ICML 2025 Papers

Contrastive Language-Image Pretraining (CLIP) has achieved remarkable success, but its performance can degrade when fine-tuned in out-of-distribution (OOD) scenarios. We model the prediction process using a Structural Causal Model (SCM) and show that the causal mechanism involving both invariant and variant factors in training environments differs from that in test environments. In contrast, the causal mechanism with solely invariant factors remains consistent across environments. We theoretically prove the existence of a linear mapping from CLIP embeddings to invariant factors, which can be estimated using interventional data. Additionally, we provide a condition to guarantee low OOD risk of the invariant predictor. Based on these insights, we propose the Invariant Causal Mechanism of CLIP (CLIP-ICM) framework. CLIP-ICM involves collecting interventional data, estimating a linear projection matrix, and making predictions within the invariant subspace. Experiments on several OOD datasets show that CLIP-ICM significantly improves the performance of CLIP. Our method offers a simple but powerful enhancement, boosting the reliability of CLIP in real-world applications.

Chinese scene text retrieval is a practical task that aims to search for images containing visual instances of a Chinese query text. This task is extremely challenging because Chinese text often features complex and diverse layouts in real-world scenes. Current efforts tend to inherit the solution for English scene text retrieval, failing to achieve satisfactory performance. In this paper, we establish a Diversified Layout benchmark for Chinese Street View Text Retrieval (DL-CSVTR), which is specifically designed to evaluate retrieval performance across various text layouts, including vertical, cross-line, and partial alignments. To address the limitations in existing methods, we propose Chinese Scene Text Retrieval CLIP (CSTR-CLIP), a novel model that integrates global visual information with multi-granularity alignment training. CSTR-CLIP applies a two-stage training process to overcome previous limitations, such as the exclusion of visual features outside the text region and reliance on single-granularity alignment, thereby enabling the model to effectively handle diverse text layouts. Experiments on existing benchmark show that CSTR-CLIP outperforms the previous state-of-the-art model by 18.82% accuracy and also provides faster inference speed. Further analysis on DL-CSVTR confirms the superior performance of CSTR-CLIP in handling various text layouts. The dataset and code will be publicly available to facilitate research in Chinese scene text retrieval.

Recently, transformer-based models have demonstrated state-of-the-art performance across various computer vision tasks, including image classification, detection, and segmentation. However, their substantial parameter count poses significant challenges for deployment in resource-constrained environments such as edge or mobile devices. Low-rank approximation (LRA) has emerged as a promising model compression technique, effectively reducing the number of parameters in transformer models by decomposing high-dimensional weight matrices into low-rank representations. Nevertheless, matrix decomposition inherently introduces information loss, often leading to a decline in model accuracy. Furthermore, existing studies on LRA largely overlook the quantization process, which is a critical step in deploying practical vision transformer (ViT) models. To address these challenges, we propose a robust LRA framework that preserves weight information after matrix decomposition and incorporates quantization tailored to LRA characteristics. First, we introduce a reparameterizable branch-based low-rank approximation (RB-LRA) method coupled with weight reconstruction to minimize information loss during matrix decomposition. Subsequently, we enhance model accuracy by integrating RB-LRA with knowledge distillation techniques. Lastly, we present an LRA-aware quantization method designed to mitigate the large outliers generated by LRA, thereby improving the robustness of the quantized model. To validate the effectiveness of our approach, we conducted extensive experiments on the ImageNet dataset using various ViT-based models. Notably, the Swin-B model with RB-LRA achieved a 31.8\% reduction in parameters and a 30.4\% reduction in GFLOPs, with only a 0.03\% drop in accuracy. Furthermore, incorporating the proposed LRA-aware quantization method reduced accuracy loss by an additional 0.83\% compared to naive quantization.

