPT` enables more accurate directional updates and achieves faster, more stable optimization. Furthermore, we derive the step size analytically, eliminating the need for expensive line-search procedures. To further accelerate convergence, we incorporate surrogate-model-based priors into `ARS-OPT`'s gradient estimation, resulting in `PARS-OPT` with enhanced performance. The superiority of our approach is supported by rigorous theoretical analysis under mild assumptions. Extensive experiments on ImageNet and CIFAR-10 demonstrate that our method surpasses thirteen state-of-the-art approaches in query efficiency. The source code will be released online.

technical paper

We are pleased to announce the Fortieth AAAI Conference on Artificial Intelligence (AAAI-26), which will be held in Singapore EXPO from January 20 to January 27, 2026.

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-26 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.<br><br>

To access this event page, you need to log in with the **email address you registered with**. <br>Access credentials will be sent to your email from Underline -  subject line "Welcome to AAAI 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

Please log in

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://aaai.org/conference/aaai/aaai-26/registration/) first.

**Online Registration Form**: https://aaai.getregistered.net/conference-2026 

Registration Required

We are pleased to announce the Fortieth AAAI Conference on Artificial Intelligence (AAAI-26), which will be held in Singapore EXPO from January 20 to January 27, 2026.

Fine-tuning large language models (LLMs) improves performance but introduces critical safety vulnerabilities: even minimal harmful data can severely compromise safety measures. We observe that perturbations orthogonal to the alignment direction—defined by weight differences between aligned (safe) and unaligned models—rapidly compromise model safety. In contrast, updates along the alignment direction largely preserve it, revealing the parameter space as a ''narrow safety basin''. To address this, we propose **SECURE** (**S**afety **E**nforcement **C**onstraint **U**sing **Re**gularized Orthogonality) to maintain safety by explicitly constraining update directions during fine-tuning. By penalizing updates orthogonal to the alignment direction, SECURE effectively constrains the model within the ''narrow safety basin," thus preserving its inherent safety. Extensive experiments on multiple datasets and models show that SECURE reduces harmful behaviors by up to 7.60\%, improves task performance by 3.44\%, and consistently outperforms existing methods across multiple tasks. Code and datasets are available at: <https://anonymous.4open.science/r/69F7-ED36/>.

AsFT: Anchoring Safety During LLM Fine-Tuning Within Narrow Safety Basin

Cross-Domain Recommendation (CDR) transfers user preferences from a source domain to alleviate data sparsity in a target domain. While disentangling representations into domain-specific and shared components is a common method, existing methods overlook user preference heterogeneity and item appeal heterogeneity. To this end, we propose **DPGCDR**, a **D**ual-**P**erspective **G**roup-aware **CDR** method that learns symmetric group-aware representations from both user and item. Conceptually, DPGCDR dynamically clusters users into groups and items into themes, then symmetrically disentangles user preferences into group-specific and cross-group shared components, and item attributes into theme-specific and cross-theme shared components. We propose a two-stage training scheme: 1) an initial warm-up stage learns preliminary representations to dynamically cluster users and items into group and theme structures which generalize cross-domain scenarios into multi-group disentanglement analogous to multi-domain settings; 2) a fusion-based aggregation stage integrates these group/theme-specific components into unified global representations. Besides, an information-theoretic alignment regularizer further ensures consistency and discriminability between global shared and group/theme-specific representations, facilitating effective knowledge transfer by explicitly modeling and preserving the inherent multi-group structure within cross-domain interactions. Extensive experiments show DPGCDR achieves state-of-the-art performance, with significant gains of up to 25\% in HR@10 over baselines on datasets with heterogeneous interaction structures. Further analyses confirm our dynamic clustering mechanism effectively adapts to underlying data patterns, enabling fine-grained cross-domain knowledge transfer.

Dual-Perspective Disentanglement: Learning Symmetric Group-Aware Representations for Cross-Domain Recommendation

Large language models (LLMs) have demonstrated impressive performance on natural language tasks, but their decision-making processes remain largely opaque. Existing explanation methods either suffer from limited faithfulness to the model's reasoning or produce explanations that are difficult for humans to understand. To address these challenges, we propose \textbf{ProtoSurE}, a novel prototype-based surrogate framework that provides faithful and understandable explanations for LLMs. ProtoSurE trains an interpretable-by-design surrogate model that aligns with the target LLM while utilizing sentence-level prototypes as understandable concepts. Extensive experiments show that ProtoSurE consistently outperforms state-of-the-art explanation methods across diverse LLMs and datasets. Importantly, ProtoSurE demonstrates strong data efficiency, requiring relatively few training examples to achieve good performance, making it practical for real-world applications. Code is available in the Appendix.

