Recent advances in large language models (LLMs) have greatly improved their reasoning and decision-making abilities when deployed as agents. Richer reasoning, however, often comes at the cost of longer chain of thought (CoT), hampering interaction efficiency in real-world scenarios. Nevertheless, there still lacks systematic definition of LLM‑Agent efficiency, hindering targeted improvements. To this end, we introduce dual‑efficiency, comprising (i) step-level efficiency, which minimizes tokens per step, and (ii) trajectory-level efficiency, which minimizes the number of steps to complete a task. Building on this definition, we propose DEPO, a dual-efficiency preference‑based optimization method that jointly rewards succinct responses and fewer action steps. Experiments on WebShop and BabyAI show that DEPO cuts token usage by up to 60.9% and steps by up to 26.9%, while achieving up to a 29.3% improvement in task performance. DEPO also generalizes to three out-of-domain math benchmarks and retains its efficiency gains when trained on only 25% of the data.

Image clustering is a classic problem in computer vision, which categorizes images into different groups. Recent studies utilize nouns as external semantic knowledge to improve clustering performance. However, these methods often overlook the inherent ambiguity of nouns, which can distort semantic representations and degrade clustering quality. To address this issue, we propose a hierarChical semAntic alignmEnt method for image clustering, dubbed CAE, which improves clustering performance in a training-free manner. In our approach, we incorporate two complementary types of textual semantics: caption-level descriptions, which convey fine-grained attributes of image content, and noun-level concepts, which represent high-level object categories. We first select relevant nouns from WordNet and descriptions from caption datasets to construct a semantic space aligned with image features. Then, we design a residual attention mechanism to further enhance the discriminability of this space. Finally, we combine the enhanced semantic and image features to perform clustering. Extensive experiments across 8 datasets demonstrate the effectiveness of our method, notably surpassing the state-of-the-art training-free approach with a 4.2% improvement in accuracy and a 2.9% improvement in adjusted rand index (ARI) on the ImageNet-1K dataset.

Xingyu Zhu, Junfeng Fang, Shuo Wang, Beier Zhu, Zhicai Wang, Yonghui Yang, Xiangnan He. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Recent advances in text-to-image (T2I) generation have led to impressive visual results. However, these models still face significant challenges when handling complex prompts-particularly those involving multiple subjects with distinct attributes. Inspired by the human drawing process, which first outlines the composition and then incrementally adds details, we propose Detail++, a training-free framework that introduces a novel Progressive Detail Injection (PDI) strategy to address this limitation. Specifically, we decompose a complex prompt into a sequence of simplified sub-prompts, guiding the generation process in stages. This staged generation leverages the inherent layout-controlling capacity of self-attention to first ensure global composition, followed by precise refinement. To achieve accurate binding between attributes and corresponding subjects, we exploit cross-attention mechanisms and further introduce a Centroid Alignment Loss at test time to reduce binding noise and enhance attribute consistency. Extensive experiments on T2I-CompBench and a newly constructed style composition benchmark demonstrate that Detail++ significantly outperforms existing methods, particularly in scenarios involving multiple objects and complex stylistic conditions.

Distribution estimation is a pivotal strategy in few-shot learning (FSL) to mitigate data scarcity by sampling from estimated distributions, utilizing statistical properties (mean and variance) transferred from related base categories. However, category-level estimation alone often fails to generate representative samples due to significant dissimilarities between base and novel categories, leading to suboptimal performance. To address this limitation, we propose Hybrid Granularity Distribution Estimation (HGDE), which integrates both coarse-grained category-level statistics and fine-grained instance-level statistics. By leveraging instance statistics from the nearest base samples, HGDE enhances the characterization of novel categories, capturing subtle features that category-level estimation overlooks. These statistics are fused through linear interpolation to form a robust distribution for novel categories, ensuring both diversity and representativeness in generated samples. Additionally, HGDE employs refined estimation techniques, such as weighted summation for mean calculation and principal component retention for covariance, to further improve accuracy. Empirical evaluations on four FSL benchmarks, including Mini-ImageNet, Tiered-ImageNet, CUB and CIFAR-FS, demonstrate that HGDE offers effective distribution estimation capabilities and leads to notable accuracy gains, with improvements of more than 1.8% in 1-shot tasks on CUB. These results highlight HGDE's ability to balance mean precision and variance diversity, making it a versatile and effective solution for FSL.

