Singapore

Multimodal large language models (MLLMs) have achieved significant results in various tasks, but their practical application is still severely constrained by hallucination issues, which are particularly prominent in reinforcement learning (RL) optimization processes. This paper systematically analyzes the causes of hallucinations in MLLM under RL training, identifying three key factors: (1) The model relies heavely on chained visual reasoning to guide decision-making during RL training. Thus, error and irrelevant information in visual reasoning can easily cause hallucinations, including inaccurate initial visual descriptions that anchor subsequent inferences to incorrect information, as well as redundant and broad inferential information; (2) Insufficient exploration diversity during the policy optimization phase, causing the model to output overly confident results; (3) The destructive conflict between different samples during optimization is a key factor that leads to false associations and unstable parameter updates. To address these issues, we propose a solution framework comprising three core modules. First, to improve the accuracy of visual localization, we add planning and caption stages before thinking and answer stages. To enhance initial visual descriptions ability, we allow LLMs to respond based solely on the caption and provide corresponding caption reward based on the quality of the response. Second, to enhance exploration capabilities, we classify samples based on the mean and variance of the reward distribution and select samples with high reward variance for training, thereby increasing the model&#39;s focus on diverse samples. Finally, to mitigate conflicts between training samples, we identify neural tangent kernel (NTK) similarity as the key factor. Rather than minimizing it uniformly, we regulate NTK similarity by grouping sample pairs based on a similarity threshold. An InfoNCE loss is then applied to pull dissimilar pairs closer and push overly similar ones apart, guiding interactions toward a balanced range. We conducted extensive empirical studies on image, video, and standard hallucination evaluation benchmarks. The experimental results demonstrate that the proposed method significantly reduces the hallucination rate and effectively improves the inference accuracy of MLLMs.

AAAI 2026

Ground What You See: Hallucination-Resistant MLLMs via Caption Feedback, Diversity-Aware Sampling, and Conflict Regularization

samples conflict

hallucination

reinforcement learning

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.

Multimodal large language models (MLLMs) have achieved significant results in various tasks, but their practical application is still severely constrained by hallucination issues, which are particularly prominent in reinforcement learning (RL) optimization processes. This paper systematically analyzes the causes of hallucinations in MLLM under RL training, identifying three key factors: (1) The model relies heavely on chained visual reasoning to guide decision-making during RL training. Thus, error and irrelevant information in visual reasoning can easily cause hallucinations, including inaccurate initial visual descriptions that anchor subsequent inferences to incorrect information, as well as redundant and broad inferential information; (2) Insufficient exploration diversity during the policy optimization phase, causing the model to output overly confident results; (3) The destructive conflict between different samples during optimization is a key factor that leads to false associations and unstable parameter updates. To address these issues, we propose a solution framework comprising three core modules. First, to improve the accuracy of visual localization, we add planning and caption stages before thinking and answer stages. To enhance initial visual descriptions ability, we allow LLMs to respond based solely on the caption and provide corresponding caption reward based on the quality of the response. Second, to enhance exploration capabilities, we classify samples based on the mean and variance of the reward distribution and select samples with high reward variance for training, thereby increasing the model's focus on diverse samples. Finally, to mitigate conflicts between training samples, we identify neural tangent kernel (NTK) similarity as the key factor. Rather than minimizing it uniformly, we regulate NTK similarity by grouping sample pairs based on a similarity threshold. An InfoNCE loss is then applied to pull dissimilar pairs closer and push overly similar ones apart, guiding interactions toward a balanced range. We conducted extensive empirical studies on image, video, and standard hallucination evaluation benchmarks. The experimental results demonstrate that the proposed method significantly reduces the hallucination rate and effectively improves the inference accuracy of MLLMs.