Making Sense of LLM Decisions: A Prototype-based Framework for Explainable Classification

We investigate the problem of synthesizing strategies that guarantee the successful execution of a high-level nondeterministic agent program $\delta$ in Golog within a nondeterministic first-order basic action theory considering the environment as adversarial. Our approach constructs a symbolic program graph that captures the control flow of $\delta$ independently of the domain, enabling strategy synthesis through the cross product of the program graph with the domain model. We formally relate graph-based transitions to standard Golog semantics and provide a synthesis procedure that is sound though incomplete (in general, the problem is undecidable, given that we have a first-order representation of the state). We also extend the framework to handle the case where the environment's possible behaviors are specified by a Golog program.

Strategic Reasoning over Golog Programs in the Nondeterministic Situation Calculus

Predicting the popularity of user-generated content (UGC) is a crucial but challenging task in social media analysis. While existing retrieval-augmented models enhance predictions by supplying rich contextual information, they remain limited by a fundamental precision-recall dilemma: enlarging the retrieval set increases coverage but introduces noisy, irrelevant context that harms prediction. In this work, we propose a unified framework that learns to retrieve, filter, and predict. Central to our approach is a Mixture-of-Logits based retrieval module that replaces static similarity metrics with a dynamic, multi-faceted scoring function, enabling the retrieval to be directly optimized by the prediction objective. Then an uncertainty-aware filter is designed to perform differentiable subset selection and refine the selected representations using the information bottleneck principle. At last, to enhance predictive robustness, we introduce a confidence-weighted test-time perturbation strategy. By learning to retrieve UGCs that are beneficial for prediction and filtering out uncertainty, our framework provides more relevant and reliable context. Extensive experiments demonstrate that the proposed framework achieves state-of-the-art performance, consistently outperforming strong baselines.

Learning to Curate Context: Jointly Optimizing Retrieval and Prediction for Multimodal Social Media Popularity

While large language models (LLMs) leverage both knowledge and reasoning during inference, the capacity to distinguish between them plays a pivotal role in model analysis, interpretability, and development. Inspired by dual-system cognitive theory, we propose a cognition attribution framework to decouple the contribution of knowledge and reasoning. In particular, the cognition of LLMs is decomposed into two distinct yet complementary phases: knowledge retrieval (Phase 1) and reasoning adjustment (Phase 2). To separate these phases, LLMs are prompted to generate answers under two different cognitive modes, fast thinking and slow thinking, respectively. The performance under different cognitive modes is analyzed to quantify the contribution of knowledge and reasoning. This architecture is employed to 15 LLMs across 3 datasets. Results reveal: (1) reasoning adjustment is domain-specific, benefiting reasoning-intensive domains (e.g., mathematics, physics, and chemistry) and potentially imparing knowledge-intensive domains. (2) Parameter scaling improves both knowledge and reasoning, with knowledge improvements being more pronounced. Additionally, parameter scaling make LLMs reasoning significantly more prudent, while moderately more intelligent. (3) Knowledge primarily resides in lower network layers, while reasoning operates in higher layers. Our framework not only helps understand LLMs from a "decoupling" perspective, but also provides new insights into existing research, including scaling laws, hierarchical knowledge editing, and limitations of small-scale-LLM reasoning.

Decoupling Knowledge and Reasoning in LLMs: An Exploration Using Cognitive Dual-System Theory

Ensuring the safety of AI-enabled systems, particularly in high-stakes domains such as autonomous driving and healthcare, has become increasingly critical. Traditional formal verification tools fall short when faced with systems that embed both opaque, black-box AI components and complex stochastic dynamics. To address these challenges, we introduce LUCID (Learning-enabled Uncertainty-aware Certification of stochastIc Dynamical systems), a verification engine for certifying safety of black-box stochastic dynamical systems from a finite dataset of random state transitions. As such, LUCID is the first known tool capable of establishing quantified safety guarantees for such systems. Thanks to its modular architecture and extensive documentation, LUCID is designed for easy extensibility. 

LUCID employs a data-driven methodology rooted in control barrier certificates, which are learned directly from system transition data, to ensure formal safety guarantees. We use conditional mean embeddings to embed data into a reproducing kernel Hilbert space (RKHS), where an RKHS ambiguity set is constructed that can be inflated to robustify the result to out-of-distribution behavior. 

A key innovation within LUCID is its use of a finite Fourier kernel expansion to reformulate a semi-infinite non-convex optimization problem into a tractable linear program. The resulting spectral barrier allows us to leverage the fast Fourier transform to generate the relaxed problem efficiently, offering a scalable yet distributionally robust framework for verifying safety. LUCID thus offers a robust and efficient verification framework, able to handle the complexities of modern black-box systems while providing formal guarantees of safety. These unique capabilities are demonstrated on challenging benchmarks.

LUCID: Learning-Enabled Uncertainty-Aware Certification of Stochastic Dynamical Systems

In this work, we propose a novel framework for the logical specification of non-Markovian rewards in Markov Decision Processes (MDPs) with large state spaces. Our approach leverages Linear Temporal Logic Modulo Theories over finite traces (LTLf$^{MT}$), a more expressive extension of classical temporal logic in which predicates are first-order formulas of arbitrary first-order theories rather than simple Boolean variables. This enhanced expressiveness enables the specification of complex tasks over unstructured and heterogeneous data domains, promoting a unified and reusable framework that eliminates the need for manual predicate encoding. However, the increased expressive power of LTLf$^{MT}$ introduces additional theoretical and computational challenges compared to standard LTLf specifications. We address these challenges from a theoretical standpoint, identifying a fragment of LTLf$^{MT}$ that is tractable but sufficiently expressive for reward specification in an infinite-state-space context. From a practical perspective, we introduce a method based on reward machines and Hindsight Experience Replay (HER) to translate first-order logic specifications and address reward sparsity. We evaluate this approach to a continuous-control setting using Non-Linear Arithmetic Theory, showing that it enables natural specification of complex tasks. Experimental results show how a tailored implementation of HER is fundamental in solving tasks with complex goals.

Do It for HER: First-Order Temporal Logic Reward Specification in Reinforcement Learning

The evolution of Visual Large Language Models (VLLMs) has revolutionized the automatic understanding of Visually Rich Documents (VRDs), which contain both textual and visual elements. Although VLLMs excel in Visual Question Answering (VQA) on multi-page VRDs, their ability to detect unanswerable questions is still an open research question. Our research delves into the robustness of the VLLMs to plausible yet unanswerable questions, i.e., questions that appear valid but cannot be answered due to subtle corruptions caused by swaps between related concepts or plausible question formulations. Corruptions are generated by replacing the original natural language entities with other ones of the same type, belonging to different document elements, and in different layout positions or pages of the related document. To this end, we present VRD-UQA (VISUALLY RICH DOCUMENT UNANSWERABLE QUESTION ANSWERING), a benchmark for evaluating VLLMs' resilience to plausible yet unanswerable questions across multiple dimensions. It automatically alters the questions of existing VQA datasets consisting of multi-page VRDs, verifies their unanswerability using a VLLM-as-a-judge approach, and then thoroughly evaluates VLLMs' performance. Experiments, run on 12 models, analyze: (1) The VLLMs' accuracy in detecting unanswerable questions at both page and document levels; (2) The effect of different types of corruption (NLP entity, document element, layout); (3) The effectiveness of different knowledge injection strategies based on in-context learning (OCR, multi-page selection, or the possibility of unanswerability). Our findings reveal VLLMs' limitations and demonstrate that VRD-UQA can serve as an evaluation framework for developing resilient document VQA systems.

Benchmarking Visual LLMs Resilience to Unanswerable Questions on Visually Rich Documents

"Fedspeak", the stylized and often nuanced language used by the U.S. Federal Reserve, encodes implicit policy signals and strategic stances. The Federal Open Market Committee strategically employs Fedspeak as a communication tool to shape market expectations and influence both domestic and global economic conditions. As such, automatically parsing and interpreting Fedspeak presents a high-impact challenge, with significant implications for financial forecasting, algorithmic trading, and data-driven policy analysis. Technically, to enrich the semantic and contextual representation of Fedspeak texts, we incorporate domain-specific reasoning grounded in the monetary policy transmission mechanism. We further introduce a dynamic uncertainty decoding module to assess the confidence of model predictions, thereby enhancing both classification accuracy and model reliability. Experimental results demonstrate that our framework achieves state-of-the-art performance on the policy stance analysis task. Moreover, statistical analysis reveals a significant positive correlation between perceptual uncertainty and model error rates, validating the effectiveness of perceptual uncertainty as a diagnostic signal.

Downloads

Next from AAAI 2026

AsFT: Anchoring Safety During LLM Fine-Tuning Within Narrow Safety Basin

Similar lecture

A Strong Baseline for Query Efficient Attacks in a Black Box Setting

Stay up to date with the latest Underline news!

PRODUCT

COMPANY

RESOURCES