Hallucinations in Large Vision-Language Models (LVLMs) significantly undermine their reliability, motivating researchers to explore the causes of hallucination. However, most studies primarily focus on the language aspect rather than the visual. In this paper, we address how LVLMs process visual information and whether this process causes hallucination. Firstly, we use the attention lens to identify the stages at which LVLMs handle visual data, discovering that the middle layers are crucial. Moreover, we find that these layers can be further divided into two stages: "visual information enrichment" and "semantic refinement" which respectively propagate visual data to object tokens and interpret it through text. By analyzing attention patterns during the visual information enrichment stage, we find that real tokens consistently receive higher attention weights than hallucinated ones, serving as a strong indicator of hallucination. Further examination of multi-head attention maps reveals that hallucination tokens often result from heads interacting with inconsistent objects. Based on these insights, we propose a simple inference-time method that adjusts visual attention by integrating information across various heads. Extensive experiments demonstrate that this approach effectively mitigates hallucinations in mainstream LVLMs without additional training costs.1

Text-driven style transfer aims to merge the style of a reference image with content described by a text prompt. Recent advancements in text-to-image models have improved the nuance of style transformations, yet significant challenges remain, particularly with overfitting to reference styles, limiting stylistic control, and misaligning with textual content. In this paper, we propose three complementary strategies to address these issues. First, we introduce a cross-modal Adaptive Instance Normalization (AdaIN) mechanism for better integration of style and text features, enhancing alignment. Second, we develop a Style-based Classifier-Free Guidance (SCFG) approach that enables selective control over stylistic elements, reducing irrelevant influences. Finally, we incorporate a teacher model during early generation stages to stabilize spatial layouts and mitigate artifacts. Our extensive evaluations demonstrate significant improvements in style transfer quality and alignment with textual prompts. Furthermore, our approach can be integrated into existing style transfer frameworks without fine-tuning.

While image-text foundation models have succeeded across diverse downstream tasks, they still face challenges in the presence of spurious correlations between the input and label. To address this issue, we propose a simple three-step approach-Project-Probe-Aggregate (PPA)-that enables parameter-efficient fine-tuning for foundation models without relying on group annotations. Building upon the failure-based debiasing scheme, our method, PPA, improves its two key components: minority samples identification and the robust training algorithm. Specifically, we first train biased classifiers by projecting image features onto the nullspace of class proxies from text encoders. Next, we infer group labels using the biased classifier and probe group targets with prior correction. Finally, we aggregate group weights of each class to produce the debiased classifier Our theoretical analysis shows that our PPA enhances minority group identification and is Bayes optimal for minimizing the balanced group error, mitigating spurious correlations. Extensive experimental results confirm the effectiveness of our PPA: it outperforms the state-of-the-art by an average worst-group accuracy while requiring less than 0.01% tunable parameters without training group labels.

With the increasing attention to pre-trained vision-language models (VLMs), \eg, CLIP, substantial efforts have been devoted to many downstream tasks, especially in test-time adaptation (TTA). However, previous works focus on learning prototypes only in the textual modality while overlooking the ambiguous semantics in class names. These ambiguities lead to textual prototypes that are insufficient to capture visual concepts, resulting in limited performance. To address this issue, we introduce \textbf{ProtoMM}, a training-free framework that constructs multimodal prototypes to adapt VLMs during the test time. By viewing the prototype as a discrete distribution over the textual descriptions and visual particles, ProtoMM has the ability to combine the multimodal features for comprehensive prototype learning. More importantly, the visual particles are dynamically updated as the testing stream flows. This allows our multimodal prototypes to continually learn from the data, enhancing their generalizability in unseen scenarios. In addition, we quantify the importance of the prototypes and test images by formulating their semantic distance as an optimal transport problem. Extensive experiments on 15 zero-shot benchmarks demonstrate the effectiveness of our method, achieving a 1.03\% average accuracy improvement over state-of-the-art methods on ImageNet and its variant datasets.