poster

Domain generalization remains a critical challenge for deploying neural networks, particularly in out-of-distribution object detection. The distributional discrepancy between training (e.g., daytime-sunny) and the realistic condition (e.g., night-rainy) inevitably produces imprecise localization and wrong classification. To address these issues, we propose a unified interaction consistency learning (UICL) framework, a novel single-source domain-generalized method designed to learn intra-class domain-invariant representations. Specifically, we put forth a cross-domain interaction mechanism to exchange region proposals between original and augmented pipelines, enriching the diversity of instance-level representations. Building upon this, we propose prediction-guided consistency learning to unify the interaction mechanism and harmonize the cross-domain representations, contributing to a discriminative prediction distribution under domain shift. In addition, we devise a cyclic interaction resilient detection strategy, which mitigates inaccurate predictions suffering from partial occlusion and ambiguous boundaries among different domains. Extensive experiments evidence that UICL significantly improves the robustness of detectors over several target domains, achieving state-of-the-art generalization performance on the diverse weather benchmark. The code is available at https://github.com/zhangpeng2001/uicl.

Unified Interaction Consistency Learning for Single-Source Domain-Generalized Object Detection in Urban Scene

In this paper, we propose a novel framework for controllable video diffusion, OmniVDiff, aiming to synthesize and comprehend multiple video visual content in a single diffusion model. To achieve this, OmniVDiff treats all video visual modalities in the color space to learn a joint distribution, while employing an adaptive control strategy that dynamically adjusts the role of each visual modality during the diffusion process, either as a generation modality or a conditioning modality. Our framework supports three key capabilities: (1) {Text-conditioned video generation}, where all modalities are jointly synthesized from a textual prompt; (2) {Video understanding}, where structural modalities are predicted from rgb inputs in a coherent manner; and (3) {X-conditioned video generation}, where video synthesis is guided by fine-grained inputs such as depth, canny and segmentation. Extensive experiments demonstrate that OmniVDiff achieves state-of-the-art performance in video generation tasks and competitive results in video understanding. Its flexibility and scalability make it well-suited for downstream applications such as video-to-video translation, modality adaptation for visual tasks, and scene reconstruction.

OmniVDiff: Omni Controllable Video Diffusion for Generation and Understanding

Multi-view understanding, the ability to reconcile visual information across diverse viewpoints for effective navigation, manipulation, and 3D scene comprehension, is a fundamental challenge in Multi-Modal Large Language Models (MLLMs) to be used as embodied agents. While recent MLLMs have shown impressive advances in high-level reasoning and planning, they frequently fall short when confronted with multi-view geometric consistency and cross-view correspondence. To comprehensively evaluate the challenges of MLLMs in multi-view scene reasoning, we introduce All-Angles Bench, a human carefully benchmark with over 2,100 question-answer pairs from 90 diverse, real-world scenes. Our broad evaluation across 38 general-purpose and 3D spatial reasoning MLLMs reveals a substantial performance gap compared to humans. More critically, our analysis identifies two root failure modes: (1) cross-view object mismatch—the inability to establish consistent object correspondence across views; and (2) cross-view spatial misalignment—the failure to infer accurate camera poses and spatial layouts. These findings underscore a lack of multi-view awareness in current MLLMs, calling for architectural innovations beyond prompt tuning alone. We believe that our benchmark offers valuable insights toward building spatially-intelligent MLLMs.

Seeing from Another Perspective: Evaluating Multi-View Understanding in MLLMs

The rapid advancement of vision-language models (VLMs) in 3D domains has accelerated research in text-query-guided point cloud processing, though existing methods underperform in point-level segmentation due to inadequate 3D-text alignment that limits local feature-text context linking. To address this limitation, we propose $\textbf{MR-COSMO}$, a Visual-Text $\textbf{M}$emory $\textbf{R}$ecall and Direct $\textbf{C}$r$\textbf{OS}$s-$\textbf{MO}$dal Alignment Method for Query-Driven 3D Segmentation, establishing explicit alignment between 3D point clouds and text/2D image data through a dedicated direct cross-modal alignment module while implementing a visual-text memory module with specialized feature banks. This direct alignment mechanism enables precise fusion of geometric and semantic features, while the memory module employs specialized banks storing text features, visual features, and their correspondence mappings to dynamically enhance scene-specific representations via attention-based knowledge recall. Comprehensive experiments across 3D instruction, reference, and semantic segmentation benchmarks confirm state-of-the-art performance.

MR-COSMO: Visual-Text Memory Recall and Direct CrOSs-MOdal Alignment Method for Query-Driven 3D Segmentation

Recent advances in Referring Expression Comprehension (REC) have been largely driven by supervised learning on curated datasets, where each expression is assumed to refer to exactly one known object. However, such assumptions rarely hold in real-world scenarios, where expressions can refer to multiple objects, fail to refer to any, or involve novel categories and complex semantics. These challenges define the task of open-world REC, which demands robust generalization and structured reasoning beyond the scope of traditional REC methods. 
In this work, we introduce a novel, training-free framework that decouples visual perception from linguistic reasoning to address open-world REC in a zero-shot setting. Our method first transforms the visual scene into a rich textual representation using an open-vocabulary multimodal perception module. 
It then employs a reasoning language model to interpret the referring expression and perform explicit logical inference over the perceived scene, enabling transparent decision-making and strong generalization in open-world scenarios. 
Experiments on three standard REC benchmarks as well as two more challenging ones, gRefCOCO and D$^3$, demonstrate that our framework achieves highly competitive zero-shot performance, often surpassing supervised baselines.

From Pixels to Logic: A Perception-Reasoning Decomposition Framework for Open-World Referring Expression Comprehension

Recent advances in vision-language-action (VLA) models
have demonstrated impressive generalization for robotic manipulation. However, these models often operate by directly
mapping visual and linguistic inputs to subsequent actions, lacking intermediate task planning, along with failure detection and recovery ability. These limitations prevent them from effectively decomposing complex tasks, recognizing problems,
and correcting erroneous actions, ultimately resulting in complete task failure. This significantly hinders their ability to
perform long-horizon tasks and generalization ability. To this end, we introduce TCoT: Trajectory Chain-of-Thought, a
unified VLA framework that enhances this direct mapping with trajectory planning as well as failure detection and recovery. TCoT leverages hierarchy trajectories as a precise and compact representation of CoT reasoning for manipulation:
global planning provides a high-level, goal-oriented trajectory to guide the robot toward its task objective, while local
planning focuses on real-time adjustments to address dynamic
changes. Moreover, we designed the Global-Local Switching Recovery algorithm that detects and effectively recovers
from failures. Experimental results reveal that TCoT surpasses the state-of-the-art methods across both real and simulated
scenarios and exhibits superior generalization capabilities. 
Code is available on https://anonymous.4open.science/r/TCoT-AB42

TCoT: Trajectory Chain-of-Thoughts for Robotic Manipulation with Failure Recovery in Vision-Language-Action Model

Accurate 3D instance segmentation is crucial for high-quality scene understanding in the 3D vision domain. However, 3D instance segmentation based on 2D-to-3D lifting approaches struggles to produce precise instance-level segmentation, due to accumulated errors introduced during the lifting process from ambiguous semantic guidance and insufficient depth constraints. 
To tackle these challenges, we propose Splitting and Growing reliable Semantic mask for high-fidelity 3D instance segmentation (SGS-3D), a novel framework that aims to improve segmentation accuracy by integrating geometric primitives and finely splitting reliable 3D semantic masks within the scene. 
Unlike existing approaches that directly rely on raw lifted masks and sacrifice segmentation accuracy, SGS-3D serves as a training-free refinement method that jointly fuses semantic and geometric information, enabling effective cooperation between the two levels of representation. 
Specifically, for semantic guidance, we introduce a mask filtering strategy that leverages the co-occurrence of 3D geometry primitives to identify and remove ambiguous masks, thereby ensuring more reliable semantic consistency with the 3D object instances. 
For the geometric refinement, we construct fine-grained object instances by exploiting both spatial continuity and high-level features, particularly in the case of interwoven objects. 
Experimental results on ScanNet200, ScanNet++, and KITTI-360 demonstrate that SGS-3D substantially improves segmentation accuracy and robustness against inaccurate masks from pre-trained models, yielding high-fidelity object instances while maintaining strong generalization across diverse indoor and outdoor environments.

SGS-3D: High-Fidelity 3D Instance Segmentation via Reliable Semantic Mask Splitting and Growing

Cross-view geo-localization (CVGL) aims to accurately localize street-view images through retrieval of corresponding geo-tagged satellite images. While prior works have achieved nearly perfect performance on certain standard datasets, their robustness in real-world corrupted environments remains under-explored. This oversight causes severe performance degradation or failure when images are affected by corruption such as blur or weather, significantly limiting practical deployment. To address this critical gap, we introduce MRGeo, the first systematic method designed for robust CVGL under corruption. MRGeo employs a hierarchical defense strategy that enhances the intrinsic quality of features and then enforces a robust geometric prior. Its core is the Spatial-Channel Enhancement Block (SCEB), which contains: (1) a Spatial Adaptive Representation Module (SARM) that models global and local features in parallel and uses a dynamic gating mechanism to arbitrate their fusion based on feature reliability; and (2) a Channel Calibration Module (CCM) that performs compensatory adjustments by modeling multi-granularity channel dependencies to counteract information loss. To prevent spatial misalignment under severe corruption, a Region-level Geometric Alignment Module (RGAM) imposes a geometric structure on the final descriptors, ensuring coarse-grained consistency. Comprehensive experiments on both robustness benchmark and standard datasets demonstrate that MRGeo not only achieves an average R@1 improvement of 2.92\% across three comprehensive robustness benchmarks (CVUSA-C-ALL, CVACT\_val-C-ALL, and CVACT\_test-C-ALL) but also establishes superior performance in cross-area evaluation, thereby demonstrating its robustness and generalization capability.

MRGeo: Robust Cross-View Geo-Localization of Corrupted Images via Spatial and Channel Feature Enhancement

Generating high-quality physically based rendering (PBR) materials is important to achieve realistic rendering in the downstream tasks, yet it remains challenging due to the intertwined effects of materials and lighting. While existing methods have made breakthroughs by incorporating material decomposition in the 3D generation pipeline, they tend to bake highlights into albedo and ignore spatially varying properties of metallicity and roughness. In this work, we present PBR3DGen, a two-stage mesh generation method with high-quality PBR materials that integrates the novel multi-view PBR material estimation model and a 3D PBR mesh reconstruction model. Specifically, PBR3DGen leverages vision language models (VLM) to guide multi-view diffusion, precisely capturing the spatial distribution and inherent attributes of reflective-metalness material. Additionally, we incorporate view-dependent illumination-aware conditions as pixel-aware priors to enhance spatially varying material properties. Furthermore, our reconstruction model reconstructs high-quality mesh with PBR materials. Experimental results demonstrate that PBR3DGen significantly outperforms existing methods, achieving new state-of-the-art results for PBR estimation and mesh generation.

PBR3DGen: A VLM-Guided Mesh Generation with High-Quality PBR Texture

The generation of realistic LiDAR point clouds plays a crucial role in the development and evaluation of autonomous driving systems. Although recent methods for 3D LiDAR point cloud generation have shown significant improvements, they still face notable limitations, including the lack of sequential generation capabilities and the inability to produce accurately positioned foreground objects and realistic backgrounds. These shortcomings hinder their practical applicability. In this paper, we introduce DriveLiDAR4D, a novel LiDAR generation pipeline consisting of multimodal conditions and a novel sequential noise prediction model LiDAR4DNet, capable of producing temporally consistent LiDAR scenes with highly controllable foreground objects and realistic backgrounds. To the best of our knowledge, this is the first work to address the sequential generation of LiDAR scenes with full scene manipulation capability in an end-to-end manner. We evaluated DriveLiDAR4D on the nuScenes and KITTI datasets, where we achieved an FRD score of 743.13 and an FVD score of 16.96 on the nuScenes dataset, surpassing the current state-of-the-art (SOTA) method, UniScene (Li et al. 2024), with an performance boost of 37.2% in FRD and 24.1% in FVD, respectively.

Content not yet available

Next from AAAI 2026

Unified Interaction Consistency Learning for Single-Source Domain-Generalized Object Detection in Urban Scene

Similar lecture

VCGD: Visual Clue Guided Decoding with Caption Model for Mitigating Hallucination in Multimodal Large Language Models