Singapore

Recently, Large Vision-Language Models (LVLMs) have been demonstrated to be vulnerable to jailbreak attacks, highlighting the urgent need for further research to comprehensively identify and mitigate these threats.
Unfortunately, existing jailbreak studies primarily focus on coarse-grained input manipulation to elicit specific responses, overlooking the exploitation of internal representations, $\textit{i.e.}$, intermediate activations, which constrains their ability to penetrate alignment safeguards and generate harmful responses.
To tackle this issue, we propose the $\textbf{Act}$ivation $\textbf{Man}$ipulation (ActMan) Attack framework, which performs fine-grained activation manipulations inspired by the perception and cognition stages of human decision-making, enhancing both the penetration capability and harmfulness of attacks.
To improve penetration capability, we introduce a Deceptive Visual Camouflage module inspired by the masking effect in human perception. This module uses a benign activation-guided attention redirection strategy to conceal abnormal activation patterns, thereby suppressing LVLM&#39;s defense detection during early-stage decoding.
To enhance harmfulness, we design a Malicious Semantic Induction module drawing from the framing effect in human cognition, which reconstructs jailbreak instructions using malicious activation guidance to change LVLM’s risk assessment during late-stage decoding, thereby amplifying the harmfulness of model responses.
Extensive experiments on six mainstream LVLMs demonstrate that our method remarkably outperforms state-of-the-art baselines, achieving an average relative ASR improvement of 12.06%.

AAAI 2026

Activation Manipulation Attack: Penetrating and Harmful Jailbreak Attack Against Large Vision-Language Models

activation manipulation

jailbreak attack

large vision-language model

## Welcome!
"I am excited to welcome you to this year’s edition of the Conference on Empirical Methods in Natural Language Processing! Importantly, it marks the 30th edition of EMNLP. With over 8,000 submissions, more than 3,000 accepted papers, and thousands of attendees, we have come a long way from that first
workshop, which had 14 accepted papers. As the field looks ahead, Suzhou is the fitting location for celebrating this milestone: rooted in a long literary tradition, yet modern and forward-looking, and home to a large share of the NLP community." 

*Message from the General Chair, Dirk Hovy*

EMNLP 2025

Celebrate 30 Years of EMNLP! 
EMNLP 2025 will be held in Suzhou, China from November 5th to November 9th, 2025.

Recently, Large Vision-Language Models (LVLMs) have been demonstrated to be vulnerable to jailbreak attacks, highlighting the urgent need for further research to comprehensively identify and mitigate these threats.
Unfortunately, existing jailbreak studies primarily focus on coarse-grained input manipulation to elicit specific responses, overlooking the exploitation of internal representations, $\textit{i.e.}$, intermediate activations, which constrains their ability to penetrate alignment safeguards and generate harmful responses.
To tackle this issue, we propose the $\textbf{Act}$ivation $\textbf{Man}$ipulation (ActMan) Attack framework, which performs fine-grained activation manipulations inspired by the perception and cognition stages of human decision-making, enhancing both the penetration capability and harmfulness of attacks.
To improve penetration capability, we introduce a Deceptive Visual Camouflage module inspired by the masking effect in human perception. This module uses a benign activation-guided attention redirection strategy to conceal abnormal activation patterns, thereby suppressing LVLM's defense detection during early-stage decoding.
To enhance harmfulness, we design a Malicious Semantic Induction module drawing from the framing effect in human cognition, which reconstructs jailbreak instructions using malicious activation guidance to change LVLM’s risk assessment during late-stage decoding, thereby amplifying the harmfulness of model responses.
Extensive experiments on six mainstream LVLMs demonstrate that our method remarkably outperforms state-of-the-art baselines, achieving an average relative ASR improvement of 12.06%.

poster

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.

To access this event page, you need to log in with the **email address you registered with**. 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.