Diffusion models (DMs) have achieved state-of-the-art generative performance but suffer from high sampling latency due to their sequential denoising nature. Existing solver-based acceleration methods often face image quality degradation under a low-latency budget. In this paper, we propose the Ensemble Parallel Direction solver (dubbed as \ours), a novel ODE solver that mitigates truncation errors by incorporating multiple parallel gradient evaluations in each ODE step. Importantly, since the additional gradient computations are independent, they can be fully parallelized, preserving low-latency sampling. Our method optimizes a small set of learnable parameters in a distillation fashion, ensuring minimal training overhead. In addition, our method can serve as a plugin to improve existing ODE samplers. Extensive experiments on various image synthesis benchmarks demonstrate the effectiveness of our \ours~in achieving high-quality and low-latency sampling. For example, at the same latency level of 5 NFE, EPD achieves an FID of 4.47 on CIFAR-10, 7.97 on FFHQ, 8.17 on ImageNet, and 8.26 on LSUN Bedroom, surpassing existing learning-based solvers by a significant margin. Codes are available in https://github.com/BeierZhu/EPD.

Despite their strong performance in multimodal emotion reasoning, existing Multimodal Large Language Models (MLLMs) often overlook the scenarios involving emotion conflicts, where emotional cues from different modalities are inconsistent. To fill this gap, we first introduce CA-MER, a new benchmark designed to examine MLLMs under realistic emotion conflicts. It consists of three subsets: video-aligned, audio-aligned, and consistent, where only one or all modalities reflect the true emotion. However, evaluations on our CA-MER reveal that current state-of-the-art emotion MLLMs systematically over-rely on audio signal during emotion conflicts, neglecting critical cues from visual modality. To mitigate this bias, we propose MoSEAR, a parameter-efficient framework that promotes balanced modality integration. MoSEAR consists of two modules: (1)MoSE, modality-specific experts with a regularized gating mechanism that reduces modality bias in the fine-tuning heads; and (2)AR, an attention reallocation mechanism that rebalances modality contributions in frozen backbones during inference. Our framework offers two key advantages: it mitigates emotion conflicts and improves performance on consistent samples-without incurring a trade-off between audio and visual modalities. Experiments on multiple benchmarks-including MER2023, EMER, DFEW, and our CA-MER-demonstrate that MoSEAR achieves state-of-the-art performance, particularly under modality conflict conditions.

Diffusion-based generative processes, formulated as differential equation solving, frequently balance computational speed with sample quality. Our theoretical investigation of ODE- and SDE-based solvers reveals complementary weaknesses: ODE solvers accumulate irreducible gradient error along deterministic trajectories, while SDE methods suffer from amplified discretization errors when the step budget is limited. Building upon this insight, we introduce AdaSDE, a novel single-step SDE solver that aims to unify the efficiency of ODEs with the error resilience of SDEs. Specifically, we introduce a single per-step learnable coefficient, estimated via lightweight distillation, which dynamically regulates the error correction strength to accelerate diffusion sampling. Notably, our framework can be integrated with existing solvers to enhance their capabilities. Extensive experiments demonstrate state-of-the-art performance: at 5 NFE, AdaSDE achieves FID scores of $4.18$ on CIFAR-10, $8.05$ on FFHQ and $6.96$ on LSUN Bedroom. Codes are available https://github.com/WLU-wry02/AdaSDE.

