Singapore

With the rapid development of multimodal large language models (MLLMs), deploying them on low-resource devices remains challenging. Beyond the model size, long multimodal inputs cause substantial memory overhead in the KV cache, making efficient cache management critical. In this paper, we propose DAVID, a KV cache eviction strategy that adapts to the degree of modality fusion across layers. By analyzing the feature distributions of vision and text tokens, we observe low fusion in early layers and high fusion in deeper layers. Based on this observation, DAVID adopts a decoupled eviction strategy in shallow layers and a super-modal eviction strategy in deeper layers. To support this dynamic switching, we design a lightweight metric that quantifies cross-modal fusion and uses a threshold to determine which layers require decoupling. Experimental results show that DAVID achieves state-of-the-art performance on multiple benchmarks and offers a new perspective on KV cache eviction for MLLMs.

AAAI 2026

DAVID: Dual-stage Adaptive Vision-text Integrated Decoupling for Multimodal KV Cache Eviction

efficient multimodal llm; language model; sparse attention;

### Welcome!
The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. Our Virtual Poster Sessions will take place online Thursday, August 22, 2024.

You are required to register for this event. **Please register [here](https://2024.aclweb.org/registration). **

If you have already registered, please check your inbox for an email from Underline granting you access to ACL 2024 content.

Please register!

ACL 2024

The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. More information will be announced soon.

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.

We propose Anomagic, a zero-shot anomaly generation method that produces semantically coherent anomalies without requiring any exemplar anomalies. By unifying both visual and textual cues through a crossmodal prompt encoding scheme, Anomagic leverages rich contextual information to steer an inpainting‐based generation pipeline. A subsequent contrastive refinement strategy enforces precise alignment between synthesized anomalies and their masks, thereby bolstering downstream anomaly detection accuracy. To facilitate training, we introduce AnomVerse, a collection of 12,987 anomaly–mask–caption triplets assembled from 13 publicly available datasets, where captions are automatically generated by multimodal large language models using structured visual prompts and template‐based textual hints. Extensive experiments demonstrate that Anomagic trained on AnomVerse can synthesize more realistic and varied anomalies than prior methods, yielding superior improvements in downstream anomaly detection. Furthermore, Anomagic can generate anomalies for any normal‐category image using user‐defined prompts, establishing a versatile foundation model for anomaly generation.

Anomagic: Crossmodal Prompt-driven Zero-shot Anomaly Generation

Reasoning-intensive ranking models built on large language models (LLMs) have made notable progress, but existing approaches often rely on very large-scale LLMs and explicit chain-of-thought reasoning, resulting in high computational cost and latency that limit real-world use. To address this, we propose TFRank, an efficient pointwise reasoning ranker based on small-scale LLMs. TFRank combines chain-of-thought data, multi-task training, and fine-grained score supervision to improve ranking performance. Furthermore, with a think-mode switch and pointwise format constraints, TFRank leverages explicit reasoning during training while delivering precise relevance scores for complex queries without generating reasoning chains at inference. Experiments show that TFRank achieves performance comparable to models with four times more parameters on the BRIGHT benchmark, and demonstrates strong competitiveness on the BEIR benchmark. Further analysis shows that TFRank achieves an effective balance between efficiency and performance, providing a practical solution for integrating advanced reasoning into real-world systems.

TFRank: Think-Free Reasoning Enables Practical Pointwise LLM Ranking

Garment-centric fashion image generation aims to synthesize realistic and controllable human models dressing a given garment, which has attracted growing interest due to its practical applications in e-commerce. The key challenges of the task lie in two aspects: (1) faithfully preserving the garment details, and (2) gaining fine-grained controllability over the model's appearance. 
Existing methods typically require performing garment deformation in the generation process, which often leads to garment texture distortions. Also, they fail to control the fine-grained attributes of the generated models, due to the lack of specifically designed mechanisms.
To address these issues, we propose FashionMAC, a novel diffusion-based deformation-free framework that achieves high-quality and controllable fashion showcase image generation. The core idea of our framework is to eliminate the need for performing garment deformation and directly outpaint the garment segmented from a dressed person, which enables faithful preservation of the intricate garment details.
Moreover, we propose a novel region-adaptive decoupled attention (RADA) mechanism along with a chained mask injection strategy to achieve fine-grained appearance controllability over the synthesized human models.
Specifically, RADA adaptively predicts the generated regions for each fine-grained text attribute and enforces the text attribute to focus on the predicted regions by a chained mask injection strategy, significantly enhancing the visual fidelity and the controllability. Extensive experiments validate the superior performance of our framework compared to existing state-of-the-art methods.

FashionMAC: Deformation-Free Fashion Image Generation with Fine-Grained Model Appearance Customization

Label distribution learning (LDL) is a novel paradigm that describe the samples by label distribution of a sample. However, acquiring LDL dataset is costly and time-consuming, which leads to the birth of incomplete label distribution learning (IncomLDL). All the previous IncomLDL methods set the description degrees of "missing" labels in an instance to 0, but remains those of other labels unchanged. This setting is unrealistic because when certain labels are missing, the degrees of the remaining labels will increase accordingly. We fix this unrealistic setting in IncomLDL and raise a new problem: LDL with hidden labels (HidLDL), which aims to recover a complete label distribution from a real-world incomplete label distribution where certain labels in an instance are omitted during annotation. To solve this challenging problem, we discover the significance of proportional information of the observed labels and capture it by an innovative constraint to utilize it during the optimization process. We simultaneously use local feature similarity and the global low-rank structure to reveal the mysterious veil of hidden labels. Moreover, we **theoretically** give the recovery bound of our method, proving the feasibility of our method in learning from hidden labels. Extensive recovery and predictive experiments on various datasets prove the superiority of our method to state-of-the-art LDL and IncomLDL methods. The code is available in the supplementary file.