The introduction of the Segment Anything Model (SAM) has paved the way for numerous semantic segmentation applications. For several tasks, quantifying the uncertainty of SAM is of particular interest. However, the ambiguous nature of the class-agnostic foundation model SAM challenges current uncertainty quantification (UQ) approaches. This paper presents a theoretically motivated uncertainty quantification model based on a Bayesian entropy formulation jointly respecting aleatoric, epistemic, and the newly introduced task uncertainty. We use this formulation to train USAM, a lightweight post-hoc UQ method. Our model traces the root of uncertainty back to under-parameterised models, insufficient prompts or image ambiguities. Our proposed deterministic USAM demonstrates superior predictive capabilities on the SA-V, MOSE, ADE20k, DAVIS, and COCO datasets, offering a computationally cheap and easy-to-use UQ alternative that can support user-prompting, enhance semi-supervised pipelines, or balance the tradeoff between accuracy and cost efficiency.

Direct Preference Optimization (DPO) aligns text-to-image (T2I) generation models with human preferences using pairwise preference data. Although substantial resources are expended in collecting and labeling datasets, a critical aspect is often neglected: *preferences vary across individuals and should be represented with more granularity.* To address this, we propose SmPO-Diffusion, a novel method for modeling preference distributions to improve the DPO objective, along with a numerical upper bound estimation for the diffusion optimization objective. First, we introduce a smoothed preference distribution to replace the original binary distribution. We employ a reward model to simulate human preferences and apply preference likelihood averaging to improve the DPO loss, such that the loss function approaches zero when preferences are similar. Furthermore, we utilize an inversion technique to simulate the trajectory preference distribution of the diffusion model, enabling more accurate alignment with the optimization objective. Our approach effectively mitigates issues of excessive optimization and objective misalignment present in existing methods through straightforward modifications. Experimental results demonstrate that our method achieves state-of-the-art performance in preference evaluation tasks, surpassing baselines across various metrics, while reducing the training costs.

Progress in 3D vision-language learning has been hindered by the scarcity of large-scale 3D datasets. We introduce UniVLG, a unified architecture for 2D and 3D vision-language understanding that bridges the gap between existing 2D-centric models and the rich 3D sensory data available in embodied systems. Our approach initializes most model weights from pre-trained 2D models and trains on both 2D and 3D vision-language data. We propose a novel language-conditioned mask decoder shared across 2D and 3D modalities to ground objects effectively in both RGB and RGB-D images, outperforming box-based approaches. To further reduce the domain gap between 2D and 3D, we incorporate 2D-to-3D lifting strategies, enabling UniVLG to utilize 2D data to enhance 3D performance. With these innovations, our model achieves state-of-the-art performance across multiple 3D vision-language grounding tasks, demonstrating the potential of transferring advances from 2D vision-language learning to the data-constrained 3D domain. Furthermore, co-training on both 2D and 3D data enhances performance across modalities without sacrificing 2D capabilities. By removing the reliance on 3D mesh reconstruction and ground-truth object proposals, UniVLG sets a new standard for realistic, embodied-aligned evaluation. Code and additional visualizations are available at https://univlg.github.io.

We present LOCATE 3D, a model for localizing objects in 3D scenes from referring expressions like "the small coffee table between the sofa and the lamp." LOCATE 3D sets a new state-of-the-art on standard referential grounding benchmarks and showcases robust generalization capabilities. Notably, LOCATE 3D operates directly on sensor observation streams (posed RGB-D frames), enabling real-world deployment on robots and AR devices. Key to our approach is 3D-JEPA, a novel self-supervised learning (SSL) algorithm applicable to sensor point clouds. It takes as input a 3D pointcloud featurized using 2D foundation models (CLIP, DINO). Subsequently, masked prediction in latent space is employed as a pretext task to aid the self-supervised learning of contextualized pointcloud features. Once trained, the 3D-JEPA encoder is finetuned alongside a language-conditioned decoder to jointly predict 3D masks and bounding boxes. Additionally, we introduce LOCATE 3D DATASET, a new dataset for 3D referential grounding, spanning multiple capture setups with over 130K annotations. This enables a systematic study of generalization capabilities as well as a stronger model. Code, models and dataset can be found at the project website: locate3d.atmeta.com