Vision-language models (VLMs) such as CLIP demonstrate strong generalization in zero-shot classification but remain highly vulnerable to adversarial perturbations. Existing methods primarily focus on adversarial fine-tuning or prompt optimization, they often overlook the gaps in CLIP’s encoded features, which is shown as the text and image features lie far apart from each other. This misalignment is significantly amplified under adversarial perturbations, leading to severe degradation in classification performance. To address this problem, we propose **C**r**O**ss-moda**L**ity **A**lignment, dubbed **COLA**, an optimal transport-based framework that explicitly addresses adversarial misalignment by restoring both global image-text alignment and local structural consistency in the feature space. (1) COLA first projects adversarial image embeddings onto a subspace spanned by class text features, effectively filtering out non-semantic distortions while preserving discriminative information. (2) It then models images and texts as discrete distributions over multiple augmented views and refines their alignment via OT, with the subspace projection seamlessly integrated into the cost computation. This design ensures stable cross-modal alignment even under adversarial conditions. COLA is training-free and compatible with existing fine-tuned models. Extensive evaluations across 14 zero-shot classification benchmarks demonstrate the effectiveness of COLA, especially with an average improvement of 6.7% on ImageNet and its variants under PGD adversarial attacks, while maintaining high accuracy on clean samples.

In this paper, we present an empirical study on image recognition unfairness, i.e., extreme class accuracy disparity on balanced data like ImageNet. We demonstrate that classes are not equal and unfairness is prevalent for image classification models across various datasets, network architectures, and model capacities. Moreover, several intriguing properties of fairness are identified. First, the unfairness lies in problematic representation rather than classifier bias distinguished from long-tailed recognition. Second, with the proposed concept of Model Prediction Bias, we investigate the origins of problematic representation during training optimization. Our findings reveal that models tend to exhibit greater prediction biases for classes that are more challenging to recognize. It means that more other classes will be confused with harder classes. Then the False Positives (FPs) will dominate the learning in optimization, thus leading to their poor accuracy. Further, we conclude that data augmentation and representation learning algorithms improve overall performance by promoting fairness to some degree in image classification.

Vision-language models such as CLIP are capable of mapping the different modality data into a unified feature space, enabling zero/few-shot inference by measuring the similarity of given images and texts. However, most existing methods overlook modality gaps in CLIP's encoded features, which is shown as the text and image features lie far apart from each other, resulting in limited classification performance. To tackle this issue, we introduce a method called Selective Vision-Language Subspace Projection (SSP), which incorporates local image features and utilizes them as a bridge to enhance the alignment between image-text pairs. Specifically, our SSP framework comprises two parallel modules: a vision projector and a language projector. Both projectors utilize local image features to span the respective subspaces for image and texts, thereby projecting the image and text features into their respective subspaces to achieve alignment. Moreover, our approach entails only training-free matrix calculations and can be seamlessly integrated into advanced CLIP-based few-shot learning frameworks. Extensive experiments on 11 datasets have demonstrated SSP's superior text-image alignment capabilities, outperforming the state-of-the-art alignment methods. The code is available at https://github.com/zhuhsingyuu/SSP

When fine-tuning zero-shot models like CLIP, our desideratum is for the fine-tuned model to excel in both in-distribution (ID) and out-of-distribution (OOD). Recently, ensemble-based models (ESM) have been shown to offer significant robustness improvement, while preserving high ID accuracy. However, our study finds that ESMs do not solve the ID-OOD trade-offs: they achieve peak performance for ID and OOD accuracy at different mixing coefficients. When optimized for OOD accuracy, the ensemble model exhibits a noticeable decline in ID accuracy, and vice versa. In contrast, we propose a sample-wise ensembling technique that can simultaneously attain the best ID and OOD accuracy without the trade-offs. Specifically, we construct a Zero-Shot Failure (ZSF) set containing training samples incorrectly predicted by the zero-shot model. For each test sample, we calculate its distance to the ZSF set and assign a higher weight to the fine-tuned model in the ensemble if the distance is small. We term our method Variance Reduction Fine-tuning (VRF), as it effectively reduces the variance in ensemble predictions, thereby decreasing residual error. On ImageNet and five derived distribution shifts, our VRF further improves the OOD accuracy by 1.5 - 2.0 pp over the ensemble baselines while maintaining or increasing ID accuracy. VRF achieves similar large robustness gains on (0.9 - 3.1 pp) on other distribution shifts 19 benchmarks. Codes are available in https://github.com/BeierZhu/VRF.

