Boyang Li

Anthony Meng Huat Tiong, Junnan Li, Boyang Li et al.

Visual question answering (VQA) is a hallmark of vision and language reasoning and a challenging task under the zero-shot setting. We propose Plug-and-Play VQA (PNP-VQA), a modular framework for zero-shot VQA. In contrast to most existing works, which require substantial adaptation of pretrained language models (PLMs) for the vision modality, PNP-VQA requires no additional training of the PLMs. Instead, we propose to use natural language and network interpretation as an intermediate representation that glues pretrained models together. We first generate question-guided informative image captions, and pass the captions to a PLM as context for question answering. Surpassing end-to-end trained baselines, PNP-VQA achieves state-of-the-art results on zero-shot VQAv2 and GQA. With 11B parameters, it outperforms the 80B-parameter Flamingo model by 8.5% on VQAv2. With 738M PLM parameters, PNP-VQA achieves an improvement of 9.1% on GQA over FewVLM with 740M PLM parameters. Code is released at https://github.com/salesforce/LAVIS/tree/main/projects/pnp-vqa

Xinyang Liu, Xuanyu Liang, Shiqi Ding et al.

The Forward-Forward (FF) algorithm offers a computationally efficient and biologically plausible alternative to backpropagation (BP) by training neural networks through purely local, layer-wise optimization. However, FF is inherently designed for classification via contrastive positive-negative sample pairs, and extending it to regression poses fundamental challenges: continuous target space lack natural "opposites" for contrastive learning, and the standard goodness function carries no information about target magnitude or ordering. We propose FFR (Forward-Forward for Regression), to our knowledge, the first framework to extend FF to real-world regression and demonstrate competitive performance across diverse real-world datasets. FFR introduces three key innovations: (1) an ordinal competitive goodness function that replaces contrastive pairs with competitive learning between partitioned neuron groups under distance-aware ordinal supervision; (2) a stratified ladder architecture where shallow layers learn coarse ordinal discrimination and deeper layers refine into fine-grained regression, with multi-scale feature aggregation for inter-layer collaboration; and (3) hierarchical prediction with uncertainty estimation, where multi-scale predictors jointly provide robust predictions and prediction confidence as a free-lunch. Extensive experimental results show FFR recovers on average 98.6% of BP's accuracy across five real-world regression benchmarks while reducing peak training memory to only 27% of BP's at depth 8 and 8% at depth 32, with per-iteration time around 72% of BP's, and substantially outperforms all BP-free competitors.

Boyang Li, Yingqian Wang, Longguang Wang et al.

Single-frame infrared small target (SIRST) detection aims at separating small targets from clutter backgrounds on infrared images. Recently, deep learning based methods have achieved promising performance on SIRST detection, but at the cost of a large amount of training data with expensive pixel-level annotations. To reduce the annotation burden, we propose the first method to achieve SIRST detection with single-point supervision. The core idea of this work is to recover the per-pixel mask of each target from the given single point label by using clustering approaches, which looks simple but is indeed challenging since targets are always insalient and accompanied with background clutters. To handle this issue, we introduce randomness to the clustering process by adding noise to the input images, and then obtain much more reliable pseudo masks by averaging the clustered results. Thanks to this "Monte Carlo" clustering approach, our method can accurately recover pseudo masks and thus turn arbitrary fully supervised SIRST detection networks into weakly supervised ones with only single point annotation. Experiments on four datasets demonstrate that our method can be applied to existing SIRST detection networks to achieve comparable performance with their fully supervised counterparts, which reveals that single-point supervision is strong enough for SIRST detection. Our code will be available at: https://github.com/YeRen123455/SIRST-Single-Point-Supervision.

Jun Chen, Ming Hu, Boyang Li et al.

Self-supervised learning for computer vision has achieved tremendous progress and improved many downstream vision tasks such as image classification, semantic segmentation, and object detection. Among these, generative self-supervised vision learning approaches such as MAE and BEiT show promising performance. However, their global masked reconstruction mechanism is computationally demanding. To address this issue, we propose local masked reconstruction (LoMaR), a simple yet effective approach that performs masked reconstruction within a small window of 7$\times$7 patches on a simple Transformer encoder, improving the trade-off between efficiency and accuracy compared to global masked reconstruction over the entire image. Extensive experiments show that LoMaR reaches 84.1% top-1 accuracy on ImageNet-1K classification, outperforming MAE by 0.5%. After finetuning the pretrained LoMaR on 384$\times$384 images, it can reach 85.4% top-1 accuracy, surpassing MAE by 0.6%. On MS COCO, LoMaR outperforms MAE by 0.5 $\text{AP}^\text{box}$ on object detection and 0.5 $\text{AP}^\text{mask}$ on instance segmentation. LoMaR is especially more computation-efficient on pretraining high-resolution images, e.g., it is 3.1$\times$ faster than MAE with 0.2% higher classification accuracy on pretraining 448$\times$448 images. This local masked reconstruction learning mechanism can be easily integrated into any other generative self-supervised learning approach. Our code is publicly available in https://github.com/junchen14/LoMaR.

Haoxin Li, Yuan Liu, Hanwang Zhang et al.

In video action recognition, shortcut static features can interfere with the learning of motion features, resulting in poor out-of-distribution (OOD) generalization. The video background is clearly a source of static bias, but the video foreground, such as the clothing of the actor, can also provide static bias. In this paper, we empirically verify the existence of foreground static bias by creating test videos with conflicting signals from the static and moving portions of the video. To tackle this issue, we propose a simple yet effective technique, StillMix, to learn robust action representations. Specifically, StillMix identifies bias-inducing video frames using a 2D reference network and mixes them with videos for training, serving as effective bias suppression even when we cannot explicitly extract the source of bias within each video frame or enumerate types of bias. Finally, to precisely evaluate static bias, we synthesize two new benchmarks, SCUBA for static cues in the background, and SCUFO for static cues in the foreground. With extensive experiments, we demonstrate that StillMix mitigates both types of static bias and improves video representations for downstream applications. Code is available at https://github.com/lihaoxin05/StillMix.

Bosheng Ding, Chengwei Qin, Linlin Liu et al.

Data annotation is the process of labeling data that could be used to train machine learning models. Having high-quality annotation is crucial, as it allows the model to learn the relationship between the input data and the desired output. GPT-3, a large-scale language model developed by OpenAI, has demonstrated impressive zero- and few-shot performance on a wide range of NLP tasks. It is therefore natural to wonder whether it can be used to effectively annotate data for NLP tasks. In this paper, we evaluate the performance of GPT-3 as a data annotator by comparing it with traditional data annotation methods and analyzing its output on a range of tasks. Through this analysis, we aim to provide insight into the potential of GPT-3 as a general-purpose data annotator in NLP.

Fei Zhang, Tianfei Zhou, Boyang Li et al.

