Singapore

Diffusion Transformers (DiTs) have demonstrated exceptional performance in high-fidelity image and video generation. To reduce their substantial computational costs, feature caching techniques have been proposed to accelerate inference by reusing hidden representations from previous timesteps. However, current methods often struggle to maintain generation quality at high acceleration ratios, where prediction errors increase sharply due to the inherent instability of long-step forecasting. In this work, we adopt an ordinary differential equation (ODE) perspective on the hidden-feature sequence, modeling layer representations along the trajectory as a feature-ODE. We attribute the degradation of existing caching strategies to their inability to robustly integrate historical features under large skipping intervals. To address this, we propose \textbf{FoCa} (Forecast-then-Calibrate), which treats feature caching as a feature-ODE solving problem. Extensive experiments across image synthesis, video generation, and super-resolution tasks demonstrate the effectiveness of FoCa, particularly under aggressive acceleration. Without additional training, FoCa achieves near-lossless speedups of 5.50$\times$ on FLUX, 6.45$\times$ on HunyuanVideo, 3.17$\times$ on Inf-DiT, and maintains high quality with a 4.53$\times$ speedup on DiT. Our code will be released upon acceptance.

AAAI 2026

Forecast Then Calibrate: Feature Caching as ODE for Efficient Diffusion Transformers

feature caching

efficient ml

diffusion transformer

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.

poster

Parallel corpora, as the foundation of machine translation, remain crucial even in the era of large language models (LLMs) for pre-training and fine-tuning.
However, annotating parallel corpora is extremely costly, as it requires annotators to be proficient in multiple languages.
To reduce this cost, prior work has explored image-pivoted corpus synthesis, generating multilingual captions for the same image as pseudo-parallel data.
Unfortunately, these pseudo corpora suffer from the serious issue of multilingual focus divergence, i.e., the model attending to distinct aspects of the image when generating captions in different languages.
To address this problem, we propose a method called PRISMS (Parallel Refracting ImageS into Multilingual descriptions with Structured visual guidance), which leverages semantic graphs as structured visual guidance to unify the focus of multilingual captions. 
To ensure adherence to this guidance, we introduce two key techniques: supervised fine-tuning using self-generated instructional data, and reinforcement learning with a reward signal based on semantic graph consistency. 
Experimental results on five languages show that our PRISMS significantly improves the image-pivot parallel corpora synthesis, enabling LLMs to achieve translation performance comparable to that of models trained on manually annotated corpora.

The Visual Prism: Refracting Images into Parallel Multilingual Descriptions with Structured Visual Guidance

Multi-person eyeblink detection in untrimmed in-the-wild videos is an emerging and challenging task. Due to its significant spatio-temporal fine-grained characteristics compared to general actions, we empirically find that general action detectors, though effective in broader domains, struggle with this task (i.e.,Blink-AP$ < $2\%). Specialized eyeblink methods alleviate it through fine-grained spatio-temporal operations. SOTA method proposes a unified model combining instance-aware face localization and eyeblink detection through joint multi-task learning and feature sharing. While effectiveness, it exhibits two critical limitations that may contribute to its unsatisfactory performance (i.e.,Blink-AP$=$10.11\%): (1) Face localization and eyeblink detection require distinct spatio-temporal feature granularities, making joint modeling in a unified feature space suboptimal. (2) Eyeblink task training could be largely affected by unstable face-eye feature learning under the joint training paradigm. 
To address this, we propose DeFB, a decomposed feature learning paradigm with favorable effectiveness and efficiency: (1) We design to model face and eye in feature spaces of different granularities, which greatly enhances fine-grained perception while reducing computational costs compared with unified feature space;
(2) To address the instability in face-eye feature learning, an asynchronous learning mechanism for the face and eye feature spaces is adopted, with eye feature learning serving as a refinement process based on well-trained coarse face features, which also maintains efficient feature sharing as in the existing unified model.
Compared with SOTA method, DeFB doubles the performance (Blink-AP: 24.65\% v.s. 10.11\%) while boosting efficiency by nearly 35\%. DeFB can also be integrated as a plugin to substantially augment the eyeblink detection capabilities of general action detectors. Code will be released to facilitate relevant fields.