Grassmannian manifolds offer a powerful carrier for geometric representation learning by modelling high-dimensional data as low-dimensional subspaces. However, existing approaches predominantly rely on static single-subspace representations, neglecting the dynamic interplay between multiple subspaces critical for capturing complex geometric structures. To address this limitation, we propose a topology-driven multi-subspace fusion framework that enables adaptive subspace collaboration on Grassmannian manifolds. Our solution introduces two key innovations: (1) an adaptive multi-subspace construction mechanism that dynamically selects and weights task-relevant subspaces via topological convergence analysis, and (2) a multi-subspace interaction block that fuses heterogeneous geometric representations through Fréchet mean optimisation on the manifold. Theoretically, we establish the convergence guarantees of adaptive subspaces under a projection metric topology, ensuring stable gradient-based optimisation. Practically, we integrate Riemannian batch normalisation and mutual information regularisation to enhance discriminability and robustness. Extensive experiments on 3D action recognition (HDM05, FPHA), EEG classification (MAMEM-SSVEPII), and graph tasks demonstrate state-of-the-art performance. Our work not only advances geometric deep learning but also successfully adapts the proven multi-channel interaction philosophy of Euclidean networks to non-Euclidean domains, achieving superior discriminability and interpretability.

Learning Topology-Driven Multi-Subspace Fusion for Grassmannian Deep Networks

Existing video-based automatic depression assessment (ADA) approaches frequently achieve video-level depression assessment by aggregating features or predictions of individual frames or equal-length segments within the given video. While their performances have been largely enhanced by recent advanced deep learning models, they typically fail to explicitly consider the varied importance of depression-related behavioural cues across different video segments, i.e., segments within one video may contain behaviours reflecting varying depression status. Underestimating segment-level variations can obscure the detection of facial behaviour cues associated with depression, thereby undermining the accuracy and interpretability of video-based depression detection systems. In this paper, we propose a novel video-based ADA approach that specifically identifies and differentiates video segments that exhibit depression-related facial behaviours across varying temporal durations, providing clear insights into how each segment contributes to the video-level depression prediction. To achieve this, a novel weakly supervised strategy is proposed to compare segment-level behaviours with video-level depression label, enabling the model to assign depression-relevant scores to multiple temporal scale video segments and attend selectively to those most indicative of depressive states. Extensive experiments on the AVEC 2013 and AVEC 2014 face video depression datasets demonstrate the effectiveness of our methods. Our code is provided in the Supplementary Material.

Explainable Depression Assessment from Face Videos by Weakly Supervised Learning

Large Reasoning Models (LRMs) have demonstrated promising performance in complex tasks. However, the resource-consuming reasoning processes may be exploited by attackers to maliciously occupy the resources of the servers, leading to a crash, like the DDoS attack in cyber. To this end, we propose a novel attack method on LRMs termed ExtendAttack to maliciously occupy the resources of servers by stealthily extending the reasoning processes of LRMs. Concretely, we systematically obfuscate characters within a benign prompt, transforming them into a complex, poly-base ASCII representation. This compels the model to perform a series of computationally intensive decoding sub-tasks that are deeply embedded within the semantic structure of the query itself. Extensive experiments demonstrate the effectiveness of our proposed ExtendAttack. Remarkably, it significantly increases response length and latency, with the former increasing by over 2.7 times for the o3 model on the HumanEval benchmark. Besides, it preserves the original meaning of the query and achieves comparable answer accuracy, showing the stealthiness.