This paper studies the problem of weakly open-vocabulary semantic segmentation (WOVSS), which learns to segment objects of arbitrary classes using mere image-text pairs. Existing works turn to enhance the vanilla vision transformer by introducing explicit grouping recognition, i.e., employing several group tokens/centroids to cluster the image tokens and perform the group-text alignment. Nevertheless, these methods suffer from a granularity inconsistency regarding the usage of group tokens, which are aligned in the all-to-one v.s. one-to-one manners during the training and inference phases, respectively. We argue that this discrepancy arises from the lack of elaborate supervision for each group token. To bridge this granularity gap, this paper explores explicit supervision for the group tokens from the prototypical knowledge. To this end, this paper proposes the non-learnable prototypical regularization (NPR) where non-learnable prototypes are estimated from source features to serve as supervision and enable contrastive matching of the group tokens. This regularization encourages the group tokens to segment objects with less redundancy and capture more comprehensive semantic regions, leading to increased compactness and richness. Based on NPR, we propose the prototypical guidance segmentation network (PGSeg) that incorporates multi-modal regularization by leveraging prototypical sources from both images and texts at different levels, progressively enhancing the segmentation capability with diverse prototypical patterns. Experimental results show that our proposed method achieves state-of-the-art performance on several benchmark datasets. The source code is available at https://github.com/Ferenas/PGSeg.

Jiayun Luo, Siddhesh Khandelwal, Leonid Sigal et al.

From image-text pairs, large-scale vision-language models (VLMs) learn to implicitly associate image regions with words, which prove effective for tasks like visual question answering. However, leveraging the learned association for open-vocabulary semantic segmentation remains a challenge. In this paper, we propose a simple, yet extremely effective, training-free technique, Plug-and-Play Open-Vocabulary Semantic Segmentation (PnP-OVSS) for this task. PnP-OVSS leverages a VLM with direct text-to-image cross-attention and an image-text matching loss. To balance between over-segmentation and under-segmentation, we introduce Salience Dropout; by iteratively dropping patches that the model is most attentive to, we are able to better resolve the entire extent of the segmentation mask. PnP-OVSS does not require any neural network training and performs hyperparameter tuning without the need for any segmentation annotations, even for a validation set. PnP-OVSS demonstrates substantial improvements over comparable baselines (+26.2% mIoU on Pascal VOC, +20.5% mIoU on MS COCO, +3.1% mIoU on COCO Stuff and +3.0% mIoU on ADE20K). Our codebase is at https://github.com/letitiabanana/PnP-OVSS.

Yidan Sun, Qin Chao, Boyang Li

Cognitive science and symbolic AI research suggest that event causality provides vital information for story understanding. However, machine learning systems for story understanding rarely employ event causality, partially due to the lack of methods that reliably identify open-world causal event relations. Leveraging recent progress in large language models, we present the first method for event causality identification that leads to material improvements in computational story understanding. Our technique sets a new state of the art on the COPES dataset (Wang et al., 2023) for causal event relation identification. Further, in the downstream story quality evaluation task, the identified causal relations lead to 3.6-16.6% relative improvement on correlation with human ratings. In the multimodal story video-text alignment task, we attain 4.1-10.9% increase on Clip Accuracy and 4.2-13.5% increase on Sentence IoU. The findings indicate substantial untapped potential for event causality in computational story understanding. The codebase is at https://github.com/insundaycathy/Event-Causality-Extraction.

Mengxuan Xiao, Yinfei Zhu, Yiming Zhu et al.

Infrared small target detection presents significant challenges due to the limited intrinsic features of the target and the overwhelming presence of visually similar background distractors. We contend that background semantics are critical for distinguishing between objects that appear visually similar in this context. To address this challenge, we propose a task, clustered infrared small target detection, and introduce DenseSIRST, a benchmark dataset that provides per-pixel semantic annotations for background regions. This dataset facilitates the shift from sparse to dense target detection. This dataset facilitates the shift from sparse to dense target detection. Building on this resource, we propose the Background-Aware Feature Exchange Network (BAFE-Net), a multi-task architecture that jointly tackles target detection and background semantic segmentation. BAFE-Net incorporates a dynamic cross-task feature hard-exchange mechanism, enabling the effective exchange of target and background semantics between the two tasks. Comprehensive experiments demonstrate that BAFE-Net significantly enhances target detection accuracy while mitigating false alarms. The DenseSIRST dataset, along with the code and trained models, is publicly available at https://github.com/GrokCV/BAFE-Net.

Xinyi Ying, Li Liu, Zaipin Lin et al.

Multi-frame infrared small target (MIRST) detection in satellite videos is a long-standing, fundamental yet challenging task for decades, and the challenges can be summarized as: First, extremely small target size, highly complex clutters & noises, various satellite motions result in limited feature representation, high false alarms, and difficult motion analyses. Second, the lack of large-scale public available MIRST dataset in satellite videos greatly hinders the algorithm development. To address the aforementioned challenges, in this paper, we first build a large-scale dataset for MIRST detection in satellite videos (namely IRSatVideo-LEO), and then develop a recurrent feature refinement (RFR) framework as the baseline method. Specifically, IRSatVideo-LEO is a semi-simulated dataset with synthesized satellite motion, target appearance, trajectory and intensity, which can provide a standard toolbox for satellite video generation and a reliable evaluation platform to facilitate the algorithm development. For baseline method, RFR is proposed to be equipped with existing powerful CNN-based methods for long-term temporal dependency exploitation and integrated motion compensation & MIRST detection. Specifically, a pyramid deformable alignment (PDA) module and a temporal-spatial-frequency modulation (TSFM) module are proposed to achieve effective and efficient feature alignment, propagation, aggregation and refinement. Extensive experiments have been conducted to demonstrate the effectiveness and superiority of our scheme. The comparative results show that ResUNet equipped with RFR outperforms the state-of-the-art MIRST detection methods. Dataset and code are released at https://github.com/XinyiYing/RFR.

Tianhao Wu, Boyang Li, Yihang Luo et al.

Space-based infrared tiny ship detection aims at separating tiny ships from the images captured by earth orbiting satellites. Due to the extremely large image coverage area (e.g., thousands square kilometers), candidate targets in these images are much smaller, dimer, more changeable than those targets observed by aerial-based and land-based imaging devices. Existing short imaging distance-based infrared datasets and target detection methods cannot be well adopted to the space-based surveillance task. To address these problems, we develop a space-based infrared tiny ship detection dataset (namely, NUDT-SIRST-Sea) with 48 space-based infrared images and 17598 pixel-level tiny ship annotations. Each image covers about 10000 square kilometers of area with 10000X10000 pixels. Considering the extreme characteristics (e.g., small, dim, changeable) of those tiny ships in such challenging scenes, we propose a multi-level TransUNet (MTU-Net) in this paper. Specifically, we design a Vision Transformer (ViT) Convolutional Neural Network (CNN) hybrid encoder to extract multi-level features. Local feature maps are first extracted by several convolution layers and then fed into the multi-level feature extraction module (MVTM) to capture long-distance dependency. We further propose a copy-rotate-resize-paste (CRRP) data augmentation approach to accelerate the training phase, which effectively alleviates the issue of sample imbalance between targets and background. Besides, we design a FocalIoU loss to achieve both target localization and shape description. Experimental results on the NUDT-SIRST-Sea dataset show that our MTU-Net outperforms traditional and existing deep learning based SIRST methods in terms of probability of detection, false alarm rate and intersection over union.

Jiaxian Guo, Junnan Li, Dongxu Li et al.

Large language models (LLMs) have demonstrated excellent zero-shot generalization to new language tasks. However, effective utilization of LLMs for zero-shot visual question-answering (VQA) remains challenging, primarily due to the modality disconnection and task disconnection between LLM and VQA task. End-to-end training on vision and language data may bridge the disconnections, but is inflexible and computationally expensive. To address this issue, we propose \emph{Img2Prompt}, a plug-and-play module that provides the prompts that can bridge the aforementioned modality and task disconnections, so that LLMs can perform zero-shot VQA tasks without end-to-end training. In order to provide such prompts, we further employ LLM-agnostic models to provide prompts that can describe image content and self-constructed question-answer pairs, which can effectively guide LLM to perform zero-shot VQA tasks. Img2Prompt offers the following benefits: 1) It can flexibly work with various LLMs to perform VQA. 2)~Without the needing of end-to-end training, it significantly reduces the cost of deploying LLM for zero-shot VQA tasks. 3) It achieves comparable or better performance than methods relying on end-to-end training. For example, we outperform Flamingo \cite{Deepmind:Flamingo2022} by 5.6\% on VQAv2. On the challenging A-OKVQA dataset, our method even outperforms few-shot methods by as much as 20\%.

Jason J. Choi, Donggun Lee, Boyang Li et al.

Control invariant sets are crucial for various methods that aim to design safe control policies for systems whose state constraints must be satisfied over an indefinite time horizon. In this article, we explore the connections among reachability, control invariance, and Control Barrier Functions (CBFs). Unlike prior formulations based on backward reachability concepts, we establish a strong link between these three concepts by examining the inevitable Forward Reachable Tube (FRT), which is the set of states such that every trajectory reaching the FRT must have passed through a given initial set of states. First, our findings show that the inevitable FRT is a robust control invariant set if it has a continuously differentiable boundary. If the boundary is not differentiable, the FRT may lose invariance. We also show that any robust control invariant set including the initial set is a superset of the FRT if the boundary of the invariant set is differentiable. Next, we formulate a differential game between the control and disturbance, where the inevitable FRT is characterized by the zero-superlevel set of the value function. By incorporating a discount factor in the cost function of the game, the barrier constraint of the CBF naturally arises in the Hamilton-Jacobi (HJ) equation and determines the optimal policy. The resulting FRT value function serves as a CBF-like function, and conversely, any valid CBF is also a forward reachability value function. We further prove that any $C^1$ supersolution of the HJ equation for the FRT value functions is a valid CBF and characterizes a robust control invariant set that outer-approximates the FRT. Building on this property, finally, we devise a novel method that learns neural control barrier functions, which learn an control invariant superset of the FRT of a given initial set.

Tong Zhang, X. Jessie Yang, Boyang Li

Research in explainable AI (XAI) aims to provide insights into the decision-making process of opaque AI models. To date, most XAI methods offer one-off and static explanations, which cannot cater to the diverse backgrounds and understanding levels of users. With this paper, we investigate if free-form conversations can enhance users' comprehension of static explanations, improve acceptance and trust in the explanation methods, and facilitate human-AI collaboration. Participants are presented with static explanations, followed by a conversation with a human expert regarding the explanations. We measure the effect of the conversation on participants' ability to choose, from three machine learning models, the most accurate one based on explanations and their self-reported comprehension, acceptance, and trust. Empirical results show that conversations significantly improve comprehension, acceptance, trust, and collaboration. Our findings highlight the importance of customized model explanations in the format of free-form conversations and provide insights for the future design of conversational explanations.

Yubin Xiao, Di Wang, Boyang Li et al.

The Traveling Salesman Problem (TSP) is a well-known combinatorial optimization problem with broad real-world applications. Recently, neural networks have gained popularity in this research area because as shown in the literature, they provide strong heuristic solutions to TSPs. Compared to autoregressive neural approaches, non-autoregressive (NAR) networks exploit the inference parallelism to elevate inference speed but suffer from comparatively low solution quality. In this paper, we propose a novel NAR model named NAR4TSP, which incorporates a specially designed architecture and an enhanced reinforcement learning strategy. To the best of our knowledge, NAR4TSP is the first TSP solver that successfully combines RL and NAR networks. The key lies in the incorporation of NAR network output decoding into the training process. NAR4TSP efficiently represents TSP encoded information as rewards and seamlessly integrates it into reinforcement learning strategies, while maintaining consistent TSP sequence constraints during both training and testing phases. Experimental results on both synthetic and real-world TSPs demonstrate that NAR4TSP outperforms five state-of-the-art models in terms of solution quality, inference speed, and generalization to unseen scenarios.

Boyang Li, Hongzhe Shou, Yuanyuan Liang et al.

Existing Chinese toxic content detection methods mainly target sentence-level classification but often fail to provide readable and contiguous toxic evidence spans. We propose \textbf{ToxiTrace}, an explainability-oriented method for BERT-style encoders with three components: (1) \textbf{CuSA}, which refines encoder-derived saliency cues into fine-grained toxic spans with lightweight LLM guidance; (2) \textbf{GCLoss}, a gradient-constrained objective that concentrates token-level saliency on toxic evidence while suppressing irrelevant activations; and (3) \textbf{ARCL}, which constructs sample-specific contrastive reasoning pairs to sharpen the semantic boundary between toxic and non-toxic content. Experiments show that ToxiTrace improves classification accuracy and toxic span extraction while preserving efficient encoder-based inference and producing more coherent, human-readable explanations. We have released the model at https://huggingface.co/ArdLi/ToxiTrace.

Xu Guo, Boyang Li, Han Yu

Prompt tuning, or the conditioning of a frozen pretrained language model (PLM) with soft prompts learned from data, has demonstrated impressive performance on a wide range of NLP tasks. However, prompt tuning requires a large training dataset to be effective and is outperformed by finetuning the entire PLM in data-scarce regimes. Previous work (Gu et al., 2022, Vu et al., 2022) proposed to transfer soft prompts pretrained on the source domain to the target domain. In this paper, we explore domain adaptation for prompt tuning, a problem setting where unlabeled data from the target domain are available during pretraining. We propose bOosting Prompt TunIng with doMain Adaptation (OPTIMA), which regularizes the decision boundary to be smooth around regions where source and target data distributions are similar. Extensive experiments demonstrate that OPTIMA significantly enhances the transferability and sample-efficiency of prompt tuning compared to strong baselines. Moreover, in few-shot settings, OPTIMA exceeds full-model tuning by a large margin.

