ICLR 2025 Papers

Graph Transformers (GTs), adept at capturing the locality and globality of graphs, have shown promising potential in node classification tasks. Most state-of-the-art GTs succeed through integrating local Graph Neural Networks (GNNs) with their global Self-Attention (SA) modules to enhance structural awareness. Nonetheless, this architecture faces limitations arising from scalability challenges and the trade-off between capturing local and global information. On the one hand, the quadratic complexity associated with the SA modules poses a significant challenge for many GTs, particularly when scaling them to large-scale graphs. Numerous GTs necessitated a compromise, relinquishing certain aspects of their expressivity to garner computational efficiency. On the other hand, GTs face challenges in maintaining detailed local structural information while capturing long-range dependencies. As a result, they typically require significant computational costs to balance the local and global expressivity. To address these limitations, this paper introduces a novel GT architecture, dubbed DUALFormer, featuring a dual-dimensional design of its GNN and SA modules. Leveraging approximation theory from Linearized Transformers and treating the query as the surrogate representation of node features, DUALFormer \emph{efficiently} performs the computationally intensive global SA module on feature dimensions. Furthermore, by such a separation of local and global modules into dual dimensions, DUALFormer achieves a natural balance between local and global expressivity. In theory, DUALFormer can reduce intra-class variance, thereby enhancing the discriminability of node representations. Extensive experiments on eleven real-world datasets demonstrate its effectiveness and efficiency over existing state-of-the-art GTs.

Video anomaly detection (VAD) aims to identify novel actions or events which are unseen during training. Existing mainstream VAD techniques typically focus on the global patterns with redundant details and struggle to generalize to unseen samples. In this paper, we propose a framework that identifies the local patterns which generalize to novel samples and models the dynamics of local patterns. The capability of extracting spatial local patterns is achieved through a two-stage process involving image-text alignment and cross-modality attention. Generalizable representations are built by focusing on semantically relevant components which can be recombined to capture the essence of novel anomalies, reducing unnecessary visual data variances. To enhance local patterns with temporal clues, we propose a State Machine Module (SMM) that utilizes earlier high-resolution textual tokens to guide the generation of precise captions for subsequent low-resolution observations. Furthermore, temporal motion estimation complements spatial local patterns to detect anomalies characterized by novel spatial distributions or distinctive dynamics. Extensive experiments on popular benchmark datasets demonstrate the achievement of state-of-the-art performance. Code is available at https://github.com/AllenYLJiang/Local-Patterns-Generalize-Better/.

Existing neural multi-objective combinatorial optimization (MOCO) methods still exhibit an optimality gap since they fail to fully exploit the intrinsic features of problem instances. A significant factor contributing to this shortfall is their reliance solely on graph-modal information. To overcome this, we propose a novel graph-image multimodal fusion (GIMF) framework that enhances neural MOCO methods by integrating graph and image information of the problem instances. Our GIMF framework comprises three key components: (1) a constructed coordinate image to better represent the spatial structure of the problem instance, (2) a problem-size adaptive resolution strategy during the image construction process to improve the cross-size generalization of the model, and (3) a multimodal fusion mechanism with modality-specific bottlenecks to efficiently couple graph and image information. We demonstrate the versatility of our GIMF by implementing it with two state-of-the-art neural MOCO backbones. Experimental results on classic MOCO problems show that our GIMF significantly outperforms state-of-the-art neural MOCO methods and exhibits superior generalization capability.