Vision-language models, such as CLIP, have shown impressive generalization capacities when using appropriate text descriptions. While optimizing prompts on downstream labeled data has proven effective in improving performance, these methods entail labor costs for annotations and are limited by their quality. Additionally, since CLIP is pre-trained on highly imbalanced Web-scale data, it suffers from inherent label bias that leads to suboptimal performance. To tackle the above challenges, we propose a label-**F**ree p**ro**mpt distribution **l**earning and b**i**as **c**orrection framework, dubbed as **Frolic**, which boosts zero-shot performance without the need for labeled data. Specifically, our Frolic learns distributions over prompt prototypes to capture diverse visual representations and adaptively fuses these with the original CLIP through confidence matching. This fused model is further enhanced by correcting label bias via a label-free logit adjustment. Notably, our method is not only training-free but also circumvents the necessity for hyper-parameter tuning. Extensive experimental results across 16 datasets demonstrate the efficacy of our approach, particularly outperforming the state-of-the-art by an average of $2.6\%$ on 10 datasets with CLIP ViT-B/16 and achieving an average margin of $1.5\%$ on ImageNet and its five distribution shifts with CLIP ViT-B/16. Codes are available in [https://github.com/zhuhsingyuu/Frolic](https://github.com/zhuhsingyuu/Frolic).

Audio-visual speech recognition (AVSR) has gained remarkable success for ameliorating the noise-robustness of speech recognition. Mainstream methods focus on fusing audio and visual inputs to obtain modality-invariant representations. However, such representations are prone to over-reliance on audio modality as it is much easier to recognize than video modality in clean conditions. As a result, the AVSR model underestimates the importance of visual stream in face of noise corruption. To this end, we leverage visual modality-specific representations to provide stable complementary information for the AVSR task. Specifically, we propose a reinforcement learning (RL) based framework called MSRL, where the agent dynamically harmonizes modality-invariant and modality-specific representations in the auto-regressive decoding process. We customize a reward function directly related to task-specific metrics (i.e., word error rate), which encourages the MSRL to effectively explore the optimal integration strategy. Experimental results on the LRS3 dataset show that the proposed method achieves state-of-the-art in both clean and various noisy conditions. Furthermore, we demonstrate the better generality of MSRL system than other baselines when test set contains unseen noises.

We present a new paradigm for fine-tuning large-scale vision-language pre-trained models on downstream task, dubbed Prompt Regularization (ProReg). Different from traditional fine-tuning which easily overfits to the downstream task data, ProReg uses the prediction by prompting the pretrained model to regularize the fine-tuning. The motivation is: by prompting the large model “a photo of a [CLASS]”, the fill-in answer is only dependent on the pretraining encyclopedic knowledge while independent of the task data distribution, which is usually biased. Specifically, given a training sample prediction during fine-tuning, we first calculate its Kullback-Leibler loss of the prompt prediction and Cross-Entropy loss of the ground-truth label, and then combine them with a proposed sample-wise adaptive trade- off weight, which automatically adjusts the transfer between the pretrained and downstream domains. On various out-of-distribution benchmarks, we show the consistently strong performance of ProReg compared with conventional fine-tuning, zero-shot prompt, prompt tuning, and other state-of-the-art methods.