Jiayun Luo, Boyang Li, Cyril Leung

In the evolution of agriculture to its next stage, Agriculture 5.0, artificial intelligence will play a central role. Controlled-environment agriculture, or CEA, is a special form of urban and suburban agricultural practice that offers numerous economic, environmental, and social benefits, including shorter transportation routes to population centers, reduced environmental impact, and increased productivity. Due to its ability to control environmental factors, CEA couples well with computer vision (CV) in the adoption of real-time monitoring of the plant conditions and autonomous cultivation and harvesting. The objective of this paper is to familiarize CV researchers with agricultural applications and agricultural practitioners with the solutions offered by CV. We identify five major CV applications in CEA, analyze their requirements and motivation, and survey the state of the art as reflected in 68 technical papers using deep learning methods. In addition, we discuss five key subareas of computer vision and how they related to these CEA problems, as well as eleven vision-based CEA datasets. We hope the survey will help researchers quickly gain a bird-eye view of the striving research area and will spark inspiration for new research and development.

Boyang Li, Qing Lu, Weiwen Jiang et al.

In most popular public accessible cryptocurrency systems, the mining pool plays a key role because mining cryptocurrency with the mining pool turns the non-profitable situation into profitable for individual miners. In many recent novel blockchain consensuses, the deep learning training procedure becomes the task for miners to prove their workload, thus the computation power of miners will not purely be spent on the hash puzzle. In this way, the hardware and energy will support the blockchain service and deep learning training simultaneously. While the incentive of miners is to earn tokens, individual miners are motivated to join mining pools to become more competitive. In this paper, we are the first to demonstrate a mining pool solution for novel consensuses based on deep learning. The mining pool manager partitions the full searching space into subspaces and all miners are scheduled to collaborate on the Neural Architecture Search (NAS) tasks in the assigned subspace. Experiments demonstrate that the performance of this type of mining pool is more competitive than an individual miner. Due to the uncertainty of miners' behaviors, the mining pool manager checks the standard deviation of the performance of high reward miners and prepares backup miners to ensure the completion of the tasks of high reward miners.

Xinlei Guan, David Arosemena, Tejaswi Dhandu et al.

The rapid growth of generative AI has introduced new challenges in content moderation and digital forensics. In particular, benign AI-generated images can be paired with harmful or misleading text, creating difficult-to-detect misuse. This contextual misuse undermines the traditional moderation framework and complicates attribution, as synthetic images typically lack persistent metadata or device signatures. We introduce a steganography enabled attribution framework that embeds cryptographically signed identifiers into images at creation time and uses multimodal harmful content detection as a trigger for attribution verification. Our system evaluates five watermarking methods across spatial, frequency, and wavelet domains. It also integrates a CLIP-based fusion model for multimodal harmful-content detection. Experiments demonstrate that spread-spectrum watermarking, especially in the wavelet domain, provides strong robustness under blur distortions, and our multimodal fusion detector achieves an AUC-ROC of 0.99, enabling reliable cross-modal attribution verification. These components form an end-to-end forensic pipeline that enables reliable tracing of harmful deployments of AI-generated imagery, supporting accountability in modern synthetic media environments. Our code is available at GitHub: https://github.com/bli1/steganography

Tong Zhang, Yong Liu, Boyang Li et al.

With the evolution of pre-trained language models, current open-domain dialogue systems have achieved great progress in conducting one-session conversations. In contrast, Multi-Session Conversation (MSC), which consists of multiple sessions over a long term with the same user, is under-investigated. In this paper, we propose History-Aware Hierarchical Transformer (HAHT) for multi-session open-domain dialogue. HAHT maintains a long-term memory of history conversations and utilizes history information to understand current conversation context and generate well-informed and context-relevant responses. Specifically, HAHT first encodes history conversation sessions hierarchically into a history memory. Then, HAHT leverages historical information to facilitate the understanding of the current conversation context by encoding the history memory together with the current context with attention-based mechanisms. Finally, to explicitly utilize historical information, HAHT uses a history-aware response generator that switches between a generic vocabulary and a history-aware vocabulary. Experimental results on a large-scale MSC dataset suggest that the proposed HAHT model consistently outperforms baseline models. Human evaluation results support that HAHT generates more human-like, context-relevant and history-relevant responses than baseline models.

Rixiang Ni, Boyang Li, Jun Chen et al.

Infrared small target detection (IRSTD) aims to separate small targets from clutter backgrounds. Extensive research is dedicated to the pixel-level supervision-guided "encoder-decoder" segmentation paradigm. Although having achieved promising performance, they neglect the fact that small targets only occupy a few pixels and are usually accompanied with blurred boundary caused by clutter backgrounds. Based on this observation, we argue that the first principle of IRSTD should be target localization instead of separating all target region accompanied with indistinguishable background noise. In this paper, we reformulate IRSTD as a centroid regression task and propose a novel Single-Point Supervision guided Infrared Probabilistic Response Encoding method (namely, SPIRE), which is indeed challenging due to the mismatch between reduced supervision network and equivalent output. Specifically, we first design a Point-Response Prior Supervision (PRPS), which transforms single-point annotations into probabilistic response map consistent with infrared point-target response characteristics, with a High-Resolution Probabilistic Encoder (HRPE) that enables encoder-only, end-to-end regression without decoder reconstruction. By preserving high-resolution features and increasing effective supervision density, SPIRE alleviates optimization instability under sparse target distributions. Finally, extensive experiments on various IRSTD benchmarks, including SIRST-UAVB and SIRST4 demonstrate that SPIRE achieves competitive target-level detection performance with consistently low false alarm rate (Fa) and significantly reduced computational cost. Code is publicly available at: https://github.com/NIRIXIANG/SPIRE-IRSTD.

Zilin Du, Yunxin Li, Xu Guo et al.

Contemporary news reporting increasingly features multimedia content, motivating research on multimedia event extraction. However, the task lacks annotated multimodal training data and artificially generated training data suffer from distribution shift from real-world data. In this paper, we propose Cross-modality Augmented Multimedia Event Learning (CAMEL), which successfully utilizes artificially generated multimodal training data and achieves state-of-the-art performance. We start with two labeled unimodal datasets in text and image respectively, and generate the missing modality using off-the-shelf image generators like Stable Diffusion and image captioners like BLIP. After that, we train the network on the resultant multimodal datasets. In order to learn robust features that are effective across domains, we devise an iterative and gradual training strategy. Substantial experiments show that CAMEL surpasses state-of-the-art (SOTA) baselines on the M2E2 benchmark. On multimedia events in particular, we outperform the prior SOTA by 4.2% F1 on event mention identification and by 9.8% F1 on argument identification, which indicates that CAMEL learns synergistic representations from the two modalities. Our work demonstrates a recipe to unleash the power of synthetic training data in structured prediction.

Boyang Li, Zheng Gong, Sylvia Herbert

In this article, we consider the infinite-horizon reach-avoid (RA) and stabilize-avoid (SA) zero-sum game problems for general nonlinear continuous-time systems, where the goal is to find the set of states that can be controlled to reach or stabilize to a target set, without violating constraints even under the worst-case disturbance. Based on the Hamilton-Jacobi reachability method, we address the RA problem by designing a new Lipschitz continuous RA value function, whose zero sublevel set exactly characterizes the RA set. We establish that the associated Bellman backup operator is contractive and that the RA value function is the unique viscosity solution of a Hamilton-Jacobi variational inequality. Finally, we develop a two-step framework for the SA problem by integrating our RA strategies with a recently proposed Robust Control Lyapunov-Value Function, thereby ensuring both target reachability and long-term stability. We numerically verify our RA and SA frameworks on a 3D Dubins car system to demonstrate the efficacy of the proposed approach.

Yidan Sun, Qin Chao, Yangfeng Ji et al.

Despite recent advances of AI, story understanding remains an open and under-investigated problem. We collect, preprocess, and publicly release a video-language story dataset, Synopses of Movie Narratives (SyMoN), containing 5,193 video summaries of popular movies and TV series with a total length of 869 hours. SyMoN captures naturalistic storytelling videos made by human creators and intended for a human audience. As a prototypical and naturalistic story dataset, SyMoN features high coverage of multimodal story events and abundant mental-state descriptions. Its use of storytelling techniques cause cross-domain semantic gaps that provide appropriate challenges to existing models. We establish benchmarks on video-text retrieval and zero-shot alignment on movie summary videos, which showcase the importance of in-domain data and long-term memory in story understanding. With SyMoN, we hope to lay the groundwork for progress in multimodal story understanding.

Chaoqun Liu, Qin Chao, Wenxuan Zhang et al.

Large Language Models (LLMs) have demonstrated remarkable performance through supervised fine-tuning or in-context learning using gold labels. However, this paradigm is limited by the availability of gold labels, while in certain scenarios, LLMs may need to perform tasks that are too complex for humans to provide such labels. To tackle this challenge, this study explores whether solely utilizing unlabeled data can elicit strong model capabilities. We propose a new paradigm termed zero-to-strong generalization. We iteratively prompt LLMs to annotate unlabeled data and retain high-quality labels by filtering. Surprisingly, we obverse that this iterative process gradually unlocks LLMs' potential on downstream tasks. Our experiments on extensive classification and reasoning tasks confirm the effectiveness of our proposed framework. Our analysis indicates that this paradigm is effective for both in-context learning and fine-tuning, and for various model sizes.

Boyang Li, Bingyu Shen, Qing Lu et al.

The continuous thriving of the Blockchain society motivates research in novel designs of schemes supporting cryptocurrencies. Previously multiple Proof-of-Deep-Learning(PoDL) consensuses have been proposed to replace hashing with useful work such as deep learning model training tasks. The energy will be more efficiently used while maintaining the ledger. However deep learning models are problem-specific and can be extremely complex. Current PoDL consensuses still require much work to realize in the real world. In this paper, we proposed a novel consensus named Proof-of-Federated-Learning-Subchain(PoFLSC) to fill the gap. We applied a subchain to record the training, challenging, and auditing activities and emphasized the importance of valuable datasets in partner selection. We simulated 20 miners in the subchain to demonstrate the effectiveness of PoFLSC. When we reduce the pool size concerning the reservation priority order, the drop rate difference in the performance in different scenarios further exhibits that the miner with a higher Shapley Value (SV) will gain a better opportunity to be selected when the size of the subchain pool is limited. In the conducted experiments, the PoFLSC consensus supported the subchain manager to be aware of reservation priority and the core partition of contributors to establish and maintain a competitive subchain.

Devaansh Gupta, Boyang Li

Combining Large Language Models (LLMs) with heuristic search algorithms like A* holds the promise of enhanced LLM reasoning and scalable inference. To accelerate training and reduce computational demands, we investigate the coreset selection problem for the training data of LLM heuristic learning. Few methods to learn the heuristic functions consider the interaction between the search algorithm and the machine learning model. In this work, we empirically disentangle the requirements of A* search algorithm from the requirements of the LLM to generalise on this task. Surprisingly, we find an overlap between their requirements; A* requires more accurate predictions on search nodes near the goal, and LLMs need the same set of nodes for effective generalisation. With these insights, we derive a data-selection distribution for learning LLM-based heuristics. On three classical planning domains, maze navigation, Sokoban and sliding tile puzzles, our technique reduces the number of iterations required to find the solutions by up to 15x, with a wall-clock speed-up of search up to 5x. The codebase is at https://github.com/devaansh100/a_star.

Qin Chao, Eunsoo Kim, Boyang Li

Investments in movie production are associated with a high level of risk as movie revenues have long-tailed and bimodal distributions. Accurate prediction of box-office revenue may mitigate the uncertainty and encourage investment. However, learning effective representations for actors, directors, and user-generated content-related keywords remains a challenging open problem. In this work, we investigate the effects of self-supervised pretraining and propose visual grounding of content keywords in objects from movie posters as a pertaining objective. Experiments on a large dataset of 35,794 movies demonstrate significant benefits of self-supervised training and visual grounding. In particular, visual grounding pretraining substantially improves learning on movies with content keywords and achieves 14.5% relative performance gains compared to a finetuned BERT model with identical architecture.

Boyang Li, Xinyi Ying, Ruojing Li et al.

In this paper, we briefly summarize the first competition on resource-limited infrared small target detection (namely, LimitIRSTD). This competition has two tracks, including weakly-supervised infrared small target detection (Track 1) and lightweight infrared small target detection (Track 2). 46 and 60 teams successfully registered and took part in Tracks 1 and Track 2, respectively. The top-performing methods and their results in each track are described with details. This competition inspires the community to explore the tough problems in the application of infrared small target detection, and ultimately promote the deployment of this technology under limited resource.

Miles Q. Li, Benjamin C. M. Fung, Boyang Li et al.

Existing white-box jailbreak attacks against aligned LLMs typically append discrete adversarial suffixes to the user prompt, which visibly alters the prompt and operates in a combinatorial token space. Prior work has avoided directly optimizing the embeddings of the original prompt tokens, presumably because perturbing them risks destroying the prompt's semantic content. We propose Prompt Embedding Optimization (PEO), a multi-round white-box jailbreak that directly optimizes the embeddings of the original prompt tokens without appending any adversarial tokens, and show that the concern is unfounded: the optimized embeddings remain close enough to their originals that the visible prompt string is preserved exactly after nearest-token projection, and quantitative analysis shows the model's responses stay on topic for the large majority of prompts. PEO combines continuous embedding-space optimization with structured continuation targets and an adaptive failure-focused schedule. Counterintuitively, later PEO rounds can benefit from heuristic composite response scaffolds that are not natural standalone templates, yet ASR-Judge shows that the resulting gains are not merely empty formatting or scaffold-only outputs. Across two standard harmful-behavior benchmarks and competing white-box attacks spanning discrete suffix search, appended adversarial embeddings, and search-based adversarial generation, PEO outperforms all of them in our experiments.

Zhiyuan Jiang, Weihao Hong, Xinlei Guan et al.

Vision-Language Models (VLMs) are increasingly deployed in settings where reliable visual grounding carries operational consequences, yet their behavior under progressively coercive prompt phrasing remains undercharacterized. Existing hallucination benchmarks predominantly rely on neutral prompts and binary detection, leaving open how both the incidence and the intensity of fabrication respond to graded linguistic pressure across structurally distinct task types. We present Ghost-100, a procedurally constructed benchmark of 800 synthetically generated images spanning eight categories across three task families -- text-illegibility, time-reading, and object-absence -- each designed under a negative-ground-truth principle that guarantees the queried target is absent, illegible, or indeterminate by construction. Every image is paired with five prompts drawn from a structured 5-Level Prompt Intensity Framework, holding the image and task identity fixed while varying only directive force, so that tone is isolated as the sole independent variable. We adopt a dual-track evaluation protocol: a rule-based H-Rate measuring the proportion of responses in which a model crosses from grounded refusal into unsupported positive commitment, and a GPT-4o-mini-judged H-Score on a 1-5 scale characterizing the confidence and specificity of fabrication once it occurs. We additionally release a three-stage automated validation workflow, which retrospectively confirms 717 of 800 images as strictly compliant. Evaluating nine open-weight VLMs, we find that H-Rate and H-Score dissociate substantially across model families, reading-style and presence-detection subsets respond to prompt pressure in qualitatively different ways, and several models exhibit non-monotonic sensitivity peaking at intermediate tone levels -- patterns that aggregate metrics obscure.

Zhuoyi Yang, Jiapeng Yu, Reuben Tan et al.

Understanding instructional videos requires recognizing fine-grained actions and modeling their temporal relations, which remains challenging for current Video Foundation Models (VFMs). This difficulty stems from noisy web supervision and a pervasive "static bias", where models rely on objects rather than motion cues. To address this, we propose InstrAction, a pretraining framework for instructional videos' action-centric representations. We first introduce a data-driven strategy, which filters noisy captions and generates action-centric hard negatives to disentangle actions from objects during contrastive learning. At the visual feature level, an Action Perceiver extracts motion-relevant tokens from redundant video encodings. Beyond contrastive learning, we introduce two auxiliary objectives: Dynamic Time Warping alignment (DTW-Align) for modeling sequential temporal structure, and Masked Action Modeling (MAM) for strengthening cross-modal grounding. Finally, we introduce the InstrAct Bench to evaluate action-centric understanding, where our method consistently outperforms state-of-the-art VFMs on semantic reasoning, procedural logic, and fine-grained retrieval tasks.

Jie Zhu, Yimin Tian, Boyang Li et al.

This paper introduces \textbf{FinMCP-Bench}, a novel benchmark for evaluating large language models (LLMs) in solving real-world financial problems through tool invocation of financial model context protocols. FinMCP-Bench contains 613 samples spanning 10 main scenarios and 33 sub-scenarios, featuring both real and synthetic user queries to ensure diversity and authenticity. It incorporates 65 real financial MCPs and three types of samples, single tool, multi-tool, and multi-turn, allowing evaluation of models across different levels of task complexity. Using this benchmark, we systematically assess a range of mainstream LLMs and propose metrics that explicitly measure tool invocation accuracy and reasoning capabilities. FinMCP-Bench provides a standardized, practical, and challenging testbed for advancing research on financial LLM agents.

Anthony Meng Huat Tiong, Junqi Zhao, Boyang Li et al.

Vision-language (VL) models, pretrained on colossal image-text datasets, have attained broad VL competence that is difficult to evaluate. A common belief is that a small number of VL skills underlie the variety of VL tests. In this paper, we perform a large-scale transfer learning experiment aimed at discovering latent VL skills from data. We reveal interesting characteristics that have important implications for test suite design. First, generation tasks suffer from a length bias, suggesting benchmarks should balance tasks with varying output lengths. Second, we demonstrate that factor analysis successfully identifies reasonable yet surprising VL skill factors, suggesting benchmarks could leverage similar analyses for task selection. Finally, we present a new dataset, OLIVE (https://github.com/jq-zh/olive-dataset), which simulates user instructions in the wild and presents challenges dissimilar to all datasets we tested. Our findings contribute to the design of balanced and broad-coverage vision-language evaluation methods.

Rengan Xu, Junjie Yang, Yifan Xu et al.

The integration of hardware accelerators has significantly advanced the capabilities of modern recommendation systems, enabling the exploration of complex ranking paradigms previously deemed impractical. However, the GPU-based computational costs present substantial challenges. In this paper, we demonstrate our development of an efficiency-driven approach to explore these paradigms, moving beyond traditional reliance on native PyTorch modules. We address the specific challenges posed by ranking models' dependence on categorical features, which vary in length and complicate GPU utilization. We introduce Jagged Feature Interaction Kernels, a novel method designed to extract fine-grained insights from long categorical features through efficient handling of dynamically sized tensors. We further enhance the performance of attention mechanisms by integrating Jagged tensors with Flash Attention. Our novel Jagged Flash Attention achieves up to 9x speedup and 22x memory reduction compared to dense attention. Notably, it also outperforms dense flash attention, with up to 3x speedup and 53% more memory efficiency. In production models, we observe 10% QPS improvement and 18% memory savings, enabling us to scale our recommendation systems with longer features and more complex architectures.

Wenyu Zhang, Wei En Ng, Lixin Ma et al.

Current vision-language models may grasp basic spatial cues and simple directions (e.g. left, right, front, back), but struggle with the multi-dimensional spatial reasoning necessary for human-like understanding and real-world applications. To address this gap, we develop SPHERE (Spatial Perception and Hierarchical Evaluation of REasoning), a hierarchical evaluation framework supported by a new human-annotated dataset. SPHERE systematically probes models across increasing levels of complexity, from fundamental skills to multi-skill integration and high-level reasoning that combines spatial, visual, and logical understanding. Benchmark evaluation of state-of-the-art models reveals significant deficiencies, especially in reasoning about distance and proximity, understanding both egocentric and allocentric perspectives, and applying spatial logic in physical contexts. These findings expose critical blind spots in existing models and underscore the need for more advanced spatial reasoning techniques, driving the development of vision-language models that align more closely with human spatial cognition. The SPHERE benchmark is available at https://github.com/zwenyu/SPHERE-VLM.

Long Cheng, Jiafei Duan, Yi Ru Wang et al. · uw

Pointing serves as a fundamental and intuitive mechanism for grounding language within visual contexts, with applications spanning robotics, assistive technologies, and interactive AI systems. While recent multimodal models have started to support pointing capabilities, existing benchmarks typically focus only on referential object localization tasks. We introduce PointArena, a comprehensive platform for evaluating multimodal pointing across diverse reasoning scenarios. PointArena comprises three components: (1) Point-Bench, a curated dataset containing approximately 1,000 pointing tasks across five reasoning categories; (2) Point-Battle, an interactive, web-based arena facilitating blind, pairwise model comparisons, which has already gathered over 4,500 anonymized votes; and (3) Point-Act, a real-world robotic manipulation system allowing users to directly evaluate multimodal model pointing capabilities in practical settings. We conducted extensive evaluations of both state-of-the-art open-source and proprietary multimodal models. Results indicate that Molmo-72B consistently outperforms other models, though proprietary models increasingly demonstrate comparable performance. Additionally, we find that supervised training specifically targeting pointing tasks significantly enhances model performance. Across our multi-stage evaluation pipeline, we also observe strong correlations, underscoring the critical role of precise pointing capabilities in enabling multimodal models to effectively bridge abstract reasoning with concrete, real-world actions. Project page: https://pointarena.github.io/

Aditya Chinchure, Sahithya Ravi, Raymond Ng et al.

The commonsense reasoning capabilities of vision-language models (VLMs), especially in abductive reasoning and defeasible reasoning, remain poorly understood. Most benchmarks focus on typical visual scenarios, making it difficult to discern whether model performance stems from keen perception and reasoning skills, or reliance on pure statistical recall. We argue that by focusing on atypical events in videos, clearer insights can be gained on the core capabilities of VLMs. Explaining and understanding such out-of-distribution events requires models to extend beyond basic pattern recognition and regurgitation of their prior knowledge. To this end, we introduce BlackSwanSuite, a benchmark for evaluating VLMs' ability to reason about unexpected events through abductive and defeasible tasks. Our tasks artificially limit the amount of visual information provided to models while questioning them about hidden unexpected events, or provide new visual information that could change an existing hypothesis about the event. We curate a comprehensive benchmark suite comprising over 3,800 MCQ, 4,900 generative and 6,700 yes/no questions, spanning 1,655 videos. After extensively evaluating various state-of-the-art VLMs, including GPT-4o and Gemini 1.5 Pro, as well as open-source VLMs such as LLaVA-Video, we find significant performance gaps of up to 32% from humans on these tasks. Our findings reveal key limitations in current VLMs, emphasizing the need for enhanced model architectures and training strategies. Our data and leaderboard is available at blackswan.cs.ubc.ca.

Landy Jimenez, Mariah Weatherspoon, Bingyu Shen et al.

Proof-of-Work (PoW) blockchain consensus consumes vast computational resources without producing useful output, while the rapid growth of large language model (LLM) agents has created unprecedented demand for GPU computation. We present HadAgent, a decentralized agentic AI serving system that replaces hash-based mining with Proof-of-Inference (PoI), a consensus mechanism in which nodes earn block-creation rights by executing deterministic LLM inference tasks. Because verification requires only re-executing a single forward pass under identical conditions, cross-node verification operates at consensus speed. HadAgent organizes validated records into a three-lane block body with dedicated DATA, MODEL, and PROOF channels, each protected by an independent Merkle root for fine-grained tamper detection. A two-tier node architecture classifies secondary nodes as trusted or non-trusted based on historical behavior: trusted nodes serve inference results in real time through optimistic execution, while non-trusted nodes must undergo full consensus verification. A harness layer monitors node behavior through heartbeat probes, anomaly detection via deterministic recomputation, and automated trust management, creating a self-correcting feedback loop that isolates malicious or unreliable participants. Experiments on a prototype implementation demonstrate 100% detection rate and 0% false positive rate for tampered records, sub-millisecond validation latency for record and hub operations, and effective harness convergence that excludes adversarial nodes within two rounds while promoting honest nodes to trusted status within five rounds.

Shuai Liu, Quanmin Liang, Zefeng Li et al.

Multi-sensor fusion is crucial for improving the performance and robustness of end-to-end autonomous driving systems. Existing methods predominantly adopt either attention-based flatten fusion or bird's eye view fusion through geometric transformations. However, these approaches often suffer from limited interpretability or dense computational overhead. In this paper, we introduce GaussianFusion, a Gaussian-based multi-sensor fusion framework for end-to-end autonomous driving. Our method employs intuitive and compact Gaussian representations as intermediate carriers to aggregate information from diverse sensors. Specifically, we initialize a set of 2D Gaussians uniformly across the driving scene, where each Gaussian is parameterized by physical attributes and equipped with explicit and implicit features. These Gaussians are progressively refined by integrating multi-modal features. The explicit features capture rich semantic and spatial information about the traffic scene, while the implicit features provide complementary cues beneficial for trajectory planning. To fully exploit rich spatial and semantic information in Gaussians, we design a cascade planning head that iteratively refines trajectory predictions through interactions with Gaussians. Extensive experiments on the NAVSIM and Bench2Drive benchmarks demonstrate the effectiveness and robustness of the proposed GaussianFusion framework. The source code will be released at https://github.com/Say2L/GaussianFusion.

Shuai Liu, Boyang Li, Zhiyu Fang et al.

Recently, significant progress has been made in the research of 3D object detection. However, most prior studies have focused on the utilization of center-based or anchor-based label assignment schemes. Alternative label assignment strategies remain unexplored in 3D object detection. We find that the center-based label assignment often fails to generate sufficient positive samples for training, while the anchor-based label assignment tends to encounter an imbalanced issue when handling objects of varying scales. To solve these issues, we introduce a dynamic cross label assignment (DCLA) scheme, which dynamically assigns positive samples for each object from a cross-shaped region, thus providing sufficient and balanced positive samples for training. Furthermore, to address the challenge of accurately regressing objects with varying scales, we put forth a rotation-weighted Intersection over Union (RWIoU) metric to replace the widely used L1 metric in regression loss. Extensive experiments demonstrate the generality and effectiveness of our DCLA and RWIoU-based regression loss. The Code will be available at https://github.com/Say2L/DCDet.git.

Shuai Liu, Boyang Li, Zhiyu Fang et al.

LiDAR-based 3D object detection has made impressive progress recently, yet most existing models are black-box, lacking interpretability. Previous explanation approaches primarily focus on analyzing image-based models and are not readily applicable to LiDAR-based 3D detectors. In this paper, we propose a feature factorization activation map (FFAM) to generate high-quality visual explanations for 3D detectors. FFAM employs non-negative matrix factorization to generate concept activation maps and subsequently aggregates these maps to obtain a global visual explanation. To achieve object-specific visual explanations, we refine the global visual explanation using the feature gradient of a target object. Additionally, we introduce a voxel upsampling strategy to align the scale between the activation map and input point cloud. We qualitatively and quantitatively analyze FFAM with multiple detectors on several datasets. Experimental results validate the high-quality visual explanations produced by FFAM. The Code will be available at \url{https://github.com/Say2L/FFAM.git}.

Zilin Du, Haoxin Li, Xu Guo et al.