A grand challenge in biology is to discover evolutionary traits---features of organisms common to a group of species with a shared ancestor in the tree of life (also referred to as phylogenetic tree). With the growing availability of image repositories in biology, there is a tremendous opportunity to discover evolutionary traits directly from images in the form of a hierarchy of prototypes. However, current prototype-based methods are mostly designed to operate over a flat structure of classes and face several challenges in discovering hierarchical prototypes, including the issue of learning over-specific prototypes at internal nodes. To overcome these challenges, we introduce the framework of Hierarchy aligned Commonality through Prototypical Networks (HComP-Net). The key novelties in HComP-Net include a novel over-specificity loss to avoid learning over-specific prototypes, a novel discriminative loss to ensure prototypes at an internal node are absent in the contrasting set of species with different ancestry, and a novel masking module to allow for the exclusion of over-specific prototypes at higher levels of the tree without hampering classification performance. We empirically show that HComP-Net learns prototypes that are accurate, semantically consistent, and generalizable to unseen species in comparison to baselines. Our code is publicly accessible at Imageomics Institute Github site: https://github.com/Imageomics/HComPNet.

''Grokking'' is a phenomenon where a neural network first memorizes training data and generalizes poorly, but then suddenly transitions to near-perfect generalization after prolonged training. While intriguing, this delayed generalization phenomenon compromises predictability and efficiency. Ideally, models should generalize directly without delay. To this end, this paper proposes GrokTransfer, a simple and principled method for accelerating grokking in training neural networks, based on the key observation that data embedding plays a crucial role in determining whether generalization is delayed. GrokTransfer first trains a smaller, weaker model to reach a nontrivial (but far from optimal) test performance. Then, the learned input embedding from this weaker model is extracted and used to initialize the embedding in the target, stronger model. We rigorously prove that, on a synthetic XOR task where delayed generalization always occurs in normal training, GrokTransfer enables the target model to generalize directly without delay. Moreover, we demonstrate that, across empirical studies of different tasks, GrokTransfer effectively reshapes the training dynamics and eliminates delayed generalization, for both fully-connected neural networks and Transformers.

Light decoder-based solvers have gained popularity for solving vehicle routing problems (VRPs) due to their efficiency and ease of integration with reinforcement learning algorithms. However, they often struggle with generalization to larger problem instances or different VRP variants. This paper revisits light decoder-based approaches, analyzing the implications of their reliance on static embeddings and the inherent challenges that arise. Specifically, we demonstrate that in the light decoder paradigm, the encoder is implicitly tasked with capturing information for all potential decision scenarios during solution construction within a single set of embeddings, resulting in high information density. Furthermore, our empirical analysis reveals that the overly simplistic decoder struggles to effectively utilize this dense information, particularly as task complexity increases, which limits generalization to out-of-distribution (OOD) settings. Building on these insights, we show that enhancing the decoder capacity, with a simple addition of identity mapping and a feed-forward layer, can considerably alleviate the generalization issue. Experimentally, our method significantly enhances the OOD generalization of light decoder-based approaches on large-scale instances and complex VRP variants, narrowing the gap with the heavy decoder paradigm. Our code is available at: https://github.com/ziweileonhuang/reld-nco.

We propose a novel formulation for dictionary learning with an overcomplete dictionary, i.e., when the number of atoms is larger than the dimension of the dictionary. The proposed formulation consists of a weighted sum of $\ell_1$ norms of the rows of the sparse coefficient matrix plus the log of the matrix volume of the dictionary matrix. The main contribution of this work is to show that this novel formulation guarantees global identifiability of the overcomplete dictionary, under a mild condition that the sparse coefficient matrix satisfies a strong scattering condition in the hypercube. Furthermore, if every column of the coefficient matrix is sparse and the dictionary guarantees $\ell_1$ recovery, then the coefficient matrix is identifiable as well. This is a major breakthrough for not only dictionary learning but also general matrix factorization models as identifiability is guaranteed even when the latent dimension is higher than the ambient dimension. We also provide a probabilistic analysis and show that if the sparse coefficient matrix is generated from the widely adopted sparse-Gaussian model, then the $m\times k$ overcomplete dictionary is globally identifiable if the sample size is bigger than a constant times $(k^2/m)\log(k^2/m)$ with overwhelming probability. Finally, we propose an algorithm based on alternating minimization to solve the new proposed formulation.

