Singapore

Contextual priming, where earlier stimuli covertly bias later judgments, offers an unexplored attack surface for large language models (LLMs). We uncover a contextual priming vulnerability in which the previous response in the dialogue can steer its subsequent behavior toward policy-violating content. While existing jailbreak attacks largely rely on single-turn or multi-turn prompt manipulations, or inject static in-context examples, these methods suffer from limited effectiveness, inefficiency, or semantic drift. We introduce Response Attack (RA), a novel framework that strategically leverages intermediate, mildly harmful responses as contextual primers within a dialogue. By reformulating harmful queries and injecting these intermediate responses before issuing a targeted trigger prompt, RA exploits a previously overlooked vulnerability in LLMs. Extensive experiments across eight state-of-the-art LLMs show that RA consistently achieves significantly higher attack success rates than nine leading jailbreak baselines. Our results demonstrate that the success of RA is directly attributable to the strategic use of intermediate responses, which induce models to generate more explicit and relevant harmful content while maintaining stealth, efficiency, and fidelity to the original query.

AAAI 2026

Response Attack: Exploiting Contextual Priming to Jailbreak Large Language Models

nlp: safety and robustness

nlp: prompt engineering / prompting

nlp: (large) language models

## 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.

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.

Retrieval-Augmented Generation (RAG) enhances the quality of question answering by integrating external knowledge with internal knowledge. A robust RAG system needs to precisely regulate the dependence of the response on the two types of knowledge. The recently proposed context-aware contrastive decoding (CCD) method attempts to achieve this goal by adjusting the knowledge reference weights by comparing the output distribution differences of LLMS when they rely on different knowledge sources. However, these methods are based on probabilistic knowledge reference adjustment strategies (such as the highest probability or entropy), only focus on the relative confidence of the output responses at each decoding step, without considering the absolute confidence of the responses, which may lead to misjudgment of the external knowledge and internal knowledge reference degree in the decoding process. To this end, we propose a novel decoding method, Evidence-guided Contrastive Decoding (ECD), which conducts evidence modeling by constructing the Dirichlet distribution and regards logits as evidence vectors, so as to regulate the reference degree of internal and external knowledge more accurately, and finally improve the quality of generated responses. Extensive evaluations across four public benchmark datasets on three mainstream LLMs have demonstrated the effectiveness and advantages of ECD.

ECD: Evidence-guided Contrastive Decoding in Retrieval-Augmented Generation with Accurate Knowledge Reference Adjustment

Trajectory similarity retrieval is the cornerstone of spatiotemporal data mining and is dominated by a trade-off: traditional metrics are computationally expensive, while learning-based methods suffer from substantial training costs and potential instability. This paper challenges this dichotomy by proposing \textbf{Geo}metric \textbf{P}rototype \textbf{T}rajectory \textbf{H}ashing (GeoPTH), a novel, lightweight, and non-learning framework for efficient trajectory retrieval. GeoPTH constructs data-dependent hash functions by using representative trajectory prototypes, i.e., small point sets preserving geometric characteristics, as anchors. The hashing process is efficient, which involves mapping a new trajectory to its closest prototype via a robust, \textit{Hausdorff}-guided metric. Extensive experiments show that GeoPTH’s retrieval accuracy is highly competitive with both traditional metrics and state-of-the-art learning methods, and it significantly outperforms binary codes generated through simple binarization of the learned embeddings. Critically, GeoPTH consistently outperforms all competitors in terms of efficiency. Our work demonstrates that a lightweight, prototype-centric approach offers a practical and powerful alternative, achieving an exceptional balance between retrieval performance and computational efficiency.

GeoPTH: A Lightweight Approach to Category-Based Trajectory Retrieval via Geometric Prototype Trajectory Hashing

As Convolutional Neural Networks (CNNs) continue to gain traction in deep learning, Winograd convolution has emerged as a key algorithm to enhance computational efficiency. Although ARM-based CPUs are increasingly prevalent in mobile devices, embedded systems and HPC servers, existing 2D Winograd convolution implementations for ARM often leave room for improvement in transformation efficiency, computational throughput, and overall versatility. Furthermore, the lack of tailored 3D Winograd convolution implementations for ARM architectures stems from the additional complexity of supporting higher-dimensional kernels. AirWino introduces a set of novel optimizations covering transformations, data layouts, micro-kernel computations, and parallelization strategies for both 2D and 3D Winograd convolution. It supports FP32 and FP16 precisions with filter sizes of 3 and 5, targeting a broad range of applications. Evaluations on four distinct ARM platforms show that AirWino consistently outperforms state-of-the-art libraries across various experimental scenarios and hardware configurations, highlighting its efficiency and portability.

AirWino: Optimized Winograd Convolution for Accelerating CNN Inference on ARMv8 Processors

3D object detection is a critical component of autonomous driving, yet its performance degrades severely in adverse weather due to the degradation of LiDAR point clouds. While existing LiDAR-4D radar fusion methods enhance robustness by incorporating weather-robust 4D radar data, they often depend on well geometric structures from LiDAR and so struggle to effectively exploit radar data in case of degraded LiDAR data. To tackle this challenge, we propose REL, a novel 4D radar-guided LiDAR geometric enhancement framework. It utilizes 4D radar features to dynamically generate virtual LiDAR points, effectively increasing the density of degraded LiDAR data. Moreover, a Position-Guided Cross Attention (PGCA) module is proposed to enhance the feature representation of virtual points, while an Adaptive Feature Fusion (AFF) module is designed to integrate virtual and real LiDAR features. Extensive experiments on the K-Radar and Vod-Fog datasets demonstrate that REL achieves state-of-the-art 3D object detection performance under diverse adverse weather conditions. Notably, REL improves the overall AP3D by 9.3% on K-Radar and boosts the cyclist class by up to 52.9% 3D mAP under the most severe foggy condition on Vod-Fog.

Towards Accurate 3D Object Detection in Adverse Weather by Leveraging 4D Radar for LiDAR Geometry Enhancement

Image composition aims to seamlessly insert foreground object into background. Despite the huge progress in generative image composition, the existing methods are still struggling with simultaneous detail preservation and foreground pose/view adjustment. To address this issue, we extend the existing generative composition model to multi-reference version, which allows using arbitrary number of foreground reference images. Furthermore, we propose to calibrate the global and local features of foreground reference images to make them compatible with the background information. The calibrated reference features can supplement the original reference features with useful global and local information of proper pose/view. Extensive experiments on MVImgNet and MureCom demonstrate that the generative model can greatly benefit from the calibrated reference features.

CareCom: Generative Image Composition with Calibrated Reference Features

Cross-time vehicle re-identification (Re-ID), especially across day and night conditions, remains a challenging problem due to drastic illumination variations that lead to significant domain shifts. While existing methods perform well under daytime scenarios, their effectiveness degrades severely in cross-domain settings, and fully supervised solutions demand costly annotations in both domains. In this paper, we introduce a new setting, Unsupervised Day-Night Vehicle Re-Identification (USL-DN-ReID), and propose a novel Cluster-Instance Alignment (CIA) framework to address it. CIA performs dual-level alignment: 1) at the cluster level, a Dictionary-Guided Graph Matching (DGM) module builds a cross-domain topological graph using soft similarities among cluster centers and solves global matching via the Hungarian algorithm; 2) at the instance level, a Multi-Factor Adaptive Alignment (MAA) module introduces a multi-factor adaptive weighting strategy that emphasizes high-confidence pairwise relations while suppressing noise. Together, these components enable robust and scalable cross-domain adaptation without requiring target-domain labels. Extensive experiments conducted on the DN-348 and DN-Wild benchmarks demonstrate the effectiveness and superiority of the proposed CIA framework, setting new state-of-the-art results on both datasets.

CIA: Cluster-Instance Alignment for Unsupervised Day-Night Vehicle Re-Identification

Reinforcement learning (RL) has shown significant promise in sequential portfolio optimization. A typical solution involves optimizing cumulative returns using historical offline data. However, it may produce less generalizable policies that merely ''memorize'' optimal buying and selling actions from the offline data while neglecting the non-stationary nature of the financial market. We frame portfolio optimization of stock data as a specific type of offline RL problem. Our method, MetaTrader, presents two key contributions. First, it introduces a novel bilevel RL algorithm that operates on both the original stock data and its transformations. The core idea is that a robust policy should generalize effectively to out-of-distribution data. Second, we propose a new temporal difference (TD) method that leverages a transformation-based conservative TD target to address value overestimation under limited offline data. Empirical results on two publicly available datasets demonstrate that MetaTrader outperforms existing methods, including both traditional stock prediction models and RL-based trading approaches.

MetaTrader: Learning to Generalize RL Trading Policies Beyond Offline Data

Inference-time scaling has emerged as a powerful technique for enhancing the reasoning performance of Large Language Models (LLMs). However, existing approaches often rely on heuristic strategies for parallel sampling, lacking a principled foundation. To address this gap, we propose a probabilistic framework that formalizes the optimality of inference-time scaling under the assumption that parallel samples are independently and identically distributed (i.i.d.), and where the Best-of-N selection strategy follows a probability distribution that can be estimated. Within this framework, we derive a theoretical lower bound on the required number of samples to achieve a target performance level, providing the first principled guidance for compute-efficient scaling. Leveraging this insight, we develop OptScale, a practical algorithm that dynamically determines the optimal number of sampled responses. OptScale employs a language model-based predictor to estimate probabilistic prior parameters, enabling the decision of the minimal number of samples needed that satisfy predefined performance thresholds and confidence levels. Extensive experiments on mathematical reasoning benchmarks (including MATH-500, GSM8K, AIME, and AMC) demonstrate that OptScale significantly reduces sampling overhead while remaining better or on par with state-of-the-art reasoning performance. Our work offers both a theoretical foundation and a practical solution for principled inference-time scaling, addressing a critical gap in the efficient deployment of LLMs for complex reasoning. The source code will be open upon acceptance at \url{https://open\_upon\_acceptance}.

OptScale: Probabilistic Optimality for Inference-time Scaling

Without manual annotations, unsupervised cross-modal hashing (UCMH) aims to achieve efficient clustering and retrieval by leveraging data interrelationships. However, the retrieval accuracy is constrained by two main aspects: 1) insufficient exploration of data relationships; 2) existing knowledge mining strategies are not well aligned with the architectural properties of multilayer perceptrons. Through summary and error analysis, the human brain is able to achieve fast learning through experience and minimal data. Inspired by this cognitive process, we propose a novel Error Notebook strategy, named ENHash, to more effectively capture similarity information between multi-modal data for fine-grained unsupervised clustering. Firstly, simulating the human process of summarizing experiences, ENHash gradually integrates the information from each batch into a global clustering representation. Secondly, drawing upon human error analysis capabilities, ENHash utilizes the summarized experiences to identify and record incorrectly predicted hash codes. Finally, by leveraging the knowledge derived from this analysis, ENHash guides the hash function to learn fine-grained patterns from the errors. To the best of our knowledge, ENHash represents the first attempt at integrating cognitively-inspired mechanisms into fine-grained UCMH optimization paradigms. We evaluate the proposed ENHash against eight state-of-the-art methods on three widely used datasets and one fine-grained cross-modal dataset. Experimental results show that ENHash achieves substantial improvements over existing approaches. To support reproducibility, the experimental code has been uploaded to the following anonymous repository: https://osf.io/tbehv/?view_only=e4470e710bdf411589391807a2914218.

ENHash: Error Notebook-Guided Fine-Grained Learning for Unsupervised Cross-Modal Hashing

Sequential knowledge editing techniques aim to continuously update knowledge in large language models at low cost, preventing models from generating outdated or incorrect information. However, existing sequential editing methods suffer from a significant decline in editing success rates after long-term editing. Through theoretical analysis and experiments, our findings reveal that as the number of edits increases, the model's output increasingly deviates from the desired target, leading to a drop in editing success rates. We refer to this issue as the **superimposed noise accumulation problem**. Our further analysis demonstrates that the problem is related to the erroneous activation of irrelevant knowledge and conflicts between activated knowledge. Based on this analysis, a method named **DeltaEdit** is proposed that reduces conflicts between knowledge through dynamic orthogonal constraint strategies. Experiments show that DeltaEdit significantly reduces superimposed noise, achieving a 16.8% improvement in editing performance over the strongest baseline.

Downloads

Next from AAAI 2026

ECD: Evidence-guided Contrastive Decoding in Retrieval-Augmented Generation with Accurate Knowledge Reference Adjustment

Similar lecture

Reasoning-Augmented Conversation for Multi-Turn Jailbreak Attacks on Large Language Models

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from AAAI 2026

ECD: Evidence-guided Contrastive Decoding in Retrieval-Augmented Generation with Accurate Knowledge Reference Adjustment

Similar lecture

Reasoning-Augmented Conversation for Multi-Turn Jailbreak Attacks on Large Language Models

Downloads