The task of multimodal relation extraction has attracted significant research attention, but progress is constrained by the scarcity of available training data. One natural thought is to extend existing datasets with cross-modal generative models. In this paper, we consider a novel problem setting, where only unimodal data, either text or image, are available during training. We aim to train a multimodal classifier from synthetic data that perform well on real multimodal test data. However, training with synthetic data suffers from two obstacles: lack of data diversity and label information loss. To alleviate the issues, we propose Mutual Information-aware Multimodal Iterated Relational dAta GEneration (MI2RAGE), which applies Chained Cross-modal Generation (CCG) to promote diversity in the generated data and exploits a teacher network to select valuable training samples with high mutual information with the ground-truth labels. Comparing our method to direct training on synthetic data, we observed a significant improvement of 24.06% F1 with synthetic text and 26.42% F1 with synthetic images. Notably, our best model trained on completely synthetic images outperforms prior state-of-the-art models trained on real multimodal data by a margin of 3.76% in F1. Our codebase will be made available upon acceptance.

Shuai Liu, Mingyue Cui, Boyang Li et al.

Fully sparse 3D detectors have recently gained significant attention due to their efficiency in long-range detection. However, sparse 3D detectors extract features only from non-empty voxels, which impairs long-range interactions and causes the center feature missing. The former weakens the feature extraction capability, while the latter hinders network optimization. To address these challenges, we introduce the Fully Sparse Hybrid Network (FSHNet). FSHNet incorporates a proposed SlotFormer block to enhance the long-range feature extraction capability of existing sparse encoders. The SlotFormer divides sparse voxels using a slot partition approach, which, compared to traditional window partition, provides a larger receptive field. Additionally, we propose a dynamic sparse label assignment strategy to deeply optimize the network by providing more high-quality positive samples. To further enhance performance, we introduce a sparse upsampling module to refine downsampled voxels, preserving fine-grained details crucial for detecting small objects. Extensive experiments on the Waymo, nuScenes, and Argoverse2 benchmarks demonstrate the effectiveness of FSHNet. The code is available at https://github.com/Say2L/FSHNet.

Adimulya Kartiyasa, Bao Gia Cao, Boyang Li

Recently there have been intensifying efforts to improve the understanding of Indonesian cultures by large language models (LLMs). An attractive source of cultural knowledge that has been largely overlooked is local journals of social science, which likely contain substantial cultural studies from a native perspective. We present a novel text dataset of journal article passages, created from 151 open-source Indonesian social science journals, called IndoSoSci. We demonstrate an effective recipe for injecting Indonesian cultural knowledge therein into LLMs: extracting the facts related to Indonesian culture, and apply retrieval-augmented generation (RAG) with LLM-generated hypothetical documents as queries during retrieval. The proposed recipe yields strong performance gains over several strong baselines on the IndoCulture benchmark. Additionally, by combining IndoSoSci with Indonesian Wikipedia, we set a new state-of-the-art accuracy on the IndoCulture benchmark.

Jin-Seop Lee, SungJoon Lee, SeongJun Jung et al.

Video Temporal Grounding (VTG) aims to localize a temporal segment in a video corresponding to a natural language query. However, existing VTG models assume that a relevant segment always exists, causing them to always predict a target segment even when the query is irrelevant to the video. While recent approaches attempt to handle irrelevant queries, they can only reject those that are entirely unrelated to the video and still fail to handle hard-irrelevant queries that are semantically similar but not actually relevant. To address this, we propose Refusal-Aware Reinforcement Fine-Tuning (RA-RFT) to effectively refuse hard-irrelevant queries in VTG. Our method is based on the Group Relative Policy Optimization (GRPO) framework and integrates four reward objectives-format, refuse-IoU, explain, and query correction-to improve both relevance discrimination and fine-grained semantic reasoning. In addition, to effectively support RA-RFT, we construct a Hard-Irrelevant VTG (HI-VTG) dataset, which includes hard-irrelevant queries and their refusal answers. We demonstrate the effectiveness of our method across various relevance-aware VTG scenarios, including hard-irrelevant VTG, simply-shuffled RA-VTG, and human-annotated RA-VTG settings. We also show that the proposed method is scalable by applying it to various LVLM-based VTG models. Our code is available at https://github.com/JINSUBY/RA-RFT.

Wenju Sun, Qingyong Li, Wen Wang et al.

Multi-task model merging aims to consolidate knowledge from multiple fine-tuned task-specific experts into a unified model while minimizing performance degradation. Existing methods primarily approach this by minimizing differences between task-specific experts and the unified model, either from a parameter-level or a task-loss perspective. However, parameter-level methods exhibit a significant performance gap compared to the upper bound, while task-loss approaches entail costly secondary training procedures. In contrast, we observe that performance degradation closely correlates with feature drift, i.e., differences in feature representations of the same sample caused by model merging. Motivated by this observation, we propose Layer-wise Optimal Task Vector Merging (LOT Merging), a technique that explicitly minimizes feature drift between task-specific experts and the unified model in a layer-by-layer manner. LOT Merging can be formulated as a convex quadratic optimization problem, enabling us to analytically derive closed-form solutions for the parameters of linear and normalization layers. Consequently, LOT Merging achieves efficient model consolidation through basic matrix operations. Extensive experiments across vision and vision-language benchmarks demonstrate that LOT Merging significantly outperforms baseline methods, achieving improvements of up to 4.4% (ViT-B/32) over state-of-the-art approaches. The source code is available at https://github.com/SunWenJu123/model-merging.

Xinyi Ying, Chao Xiao, Ruojing Li et al.

Small object detection (SOD) has been a longstanding yet challenging task for decades, with numerous datasets and algorithms being developed. However, they mainly focus on either visible or thermal modality, while visible-thermal (RGBT) bimodality is rarely explored. Although some RGBT datasets have been developed recently, the insufficient quantity, limited category, misaligned images and large target size cannot provide an impartial benchmark to evaluate multi-category visible-thermal small object detection (RGBT SOD) algorithms. In this paper, we build the first large-scale benchmark with high diversity for RGBT SOD (namely RGBT-Tiny), including 115 paired sequences, 93K frames and 1.2M manual annotations. RGBT-Tiny contains abundant targets (7 categories) and high-diversity scenes (8 types that cover different illumination and density variations). Note that, over 81% of targets are smaller than 16x16, and we provide paired bounding box annotations with tracking ID to offer an extremely challenging benchmark with wide-range applications, such as RGBT fusion, detection and tracking. In addition, we propose a scale adaptive fitness (SAFit) measure that exhibits high robustness on both small and large targets. The proposed SAFit can provide reasonable performance evaluation and promote detection performance. Based on the proposed RGBT-Tiny dataset and SAFit measure, extensive evaluations have been conducted, including 23 recent state-of-the-art algorithms that cover four different types (i.e., visible generic detection, visible SOD, thermal SOD and RGBT object detection). Project is available at https://github.com/XinyiYing/RGBT-Tiny.