ExtendAttack: Attacking Servers of LRMs via Extending Reasoning

Face anti-spoofing (FAS) is crucial for protecting facial recognition systems from presentation attacks. Previous methods approached this task as a classification problem, lacking interpretability and reasoning behind the predicted results. Recently, multimodal large language models (MLLMs) have shown strong capabilities in perception, reasoning, and decision-making in visual tasks. However, there are currently no MLLMs for FAS and existing FAS datasets only contain simple classification labels without corresponding explanations. To address this gap, we propose FaceShield, a MLLM for FAS, along with the corresponding pre-training and supervised fine-tuning (SFT) datasets, FaceShield-pre10K and FaceShield-sft45K. FaceShield is capable of determining the authenticity of faces, identifying types of spoofing attacks, providing reasoning for its judgments, and detecting attack areas. Specifically, we employ spoof-aware vision perception (SAVP) that incorporates both the original image and auxiliary information based on prior knowledge. We then use an prompt-guided vision token masking (PVTM) strategy to random mask vision tokens, thereby improving the model's generalization ability. We conducted extensive experiments on three benchmark datasets, demonstrating that FaceShield significantly outperforms previous deep learning models and general MLLMs on four FAS tasks, i.e., coarse-grained classification, fine-grained classification, reasoning, and attack localization . Our instruction datasets, protocols, and codes will be released soon.

FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models

State-of-the-art (SOTA) gradient-based adversarial attacks on spiking neural networks (SNNs), which largely rely on extending FGSM and PGD frameworks, face a critical limitation: substantial attack latency from multi-timestep processing, rendering them infeasible for practical real-time applications. This inefficiency stems from their design as direct extensions of ANN paradigms, which fail to exploit key SNN properties. In this paper, we propose the timestep compressed attack (TCA), a novel framework that significantly reduces attack latency. TCA introduces two components founded on key insights into SNN behavior. First, timestep-level backpropagation (TLBP) is based on our finding that global temporal information in backpropagation to generate perturbations is not critical for an attack’s success, enabling per-timestep evaluation for early stopping. Second, adversarial membrane potential reuse (A-MPR) is motivated by the observation that initial timesteps are inefficiently spent accumulating membrane potential, a warm-up phase that can be pre-calculated and reused. Our experiments on VGG-11 and ResNet-17 with the CIFAR-10/100 and CIFAR10-DVS datasets show that TCA significantly reduces the required attack latency by up to 56.6% and 57.1% compared to SOTA methods in white-box and black-box settings, respectively, while maintaining a comparable attack success rate.

Timestep-Compressed Attack on Spiking Neural Networks Through Timestep-Level Backpropagation

Generating behaviors that align with human expectations is a key requirement for human-robot collaboration. Potential behavior misalignment could lead to the robot performing actions with unanticipated, potentially dangerous side effects even while pursuing human goals. In this paper, we introduce a novel metric called Goal State Divergence $\mathcal{(GSD)}$ which quantifies the difference between the state a robot achieved in response to a human-specified goal and what the human expected. In cases where $\mathcal{GSD}$ cannot be directly calculated, we show how it can be approximated using maximal and minimal bounds. We then leverage $\mathcal{GSD}$ in our novel human-robot goal alignment design (HRGAD) problem, which identifies a minimal set of environment modifications that can reduce such mismatches. We show the effectiveness of our method in reducing the goal state divergence by empirically evaluating our approach on several planning benchmarks.

Reducing Goal State Divergence with Environment Design

Human writers often begin their stories with an overarching mental scene, where they envision the interactions between characters and their environment. Inspired by this creative process, we propose a novel approach to long-form story generation, termed hybrid bottom-up long-form story generation, using multi-agent simulations. In our method, agents interact within a dynamic sandbox environment, where their behaviors and interactions with one another and the environment generate emergent events. These events form the foundation for the story, enabling organic character development and plot progression. Unlike traditional top-down approaches that impose rigid structures, our hybrid bottom-up approach allows for the natural unfolding of events, fostering more spontaneous and engaging storytelling. The system is capable of generating stories exceeding 10,000 words while maintaining coherence and consistency, addressing some of the key challenges faced by current story generation models. We achieve state-of-the-art performance across several metrics. This approach offers a scalable and innovative solution for creating dynamic, immersive long-form stories that evolve organically from agent-driven interactions.