LiDAR-based 3D panoptic segmentation often struggles with the inherent sparsity of data from LiDAR sensors, which makes it challenging to accurately recognize distant or small objects. Recently, a few studies have sought to overcome this challenge by integrating LiDAR inputs with camera images, leveraging the rich and dense texture information provided by the latter. While these approaches have shown promising results, they still face challenges, such as misalignment during data augmentation and the reliance on post-processing steps. To address these issues, we propose **I**mage-**A**ssists-**L**iDAR (**IAL**), a novel multi-modal 3D panoptic segmentation framework. In IAL, we first introduce a modality-synchronized data augmentation strategy, PieAug, to ensure alignment between LiDAR and image inputs from the start. Next, we adopt a transformer decoder to directly predict panoptic segmentation results. To effectively fuse LiDAR and image features into tokens for the decoder, we design a Geometric-guided Token Fusion (GTF) module. Additionally, we leverage the complementary strengths of each modality as priors for query initialization through a Prior-based Query Generation (PQG) module, enhancing the decoder’s ability to generate accurate instance masks. Our IAL framework achieves state-of-the-art performance compared to previous multi-modal 3D panoptic segmentation methods on two widely used benchmarks. Code and models are publicly available at https://github.com/IMPL-Lab/IAL.git.

In this work, we explore a cost-effective framework for multilingual image generation. We find that, unlike models tuned on high-quality images with multilingual annotations, leveraging text encoders pre-trained on widely available, noisy Internet image-text pairs significantly enhances data efficiency in text-to-image (T2I) generation across multiple languages. Based on this insight, we introduce MuLan, Multi-Language adapter, a lightweight language adapter with fewer than 20M parameters, trained alongside a frozen text encoder and image diffusion model. Compared to previous multilingual T2I models, this framework offers: (1) Cost efficiency. Using readily accessible English data and off-the-shelf multilingual text encoders minimizes the training cost; (2) High performance. Achieving comparable generation capabilities in over 110 languages with CLIP similarity scores nearly matching those in English (39.57 for English vs. 39.61 for other languages); and (3) Broad applicability. Seamlessly integrating with compatible community tools like LoRA, LCM, ControlNet, and IP-Adapter, expanding its potential use cases.

The success of large language models (LLMs) in text generation has inspired their application to image generation. However, existing methods either rely on specialized designs with inductive biases or adopt LLMs without fully exploring their potential in vision tasks. In this work, we systematically investigate the design space of LLMs for image generation and demonstrate that LLMs can achieve near state-of-the-art performance without domain-specific designs, simply by making proper choices in tokenization methods, modeling approaches, scan patterns, vocabulary design, and sampling strategies. We further analyze autoregressive models' learning and scaling behavior, revealing how larger models effectively capture more useful information than the smaller ones. Additionally, we explore the inherent differences between text and image modalities, highlighting the potential of LLMs across domains. The exploration provides valuable insights to inspire more effective designs when applying LLMs to other domains. With extensive experiments, our proposed model, **ELM** achieves an FID of 1.54 on 256$\times$256 ImageNet and an FID of 3.29 on 512$\times$512 ImageNet, demonstrating the powerful generative potential of LLMs in vision tasks.

Incorporating geometric invariance into neural networks enhances parameter efficiency but typically increases computational costs. This paper introduces new equivariant neural networks that preserve symmetry while maintaining a comparable number of floating-point operations (FLOPs) per parameter to standard non-equivariant networks. We focus on horizontal mirroring (flopping) invariance, common in many computer vision tasks. The main idea is to parametrize the feature spaces in terms of mirror-symmetric and mirror-antisymmetric features, i.e., irreps of the flopping group. This decomposes the linear layers to be block-diagonal, requiring half the number of FLOPs. Our approach reduces both FLOPs and wall-clock time, providing a practical solution for efficient, scalable symmetry-aware architectures.