Thanks to the large pre-trained vision-language models (VLMs) like CLIP [37], we can craft a zero-shot classifier by discrete prompt design, e.g., the confidence score of an image being "[CLASS]" can be obtained by using the VLM provided similarity between the image and the prompt sentence "a photo of a [CLASS]". Furthermore, prompting shows great potential for fast adaptation of VLMs to downstream tasks if we fine-tune the soft prompts with few samples. However, we find a common failure that improper fine-tuning or learning with extremely few-shot samples may even under-perform the zero-shot prediction. Existing methods still address this problem by using traditional anti-overfitting techniques such as early stopping and data augmentation, which lack a principled solution specific to prompting. In this paper, we present Prompt-aligned Gradient, dubbed ProGrad to prevent prompt tuning from forgetting the general knowledge learned from VLMs. In particular, ProGrad only updates the prompt whose gradient is aligned (or non-conflicting) to the general knowledge, which is represented as the optimization direction offered by the pre-defined prompt predictions. Extensive experiments under the few-shot learning, domain generalization, base-to-new generalization and cross-dataset transfer settings demonstrate the stronger few-shot generalization ability of ProGrad over state-of-the-art prompt tuning methods.

Foundation models like CLIP allow zero-shot transfer on various tasks without additional training data. Yet, the zero-shot performance is less competitive than a fully supervised one. Thus, to enhance the performance, fine-tuning and ensembling are also commonly adopted to better fit the downstream tasks. However, we argue that such prior work has overlooked the inherent biases in foundation models. Due to the highly imbalanced Web-scale training set, these foundation models are inevitably skewed toward frequent semantics, and thus the subsequent fine-tuning or ensembling is still biased. In this study, we systematically examine the biases in foundation models and demonstrate the efficacy of our proposed Generalized Logit Adjustment (GLA) method. Note that bias estimation in foundation models is challenging, as most pre-train data cannot be explicitly assessed like in traditional long-tailed classification tasks. To this end, GLA has an optimization-based bias estimation approach for debiasing foundation models. As our work resolves a fundamental flaw in the pre-training, the proposed GLA demonstrates significant improvements across a diverse range of tasks: it achieves 1.5 pp accuracy gains on ImageNet, an large average improvement (1.4-4.6 pp) on 11 few-shot datasets, 2.4 pp gains on long-tailed classification. Codes are in https://github.com/BeierZhu/GLA.

We address the overlooked unbiasedness in existing long-tailed classification methods: we find that their overall improvement is mostly attributed to the biased preference of "tail" over "head", as the test distribution is assumed to be balanced; however, when the test is as imbalanced as the long-tailed training data---let the test respect Zipf's law of nature---the "tail" bias is no longer beneficial overall because it hurts the "head" majorities. In this paper, we propose Cross-Domain Empirical Risk Minimization (xERM) for training an unbiased test-agnostic model to achieve strong performances on both test distributions, which empirically demonstrates that xERM fundamentally improves the classification by learning better feature representation rather than the "head vs. tail" game. Based on causality, we further theoretically explain why xERM achieves unbiasedness: the bias caused by the domain selection is removed by adjusting the empirical risks on the imbalanced domain and the balanced but unseen domain.

In this paper, we propose a structure-coherent deep feature learning method for face alignment. Unlike most existing face alignment methods which overlook the facial structure cues, we explicitly exploit the relation among facial landmarks to make the detector robust to hard cases such as occlusion and large pose. Specifically, we leverage a landmark-graph relational network to enforce the structural relationships among landmarks. We consider the facial landmarks as structural graph nodes and carefully design the neighborhood to passing features among the most related nodes. Our method dynamically adapts the weights of node neighborhood to eliminate distracted information from noisy nodes, such as occluded landmark point. Moreover, different from most previous works which only tend to penalize the landmarks absolute position during the training, we propose a relative location loss to enhance the information of relative location of landmarks. This relative location supervision further regularizes the facial structure. Our approach considers the interactions among facial landmarks and can be easily implemented on top of any convolutional backbone to boost the performance. Extensive experiments on three popular benchmarks, including WFLW, COFW and 300W, demonstrate the effectiveness of the proposed method. In particular, due to explicit structure modeling, our approach is especially robust to challenging cases resulting in impressive low failure rate on COFW and WFLW datasets. The model and code are publicly available at https://github.com/BeierZhu/Sturcture-Coherency-Face-Alignment.