StoryBox: Collaborative Multi-Agent Simulation for Hybrid Bottom-Up Long-Form Story Generation Using Large Language Models

The ability of Large Language Models (LLMs) to use ex
ternal tools unlocks powerful real-world interactions, mak
ing rigorous evaluation essential. However, current bench
marks primarily report final accuracy, revealing what mod
els can do but obscuring the cognitive bottlenecks that define
their true capability boundaries. To move from simple per
formance scoring to a diagnostic tool, we introduce a frame
workgroundedinCognitive LoadTheory.Ourframeworkde
constructs task complexity into two quantifiable components:
Intrinsic Load, the inherent structural complexity of the solu
tion path, formalized with a novel Tool Interaction Graph; and
Extraneous Load, the difficulty arising from ambiguous task
presentation. To enable controlled experiments, we construct
ToolLoad-Bench, the first benchmark with parametrically ad
justable cognitive load. Our evaluation reveals distinct per
formance cliffs as cognitive load increases, allowing us to
precisely map each model’s capability boundary. We validate
that our framework’s predictions are highly calibrated with
empirical results, establishing a principled methodology for
understanding an agent’s limits and a practical foundation for
building more efficient systems.

Beyond Accuracy: A Cognitive Load Framework for Mapping the Capability Boundaries of Tool-use Agents

Recent advances in vision-language models have opened up new possibilities for reasoning-driven image geolocalization. However, existing approaches often rely on synthetic reasoning annotations or external image retrieval, which can limit interpretability and generalizability. In this paper, we present Geo-R1, a retrieval-free framework that uncovers structured reasoning paths from existing ground-truth coordinates and optimizes geolocation accuracy via reinforcement learning. We propose the Chain of Region, a rule-based hierarchical reasoning paradigm that generates precise, interpretable supervision by mapping GPS coordinates to geographic entities (e.g., country, province, city) without relying on model-generated or synthetic labels. Building on this, we introduce a lightweight reinforcement learning strategy with coordinate-aligned rewards based on Haversine distance, enabling the model to refine predictions through spatially meaningful feedback. Our approach bridges structured geographic reasoning with direct spatial supervision, yielding improved localization accuracy, stronger generalization, and more transparent inference. Experimental results across multiple benchmarks confirm the effectiveness of Geo-R1, establishing a new retrieval-free paradigm for scalable and interpretable image geolocalization. To facilitate further research and ensure reproducibility, both the model and code will be made publicly available.

Vision-Language Reasoning for Geolocalization: A Reinforcement Learning Approach

Flawed planning from VLM-driven embodied agents poses significant safety hazards, hindering their deployment in real-world household tasks. However, existing static, non-interactive evaluation paradigms fail to adequately assess risks within these interactive environments, since they cannot simulate dynamic risks that emerge from an agent’s actions and rely on unreliable post-hoc evaluations that ignore unsafe intermediate steps. To bridge this critical gap, we propose evaluating an agent’s interactive safety: its ability to perceive emergent risks and execute mitigation steps in the correct procedural order. We thus present IS-Bench, the first multi-modal benchmark designed for interactive safety, featuring 161 challenging scenarios with 388 unique safety risks instantiated in a high-fidelity simulator. Crucially, it facilitates a novel process-oriented evaluation that verifies whether risk mitigation actions are performed before/after specific risk-prone steps. Extensive experiments on 16 leading VLMs, including the GPT-4o and Gemini-2.5 series, reveal that current agents lack interactive safety awareness. While safety-aware Chain-of-Thought can improve performance, it often compromises task completion. By highlighting these critical limitations, IS-Bench provides a foundation for developing safer and more reliable embodied AI systems.

Downloads

Next from AAAI 2026

Learning Topology-Driven Multi-Subspace Fusion for Grassmannian Deep Networks

Similar lecture

PBI-Attack: Prior-Guided Bimodal Interactive Black-Box Jailbreak Attack for Toxicity Maximization

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES