ICML 2025 Papers

Federated Prompt Learning (FPL) adapts pre-trained Vision-Language Models (VLMs) to federated learning through prompt tuning, leveraging their transferable representations and strong generalization capabilities. Traditional methods often require uniform prompt lengths for federated aggregation, limiting adaptability to clients with diverse prompt lengths and distribution biases. In this paper, we propose **Fed**erated **P**rompt Learning for **H**eterogeneous Client **A**daptation (FedPHA), a novel framework that combines a fixed-length global prompt for efficient aggregation with local prompts of varying lengths to capture client-specific data characteristics. Additionally, FedPHA designs Singular Value Decomposition (SVD) based projection and bidirectional alignment to disentangle global conflicts arising from client heterogeneity, ensuring that personalized client tasks effectively utilize non-harmful global knowledge. This approach ensures that global knowledge improves model generalization while local knowledge preserves local optimization. Experimental results validate the effectiveness of FedPHA in achieving a balance between global and personalized knowledge in federated learning scenarios.

The emergence of multimodal large language models (MLLMs) advances multimodal emotion recognition (MER) to the next level—from naive discriminative tasks to complex emotion understanding with advanced video understanding abilities and natural language description. However, the current community suffers from a lack of large-scale datasets with intensive, descriptive emotion annotations, as well as a multimodal-centric framework to maximize the potential of MLLMs for emotion understanding. To address this, we establish a new benchmark for MLLM-based emotion understanding with a novel dataset (MER-Caption) and a new model (AffectGPT). Utilizing our model-based crowd-sourcing data collection strategy, we construct the largest descriptive emotion dataset to date (by far), featuring over 2K fine-grained emotion categories across 115K samples. We also introduce the AffectGPT model, designed with pre-fusion operations to enhance multimodal integration. Finally, we present MER-UniBench, a unified benchmark with evaluation metrics tailored for typical MER tasks and the free-form, natural language output style of MLLMs. Extensive experimental results show AffectGPT's robust performance across various MER tasks. We have released both the code and the dataset to advance research and development in emotion understanding: https://github.com/zeroQiaoba/AffectGPT.

We introduce SWE-Lancer, a benchmark of over 1400 freelance software engineering tasks from Upwork, valued at \\\$1 million USD total in real-world payouts. SWE-Lancer encompasses both independent engineering tasks — ranging from \\\$50 bug fixes to \\\$32000 feature implementations — and managerial tasks, where models choose between technical implementation proposals. Independent tasks are graded with end-to-end tests triple-verified by experienced software engineers, while managerial decisions are assessed against the choices of the original hired engineering managers. We evaluate model performance and find that frontier models are still unable to solve the majority of tasks. To facilitate future research, we open-source a unified Docker image and a public evaluation split. By mapping model performance to monetary value, we hope SWE-Lancer enables greater research into the economic impact of AI model development.

The weighted Euclidean norm $||x||_w$ of a vector $x\in \mathbb{R}^d$ with weights $w\in \mathbb{R}^d$ is the Euclidean norm where the contribution of each dimension is scaled by a given weight. Approaches to dimensionality reduction that satisfy the Johnson–Lindenstrauss (JL) lemma can be easily adapted to the weighted Euclidean distance if weights are known and fixed: it suffices to scale each dimension of the input vectors according to the weights, and then apply any standard approach. However, this is not the case when weights are unknown during the dimensionality reduction or might dynamically change. In this paper, we address this issue by providing a linear function that maps vectors into a smaller complex vector space and allows to retrieve a JL-like estimate for the weighted Euclidean distance once weights are revealed. Our results are based on the decomposition of the complex dimensionality reduction into several Rademacher chaos random variables, which are studied using novel concentration inequalities for sums of independent Rademacher chaoses.

With advancements in visual language models (VLMs) and large language models (LLMs), video anomaly detection (VAD) has progressed beyond binary classification to fine-grained categorization and multidimensional analysis. However, existing methods focus mainly on coarse-grained detection, lacking anomaly explanations. To address these challenges, we propose Ex-VAD, an Explainable Fine-grained Video Anomaly Detection approach that combines fine-grained classification with detailed explanations of anomalies. First, we use a VLM to extract frame-level captions, and an LLM converts them to video-level explanations, enhancing the model's explainability. Second, integrating textual explanations of anomalies with visual information greatly enhances the model's anomaly detection capability. Finally, we apply label-enhanced alignment to optimize feature fusion, enabling precise fine-grained detection. Extensive experimental results on the UCF-Crime and XD-Violence datasets demonstrate that Ex-VAD significantly outperforms existing State-of-The-Art methods.

This paper studies phase transitions for the existence of unregularized M-estimators under proportional asymptotics where the sample size $n$ and feature dimension $p$ grow proportionally with $n/p \to \delta \in (1, \infty)$. We study the existence of M-estimators in single-index models where the response $y_i$ depends on covariates $x_i \sim N(0, I_p)$ through an unknown index ${w} \in \mathbb{R}^p$ and an unknown link function. An explicit expression is derived for the critical threshold $\delta_\infty$ that determines the phase transition for the existence of the M-estimator, generalizing the results of Candés & Sur (2020) for binary logistic regression to other single-index models. Furthermore, we investigate the existence of a solution to the nonlinear system of equations governing the asymptotic behavior of the M-estimator when it exists. The existence of solution to this system for $\delta > \delta_\infty$ remains largely unproven outside the global null in binary logistic regression. We address this gap with a proof that the system admits a solution if and only if $\delta > \delta_\infty$, providing a comprehensive theoretical foundation for proportional asymptotic results that require as a prerequisite the existence of a solution to the system.

3D vision-language grounding faces a fundamental data bottleneck: while 2D models train on billions of images, 3D models have access to only thousands of labeled scenes--a six-order-of-magnitude gap that severely limits performance. We introduce \textbf{\emph{LIFT-GS}}, a practical distillation technique that overcomes this limitation by using differentiable rendering to bridge 3D and 2D supervision. LIFT-GS predicts 3D Gaussian representations from point clouds and uses them to render predicted language-conditioned 3D masks into 2D views, enabling supervision from 2D foundation models (SAM, CLIP, LLaMA) without requiring any 3D annotations. This render-supervised formulation enables end-to-end training of complete encoder-decoder architectures and is inherently model-agnostic. LIFT-GS achieves state-of-the-art results with 25.7\% mAP on open-vocabulary instance segmentation (vs. 20.2\% prior SOTA) and consistent 10-30\% improvements on referential grounding tasks. Remarkably, pretraining effectively multiplies fine-tuning datasets by 2×, demonstrating strong scaling properties that suggest 3D VLG currently operates in a severely data-scarce regime. Project page: \url{https://liftgs.github.io}.

The advancement of Large Vision-Language Models (LVLMs) has increasingly highlighted the critical issue of their tendency to hallucinate non-existing objects in the images. To address this issue, previous works focused on using specially curated datasets or powerful LLMs to rectify the outputs of LVLMs. However, these approaches require either costly training or fine-tuning, or API access to proprietary LLMs for post-generation correction. In response to these limitations, we propose Mitigating hallucinAtion via image-gRounded guIdaNcE (MARINE), a framework that is both training-free and API-free. MARINE effectively and efficiently reduces object hallucinations during inference by introducing image-grounded guidance to LVLMs. This is achieved by leveraging open-source vision models to extract object-level information, thereby enhancing the precision of LVLM-generated content. Our framework's flexibility further allows for the integration of multiple vision models, enabling more reliable and robust object-level guidance. Through comprehensive evaluations across 5 popular LVLMs with diverse evaluation metrics and benchmarks, we demonstrate the effectiveness of MARINE, which even outperforms existing fine-tuning-based methods. Remarkably, it reduces hallucinations consistently in GPT-4V-assisted evaluation while maintaining the detailedness of LVLMs' generations. We release our code at https://github.com/Linxi-ZHAO/MARINE.

Achieving robustness in image segmentation models is challenging due to the fine-grained nature of pixel-level classification. These models, which are crucial for many real-time perception applications, particularly struggle when faced with natural corruptions in the wild for autonomous systems. While sensitivity analysis can help us understand how input variables influence model outputs, its application to natural and uncontrollable corruptions in training data is computationally expensive. In this work, we present an adaptive, sensitivity-guided augmentation method to enhance robustness against natural corruptions. Our sensitivity analysis on average runs 10 times faster and requires about 200 times less storage than previous sensitivity analysis, enabling practical, on-the-fly estimation during training for a model-free augmentation policy. With minimal fine-tuning, our sensitivity-guided augmentation method achieves improved robustness on both real-world and synthetic datasets compared to state-of-the-art data augmentation techniques in image segmentation.

Blind face restoration (BFR), which involves converting low-quality (LQ) images into high-quality (HQ) images, remains challenging due to complex and unknown degradations. While previous diffusion-based methods utilize feature extractors from LQ images as guidance, using raw LQ images directly as the starting point for the reverse diffusion process offers a theoretically optimal solution. In this work, we propose Pseudo-Hashing Image-to-image Schrödinger Bridge (P-I2SB), a novel framework inspired by optimal mass transport problems, which enhances the restoration potential of Schrödinger Bridge (SB) by correcting data distributions and effectively learning the optimal transport path between any two data distributions. Notably, we theoretically explore and identify that existing methods are limited by the optimality and reversibility of solutions in SB, leading to suboptimal performance. Our approach involves preprocessing HQ images during training by hashing them into pseudo-samples according to a rule related to LQ images, ensuring structural similarity in distribution. This guarantees optimal and reversible solutions in SB, enabling the inference process to learn effectively and allowing P-I2SB to achieve state-of-the-art results in BFR, with more natural textures and retained inference speed compared to previous methods.

The Platonic Representation Hypothesis claims that recent foundation models are converging to a shared representation space as a function of their downstream task performance, irrespective of the objectives and data modalities used to train these models (Huh et al., 2024). Representational similarity is generally measured for individual datasets and is not necessarily consistent across datasets. Thus, one may wonder whether this convergence of model representations is confounded by the datasets commonly used in machine learning. Here, we propose a systematic way to measure how representational similarity between models varies with the set of stimuli used to construct the representations. We find that the objective function is a crucial factor in determining the consistency of representational similarities across datasets. Specifically, self-supervised vision models learn representations whose relative pairwise similarities generalize better from one dataset to another compared to those of image classification or image-text models. Moreover, the correspondence between representational similarities and the models' task behavior is dataset-dependent, being most strongly pronounced for single-domain datasets. Our work provides a framework for analyzing similarities of model representations across datasets and linking those similarities to differences in task behavior.

Curvature regularization techniques like Sharpness Aware Minimization (SAM) have shown great promise in improving generalization on vision tasks. However, we find that SAM performs poorly in domains like natural language processing (NLP), often degrading performance --- even with twice the compute budget. We investigate the discrepancy across domains and find that in the NLP setting, SAM is dominated by regularization of the logit statistics --- instead of improving the geometry of the function itself. We use this observation to develop an alternative algorithm we call Functional SAM, which regularizes curvature only through modification of the statistics of the overall function implemented by the neural network, and avoids spurious minimization through logit manipulation. Furthermore, we argue that preconditioning the SAM perturbation also prevents spurious minimization, and when combined with Functional SAM, it gives further improvements. Our proposed algorithms show improved performance over AdamW and SAM baselines when trained for an equal number of steps, in both fixed-length and Chinchilla-style training settings, at various model scales (including billion-parameter scale). On the whole, our work highlights the importance of more precise characterizations of sharpness in broadening the applicability of curvature regularization to large language models (LLMs)

Multi-Modal Large Language Models (MLLMs) have exhibited remarkable performance on various vision-language tasks such as Visual Question Answering (VQA). Despite accumulating evidence of privacy concerns associated with task-relevant content, it remains unclear whether MLLMs inadvertently memorize private content that is entirely irrelevant to the training tasks. In this paper, we investigate how randomly generated task-irrelevant private content can become spuriously correlated with downstream objectives due to partial mini-batch training dynamics, thus causing inadvertent memorization. Concretely, we randomly generate task-irrelevant watermarks into VQA fine-tuning images at varying probabilities and propose a novel probing framework to determine whether MLLMs have inadvertently encoded such content. Our experiments reveal that MLLMs exhibit notably different training behaviors in partial mini-batch settings with task-irrelevant watermarks embedded. Furthermore, through layer-wise probing, we demonstrate that MLLMs trigger distinct representational patterns when encountering previously seen task-irrelevant knowledge, even if this knowledge does not influence their output during prompting. Our code is available at https://github.com/illusionhi/ProbingPrivacy.

Recent work has demonstrated that large-scale, multi-animal models are powerful tools for characterizing the relationship between neural activity and behavior. Current large-scale approaches, however, focus exclusively on either predicting neural activity from behavior (encoding) or predicting behavior from neural activity (decoding), limiting their ability to capture the bidirectional relationship between neural activity and behavior. To bridge this gap, we introduce a multimodal, multi-task model that enables simultaneous Neural Encoding and Decoding at Scale (NEDS). Central to our approach is a novel multi-task-masking strategy, which alternates between neural, behavioral, within-modality, and cross-modality masking. We pretrain our method on the International Brain Laboratory (IBL) repeated site dataset, which includes recordings from 83 animals performing the visual decision-making task. In comparison to other large-scale modeling approaches, we demonstrate that NEDS achieves state-of-the-art performance for both encoding and decoding when pretrained on multi-animal data and then fine-tuned on new animals. Surprisingly, NEDS's learned embeddings exhibit emergent properties: even without explicit training, they are highly predictive of the brain regions in each recording. Altogether, our approach is a step towards a foundation model of the brain that enables seamless translation between neural activity and behavior.

Current Large Vision-Language Models (LVLMs) have demonstrated remarkable capabilities in understanding multimodal data, but their potential remains underexplored for deepfake detection due to the misalignment of their knowledge and forensics patterns. To this end, we present a novel framework that unlocks LVLMs' potential capabilities for deepfake detection. Our framework includes a Knowledge-guided Forgery Detector (KFD), a Forgery Prompt Learner (FPL), and a Large Language Model (LLM). The KFD is used to calculate correlations between image features and pristine/deepfake image description embeddings, enabling forgery classification and localization. The outputs of the KFD are subsequently processed by the Forgery Prompt Learner to construct fine-grained forgery prompt embeddings. These embeddings, along with visual and question prompt embeddings, are fed into the LLM to generate textual detection responses. Extensive experiments on multiple benchmarks, including FF++, CDF2, DFD, DFDCP, DFDC, and DF40, demonstrate that our scheme surpasses state-of-the-art methods in generalization performance, while also supporting multi-turn dialogue capabilities.

A core data-centric learning challenge is the identification of training samples that are detrimental to model performance. Influence functions serve as a prominent tool for this task and offer a robust framework for assessing training data influence on model predictions. Despite their widespread use, their high computational cost associated with calculating the inverse of the Hessian matrix pose constraints, particularly when analyzing large-sized deep models. In this paper, we establish a bridge between identifying detrimental training samples via influence functions and outlier gradient detection. This transformation not only presents a straightforward and Hessian-free formulation but also provides insights into the role of the gradient in sample impact. Through systematic empirical evaluations, we first validate the hypothesis of our proposed outlier gradient analysis approach on synthetic datasets. We then demonstrate its effectiveness in detecting mislabeled samples in vision models and selecting data samples for improving performance of natural language processing transformer models. We also extend its use to influential sample identification for fine-tuning Large Language Models.

The ability to organically reason over and with both text and images is a pillar of human intelligence, yet the ability of Multimodal Large Language Models (MLLMs) to perform such multimodal reasoning remains under-explored. Existing benchmarks often emphasize text-dominant reasoning or rely on shallow visual cues, failing to adequately assess integrated visual and textual reasoning. We introduce EMMA (Enhanced MultiModal reAsoning), a benchmark targeting organic multimodal reasoning across mathematics, physics, chemistry, and coding. EMMA tasks demand advanced cross-modal reasoning that cannot be addressed by reasoning independently in each modality, offering an enhanced test suite for MLLMs' reasoning capabilities. Our evaluation of state-of-the-art MLLMs on EMMA reveals significant limitations in handling complex multimodal and multi-step reasoning tasks, even with advanced techniques like Chain-of-Thought prompting and test-time compute scaling underperforming. These findings underscore the need for improved multimodal architectures and training paradigms to close the gap between human and model reasoning in multimodality.

Many video-to-audio (VTA) methods have been proposed for dubbing silent AI-generated videos. An efficient quality assessment method for AI-generated audio-visual content (AGAV) is crucial for ensuring audio-visual quality. Existing audio-visual quality assessment methods struggle with unique distortions in AGAVs, such as unrealistic and inconsistent elements. To address this, we introduce **AGAVQA-3k**, the first large-scale AGAV quality assessment dataset, comprising $3,382$ AGAVs from $16$ VTA methods. AGAVQA-3k includes two subsets: AGAVQA-MOS, which provides multi-dimensional scores for audio quality, content consistency, and overall quality, and AGAVQA-Pair, designed for optimal AGAV pair selection. We further propose **AGAV-Rater**, a LMM-based model that can score AGAVs, as well as audio and music generated from text, across multiple dimensions, and selects the best AGAV generated by VTA methods to present to the user. AGAV-Rater achieves state-of-the-art performance on AGAVQA-3k, Text-to-Audio, and Text-to-Music datasets. Subjective tests also confirm that AGAV-Rater enhances VTA performance and user experience. The dataset and code is available at https://github.com/charlotte9524/AGAV-Rater.

Motivated by the recent success of time-series foundation models for zero-shot forecasting, we present a methodology for _in-context fine-tuning_ of a time-series foundation model. In particular, we design a pretrained foundation model that can be prompted (at inference time) with multiple time-series examples, in order to forecast a target time-series into the future. Our foundation model is specifically trained to utilize examples from multiple related time-series in its context window (in addition to the history of the target time-series) to help it adapt to the specific distribution of the target domain at inference time. We show that such a foundation model that uses in-context examples at inference time can obtain much better performance on popular forecasting benchmarks compared to supervised deep learning methods, statistical models, and other time series foundation models. Interestingly, our in-context fine-tuning approach even matches the performance of a foundation model that is explicitly fine-tuned on the target domain.