We introduce FlexTok, a tokenizer that projects 2D images into variable-length, ordered 1D token sequences. For example, a 256x256 image can be resampled into anywhere from 1 to 256 discrete tokens, hierarchically and semantically compressing its information. By training a rectified flow model as the decoder and using nested dropout, FlexTok produces plausible reconstructions regardless of the chosen token sequence length. We evaluate our approach in an autoregressive generation setting using a simple GPT-style Transformer. On ImageNet, this approach achieves an FID<2 across 8 to 128 tokens, outperforming TiTok and matching state-of-the-art methods with far fewer tokens. We further extend the model to support to text-conditioned image generation and examine how FlexTok relates to traditional 2D tokenization. A key finding is that FlexTok enables next-token prediction to describe images in a coarse-to-fine "visual vocabulary", and that the number of tokens to generate depends on the complexity of the generation task.

Creating Computer-Aided Design (CAD) models requires significant expertise and effort. Text-to-CAD, which converts textual descriptions into CAD parametric sequences, is crucial in streamlining this process. Recent studies have utilized ground-truth parametric sequences, known as sequential signals, as supervision to achieve this goal. However, CAD models are inherently multimodal, comprising parametric sequences and corresponding rendered visual objects. Besides, the rendering process from parametric sequences to visual objects is many-to-one. Therefore, both sequential and visual signals are critical for effective training. In this work, we introduce CADFusion, a framework that uses Large Language Models (LLMs) as the backbone and alternates between two training stages: the sequential learning (SL) stage and the visual feedback (VF) stage. In the SL stage, we train LLMs using ground-truth parametric sequences, enabling the generation of logically coherent parametric sequences. In the VF stage, we reward parametric sequences that render into visually preferred objects and penalize those that do not, allowing LLMs to learn how rendered visual objects are perceived and evaluated. These two stages alternate throughout the training, ensuring balanced learning and preserving benefits of both signals. Experiments demonstrate that CADFusion significantly improves performance, both qualitatively and quantitatively.

The advancements in Multimodal Large Language Models (MLLMs) have enabled various multimodal tasks to be addressed under a zero-shot paradigm. This paradigm sidesteps the cost of model fine-tuning, emerging as a dominant trend in practical application. Nevertheless, Multimodal Sentiment Analysis (MSA), a pivotal challenge in the quest for general artificial intelligence, fails to accommodate this convenience. The zero-shot paradigm exhibits undesirable performance on MSA, casting doubt on whether MLLMs can perceive sentiments as competent as supervised models. By extending the zero-shot paradigm to In-Context Learning (ICL) and conducting an in-depth study on configuring demonstrations, we validate that MLLMs indeed possess such capability. Specifically, three key factors that cover demonstrations' retrieval, presentation, and distribution are comprehensively investigated and optimized. A sentimental predictive bias inherent in MLLMs is also discovered and later effectively counteracted. By complementing each other, the devised strategies for three factors result in average accuracy improvements of 15.9% on six MSA datasets against the zero-shot paradigm and 11.2% against the random ICL baseline.

Text-to-visual retrieval often struggles with semantic redundancy and granularity mismatches between textual queries and visual content. Unlike existing methods that address these challenges during training, we propose VISual Abstraction (VISA), a test-time approach that enhances retrieval by transforming visual content into textual descriptions using off-the-shelf large models. The generated text descriptions, with their dense semantics, naturally filter out low-level redundant visual information. To further address granularity issues, VISA incorporates a question-answering process, enhancing the text description with the specific granularity information requested by the user. Extensive experiments demonstrate that VISA brings substantial improvements in text-to-image and text-to-video retrieval for both short- and long-context queries, offering a plug-and-play enhancement to existing retrieval systems.

3D few-shot class incremental learning (FSCIL) aims to learn new point cloud categories from limited samples while preventing the forgetting of previously learned categories. This research area significantly enhances the capabilities of self-driving vehicles and computer vision systems. Existing 3D FSCIL approaches primarily utilize multimodal pre-trained models to extract the semantic features, heavily dependent on meticulously designed high-quality prompts and fine-tuning strategies. To reduce this dependence, this paper proposes a novel method for **3D** **F**SCI**L** with **E**mbedded **G**eometric features (**3D-FLEG**). Specifically, 3D-FLEG develops a point cloud *geometric feature extraction module* to capture category-related geometric characteristics. To address the modality heterogeneity issues that arise from integrating geometric and text features, 3D-FLEG introduces a *geometric feature embedding module*. By augmenting text prompts with spatial geometric features through these modules, 3D-FLEG can learn robust representations of new categories even with limited samples, while mitigating forgetting of the previously learned categories. Experiments conducted on several publicly available 3D point cloud datasets, including ModelNet, ShapeNet, ScanObjectNN, and CO3D, demonstrate 3D-FLEG's superiority over existing state-of-the-art 3D FSCIL methods. Code is available at https://github.com/lixiangqi707/3D-FLEG.

The substantial computational demands of detection transformers (DETRs) hinder their deployment in resource-constrained scenarios, with the encoder consistently emerging as a critical bottleneck. A promising solution lies in reducing token redundancy within the encoder. However, existing methods perform static sparsification while ignoring the varying importance of tokens across different levels and encoder blocks for object detection, leading to suboptimal sparsification and performance degradation. In this paper, we propose **Dynamic DETR** (**Dynamic** token aggregation for **DE**tection **TR**ansformers), a novel strategy that leverages inherent importance distribution to control token density and performs multi-level token sparsification. Within each stage, we apply a proximal aggregation paradigm for low-level tokens to maintain spatial integrity, and a holistic strategy for high-level tokens to capture broader contextual information. Furthermore, we propose center-distance regularization to align the distribution of tokens throughout the sparsification process, thereby facilitating the representation consistency and effectively preserving critical object-specific patterns. Extensive experiments on canonical DETR models demonstrate that Dynamic DETR is broadly applicable across various models and consistently outperforms existing token sparsification methods.

This paper introduces MOTE (MOre Than meets the Eye), a novel multi-object tracking (MOT) algorithm designed to address the challenges of tracking occluded objects. By integrating deformable detection transformers with a custom disocclusion matrix, MOTE significantly enhances the ability to track objects even when they are temporarily hidden from view. The algorithm leverages optical flow to generate features that are processed through a softmax splatting layer, which aids in the creation of a disocclusion matrix. This matrix plays a crucial role in maintaining track consistency by estimating the motion of occluded objects. MOTE's architecture includes modifications to the enhanced track embedding module (ETEM), which allows it to incorporate these advanced features into the track query layer embeddings. This integration ensures that the model not only tracks visible objects but also accurately predicts the trajectories of occluded ones, much like the human visual system. The proposed method is evaluated on multiple datasets, including MOT17, MOT20, and DanceTrack, where it achieves impressive tracking metrics--82.0 MOTA and 66.3 HOTA on the MOT17 dataset, 81.7 MOTA and 65.8 HOTA on the MOT20 dataset, and 93.2 MOTA and 74.2 HOTA on the DanceTrack dataset. Notably, MOTE excels in reducing identity switches and maintaining consistent tracking in complex real-world scenarios with frequent occlusions, outperforming existing state-of-the-art methods across all tested benchmarks.

Recent diffusion-based talking face generation models have demonstrated impressive potential in synthesizing videos that accurately match a speech audio clip with a given reference identity. However, existing approaches still encounter significant challenges due to uncontrollable factors, such as inaccurate lip-sync, inappropriate head posture and the lack of fine-grained control over facial expressions. In order to introduce more face-guided conditions beyond speech audio clips, a novel two-stage training framework Playmate is proposed to generate more lifelike facial expressions and talking faces. In the first stage, we introduce a decoupled implicit 3D representation along with a meticulously designed motion-decoupled module to facilitate more accurate attribute disentanglement and generate expressive talking videos directly from audio cues. Then, in the second stage, we introduce an emotion-control module to encode emotion control information into the latent space, enabling fine-grained control over emotions and thereby achieving the ability to generate talking videos with desired emotion. Extensive experiments demonstrate that Playmate not only outperforms existing state-of-the-art methods in terms of video quality, but also exhibits strong competitiveness in lip synchronization while offering improved flexibility in controlling emotion and head pose. The code will be available at https://github.com/Playmate111/Playmate.