Towards Better IncomLDL: We Are Unaware of Hidden Labels in Advance

Deep Neural Networks (DNNs), as valuable intellectual property, face unauthorized use. Existing protections, such as digital watermarking, are largely passive; they provide only post-hoc ownership verification and cannot actively prevent the illicit use of a stolen model. This work proposes a proactive protection scheme, dubbed ``Authority backdoors," which embeds access constraints directly into the model. In particular, the scheme utilizes a backdoor learning framework to intrinsically lock a model's utility, such that it performs normally only in the presence of a specific trigger (e.g., a hardware fingerprint). But in its absence, the DNN's performance degrades to be useless. To further enhance the security of the proposed authority scheme, the certified robustness is integrated to prevent an adaptive attacker from removing the implanted backdoor. The resulting framework establishes a provably secure authority mechanism for DNNs, combining access control with robustness guarantees against adversarial attacks. 
Extensive experiments on diverse architectures and datasets validate the effectiveness and robustness of the proposed framework. $\textit{The source code for our framework will be made available upon publication.}$

Authority Backdoor: A Certifiable Backdoor Mechanism for Authoring DNNs

Visual generative models have achieved remarkable progress in synthesizing photorealistic images and videos, yet aligning their outputs with human preferences across critical dimensions remains a persistent challenge.
Though reinforcement learning from human feedback offers promise for preference alignment, existing reward models for visual generation face limitations, including black-box scoring without interpretability and potentially resultant unexpected biases.
We present VisionReward, a general framework for learning human visual preferences in both image and video generation.
Specifically, we employ a hierarchical visual assessment framework to capture fine-grained human preferences, and leverages linear weighting to enable interpretable preference learning.
Furthermore, we propose a multi-dimensional consistent strategy when using VisionReward as a reward model during preference optimization for visual generation.
Experiments show that VisionReward can significantly outperform existing image and video reward models on both machine metrics and human evaluation.
Notably, VisionReward surpasses VideoScore by 17.2\% in preference prediction accuracy, and text-to-video models with VisionReward achieve a 31.6\% higher pairwise win rate compared to the same models using VideoScore.
The models and dataset will be open-sourced.

VisionReward: Fine-Grained Multi-Dimensional Human Preference Learning for Image and Video Generation

Vision-language models (VLMs) have gained widespread attention for their strong zero-shot capabilities across numerous downstream tasks. However, these models assume that each test image’s class label is drawn from a predefined label set and lack a reliable mechanism to reject samples from emerging unknown classes when only unlabeled data are available. To address this gap, open-set domain adaptation methods retrain models to push potential unknowns away from known clusters. Yet, some unknown samples remain stably anchored to specific known classes in the VLM feature space due to semantic relevance, which is termed as emantic Affinity Anchoring (SAA). Forcibly repelling these samples unavoidably distorts the native geometry of VLMs and degrades performance. Meanwhile, existing score‑based unknown detectors use simplistic thresholds and suffer from threshold sensitivity, resulting in sub‑optimal performance. To address aforementioned issues, we propose VLM-OpenXpert, which comprises two training‑free, plug‑and‑play inference modules. SUFF performs SVD on high-confidence unknowns to extract a low-rank "unknown subspace". Each sample’s projection onto this subspace is weighted and softly removed from its feature, suppressing unknown components while preserving semantics. BGAT corrects score skewness via a Box–Cox transform, then fits a bimodal Gaussian mixture to adaptively estimate the optimal threshold balancing known-class recognition and unknown-class rejection. Experiments on 9 benchmarks and three backbones (CLIP, SigLIP, ALIGN) under source-free OSDA settings show that our training-free pipeline matches or outperforms retraining-heavy state-of-the-art methods, establishing a powerful lightweight inference calibration paradigm for open-set VLM deployment.

Beyond Retraining: Training-Free Unknown Class Filtering for Source-Free Open Set Domain Adaptation of Vision–Language Models

Social media has fundamentally transformed how people access information and form social connections, with content expression playing a critical role in driving information diffusion. While prior research has focused largely on network structures and tipping point identification, it provides limited tools for automatically generating content tailored for virality within a specific audience. To fill this gap, we propose the novel task of Diffusion-Oriented Content Generation (DOCG) and introduce an information enhancement algorithm for generating content optimized for diffusion. Our method includes an influence indicator that enables content-level diffusion assessment without requiring access to network topology, and an information editor that employs reinforcement learning to explore interpretable editing strategies. The editor leverages generative models to produce semantically faithful, audience-aware textual or visual content. Experiments on real-world social media datasets and user study demonstrate that our approach significantly improves diffusion effectiveness while preserving the core semantics of the original content.

Designed to Spread: A Generative Approach to Enhance Information Diffusion

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.

Response Attack: Exploiting Contextual Priming to Jailbreak Large Language Models

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.

Content not yet available

Next from AAAI 2026

Anomagic: Crossmodal Prompt-driven Zero-shot Anomaly Generation

Similar lecture

NACL: A General and Effective KV Cache Eviction Framework for LLM at Inference Time