DeFB: Decomposed Feature Learning for Real-Time Multi-Person Eyeblink Detection in Untrimmed In-the-Wild Videos

Urban air pollution is a major health crisis causing millions of premature deaths annually, underscoring the urgent need for accurate and scalable monitoring of air quality (AQ). 
While low-cost sensors (LCS) offer a scalable alternative to expensive reference-grade stations, their readings are affected by drift, calibration errors, and environmental interference. 
To address these challenges, we introduce Veli (Reference-free Variational Estimation via Latent Inference), an unsupervised Bayesian model that leverages variational inference to correct LCS readings without requiring co-location with reference stations, eliminating a major deployment barrier.
Specifically, Veli constructs a disentangled representation of the LCS readings, effectively separating the true pollutant reading from the sensor noise. To build our model and address the lack of standardized benchmarks in AQ monitoring, we also introduce the Air Quality Sensor Data Repository (AQ-SDR).
AQ-SDR is the largest AQ sensor benchmark to date, with readings from 23,737 LCS and reference stations across multiple regions. Veli demonstrates strong generalization across both in-distribution and out-of-distribution settings, effectively handling sensor drift and erratic sensor behavior. We will publicly release the model code and the dataset.

Veli: Unsupervised Method and Unified Benchmark for Low-Cost Air Quality Sensor Correction

We propose unsupervised multi-scenario (UMS) person re-identification (ReID) as a new task that expands ReID across diverse scenarios (cross-resolution, clothing change, etc.) within a single coherent framework. To tackle UMS-ReID, we introduce image-text knowledge modeling (ITKM) -- a three-stage framework that effectively exploits the representational power of vision-language models. We start with a pre-trained CLIP model with an image encoder and a text encoder. In Stage I, we introduce a scenario embedding in the image encoder and fine-tune the encoder to adaptively leverage knowledge from multiple scenarios. In Stage II, we optimize a set of learned text embeddings to associate with pseudo-labels from Stage I and introduce a multi-scenario separation loss to increase the divergence between inter-scenario text representations. In Stage III, we first introduce cluster-level and instance-level heterogeneous matching modules to obtain reliable heterogeneous positive pairs (e.g., a visible image and an infrared image of the same person) within each scenario. Next, we propose a dynamic text representation update strategy to maintain consistency between text and image supervision signals. Experimental results across multiple scenarios demonstrate the superiority and generalizability of ITKM; it not only outperforms existing scenario-specific methods but also enhances overall performance by integrating knowledge from multiple scenarios.

Image-Text Knowledge Modeling for Unsupervised Multi-Scenario Person Re-Identification

Building a unified target representation that simultaneously achieves short-term adaptability and long-term stability is crucial for robust visual tracking. 
However, existing trackers typically face an inherent trade-off. Methods primarily relying on short-term appearance and motion cues achieve rapid adaptation, but they often struggle with long-term identity consistency. Conversely, trackers that emphasize extensive temporal context provide strong robustness, yet this approach can compromise their short-term adaptability.
To bridge this gap, we propose a novel tracker, MUTrack, which comprehensively integrates both long-term and short-term memories into a unified target representation for more robust tracking. 
Specifically, we design a unified memory bank that stores and manages long-term memory for maintaining long-term identity consistency, and short-term memory for adapting to instantaneous appearance changes. 
To fully leverage the complementary nature of both long-term and short-term temporal information, we introduce a perception interaction module that dynamically fuses these memory types through deep and bidirectional interactions, enabling mutual refinement where one guides the other.
This ultimately generates a highly adaptive target representation, which effectively balances adaptability to instantaneous changes with robustness against long-term identity drift.
Extensive experiments on GOT10k, TrackingNet, LaSOT, LaSOT_ext, NfS, and OTB100 consistently demonstrate that MUTrack achieves SOTA performance.

MUTrack: A Memory-Aware Unified Representation Framework for Visual Tracking

Despite the rapid progress of Vision-Language Models (VLMs), their capabilities are inadequately assessed by existing benchmarks, which are predominantly English-centric, feature simplistic layouts, and support limited tasks. Consequently, they fail to evaluate model performance for Visually Rich Document Understanding (VRDU), a critical challenge involving complex layouts and dense text. To address this, we introduce DocWeaver, a novel multi-agent pipeline that leverages Large Language Models to automatically generate a new benchmark. The result is MosaicDoc, a large-scale, bilingual (Chinese and English) resource designed to push the boundaries of VRDU. Sourced from newspapers and magazines, MosaicDoc features diverse and complex layouts (including multi-column and non-Manhattan), rich stylistic variety from 196 publishers, and comprehensive multi-task annotations (OCR, VQA, reading order, and localization). With 72K images and over 600K QA pairs, MosaicDoc serves as a definitive benchmark for the field. Our extensive evaluation of state-of-the-art models on this benchmark reveals their current limitations in handling real-world document complexity and charts a clear path for future research.

MosaicDoc: A Large-Scale Bilingual Benchmark for Visually Rich Document Understanding

Retrieval-augmented generation (RAG) is widely adopted for knowledge-intensive tasks, but unverified external knowledge can pose risks such as data injection and retrieval pollution, leading to unexpected generation. Existing defenses rely on patch-based fixes, which limit generalization and increase system latency. To address these issues, we propose **RAG2RAG**, the first **framework-level** security solution designed specifically for RAG. Inspired by human intuition to reason about "what can and cannot be said" during RAG phase, RAG2RAG augments the main RAG module with a lightweight security expert module composed of two components: (1) a Detective that dynamically retrieves supporting evidence, and (2) a Judge that makes final decisions based on retrieved context. The main and expert modules operate in parallel without causing noticeable delays. Experiments across two languages, 6 domains, and 7 types of poisoning attacks demonstrate that RAG2RAG consistently achieves higher accuracy and lower attack success rates than 7 mainstream baselines. Furthermore, it integrates seamlessly with various RAG architectures, offering generalizable and efficient protection across diverse threat scenarios.

Safe RAG by RAG: Untying the Bell That RAG Rang with the RAG Hand

Out-of-distribution (OOD) detection plays a critical role in ensuring the robustness of machine learning models in open-world settings. While extensive efforts have been made in vision, language, and graph domains, the challenge of OOD detection in hypergraph-structured data remains unexplored. In this work, we formalize the problem of hypergraph out-of-distribution (HOOD) detection, which aims to identify nodes or hyperedges whose high-order relational contexts differ significantly from those seen during training. We propose HyperGOOD, a unified energy-based detection framework that integrates multi-scale spectral decomposition with structure-aware uncertainty propagation. By preserving both low- and high-frequency signals and diffusing uncertainty across the hypergraph, HyperGOOD effectively captures subtle and relationally entangled anomalies. Experimental results on nine hypergraph datasets demonstrate the effectiveness of our approach, establishing a new foundation for robust hypergraph learning under distributional shifts.

HyperGOOD: Towards Out-of-Distribution Detection in Hypergraphs

The recently emerging conditional diffusion models seem promising for mitigating the labor and expenses in building large 3D medical imaging datasets. However, previous studies on 3D CT generation primarily focus on specific organs characterized by a local structure and fixed contrast and have yet to fully capitalize on the benefits of both semantic and textual conditions. In this paper, we present GuideGen, a controllable framework based on easily-acquired text prompts to generate anatomical masks and corresponding CT volumes for the entire torso—from chest to pelvis. Our approach includes three core components: a text-conditional semantic synthesizer for creating realistic full-torso anatomies; an anatomy-aware high-dynamic-range (HDR) autoencoder for high-fidelity feature extraction across varying intensity levels; and a latent feature generator that ensures alignment between CT images, anatomical semantics and input prompts. Combined, these components enable data synthesis for segmentation tasks from only textual instructions. To train and evaluate GuideGen, we compile a multi-modality cancer imaging dataset with paired CT and clinical descriptions from 12 public TCIA datasets and one private real-world dataset. Comprehensive evaluations across generation quality, cross-modality alignment, and data usability on multi-organ and tumor segmentation tasks demonstrate GuideGen's superiority over existing CT generation methods.

GuideGen: A Text-Guided Framework for Paired Full-torso Anatomy and CT Volume Generation

Despite recent advances in text-to-image (T2I) generation, models still struggle to accurately render prompt-specified text with correct spatial layout—especially in multi-span, structured settings. 
This challenge is driven not only by the lack of datasets that align prompts with the exact text and layout expected in the image, but also by the absence of effective metrics for evaluating layout quality.
To address these issues, we introduce TextGround4M, a large-scale dataset of over 4 million prompt-image pairs, each annotated with span-level text grounded in the prompt and corresponding bounding boxes.
This enables fine-grained supervision for layout-aware, prompt-grounded text rendering.
Building on this, we propose a lightweight training strategy for autoregressive T2I models that appends layout-aware span tokens during training, without altering model architecture or inference behavior.
We further construct a benchmark with stratified layout complexity to evaluate both open-source and proprietary models in a zero-shot setting. 
In addition, we introduce two layout-aware metrics to address the long-standing lack of spatial evaluation in text rendering.
Our results show that models trained on TextGround4M outperform strong baselines in text fidelity, spatial accuracy, and prompt consistency, highlighting the importance of fine-grained layout supervision for grounded T2I generation.

Content not yet available

Next from AAAI 2026

The Visual Prism: Refracting Images into Parallel Multilingual Descriptions with Structured Visual Guidance

Similar lecture

SpecDiff: Accelerating Diffusion Model Inference with Self-Speculation