How to accurately and efficiently assess AI-generated images (AIGIs) remains a critical challenge for generative models. Given the high costs and extensive time commitments required for user studies, many researchers have turned towards employing large multi-modal models (LMMs) as AIGI evaluators, the precision and validity of which are still questionable. Furthermore, traditional benchmarks often utilize mostly natural-captured content rather than AIGIs to test the abilities of LMMs, leading to a noticeable gap for AIGIs. Therefore, we introduce **A-Bench** in this paper, a benchmark designed to diagnose *whether LMMs are masters at evaluating AIGIs*. Specifically, **A-Bench** is organized under two key principles: 1) Emphasizing both high-level semantic understanding and low-level visual quality perception to address the intricate demands of AIGIs. 2) Various generative models are utilized for AIGI creation, and various LMMs are employed for evaluation, which ensures a comprehensive validation scope. Ultimately, 2,864 AIGIs from 16 text-to-image models are sampled, each paired with question-answers annotated by human experts. We hope that **A-Bench** will significantly enhance the evaluation process and promote the generation quality for AIGIs.

Large Language Models~(LLMs) have demonstrated superior performance across a wide range of tasks, but they still exhibit undesirable errors due to incorrect knowledge learned from the training data. To avoid this, knowledge editing methods emerged to precisely edit the specific model knowledge via efficiently modifying a very small percentage of parameters. However, those methods can lead to the problem of **Specificity Failure**, where the existing knowledge and capabilities are severely degraded due to editing. Our preliminary indicates that Specificity Failure primarily stems from the model's attention heads assigning excessive attention scores to entities related to the edited knowledge, thereby unduly focusing on specific snippets within the context, which we denote as the **Attention Drift** phenomenon. To mitigate such Attention Drift issue, we introduce a simple yet effective method **S**elective **A**ttention **D**rift **R**estriction(**SADR**), which introduces an additional regularization term during the knowledge editing process to restrict changes in the attention weight distribution, thereby preventing undue focus on the edited entity. Experiments on five frequently-used strong LLMs demonstrate the effectiveness of our method, where SADR can significantly mitigate Specificity Failure in the predominant knowledge editing tasks.

Antibody design is crucial for developing therapies against diseases such as cancer and viral infections. Recent deep generative models have significantly advanced computational antibody design, particularly in enhancing binding affinity to target antigens. However, beyond binding affinity, antibodies should exhibit other favorable biophysical properties such as non-antigen binding specificity and low self-association, which are important for antibody developability and clinical safety. To address this challenge, we propose AbNovo, a framework that leverages constrained preference optimization for multi-objective antibody design. First, we pre-train an antigen-conditioned generative model for antibody structure and sequence co-design. Then, we fine-tune the model using binding affinity as a reward while enforcing explicit constraints on other biophysical properties. Specifically, we model the physical binding energy with continuous rewards rather than pairwise preferences and explore a primal-and-dual approach for constrained optimization. Additionally, we incorporate a structure-aware protein language model to mitigate the issue of limited training data. Evaluated on independent test sets, AbNovo outperforms existing methods in metrics of binding affinity such as Rosetta binding energy and evolutionary plausibility, as well as in metrics for other biophysical properties like stability and specificity.

Transformers achieve state-of-the-art accuracy and robustness across many tasks, but an understanding of their inductive biases and how those biases differ from other neural network architectures remains elusive. In this work, we identify the sensitivity of the model to token-wise random perturbations in the input as a unified metric which explains the inductive bias of transformers across different data modalities and distinguishes them from other architectures. We show that transformers have lower sensitivity than MLPs, CNNs, ConvMixers and LSTMs, across both vision and language tasks. We also show that this low-sensitivity bias has important implications: i) lower sensitivity correlates with improved robustness; it can also be used as an efficient intervention to further improve the robustness of transformers; ii) it corresponds to flatter minima in the loss landscape; and iii) it can serve as a progress measure for grokking. We support these findings with theoretical results showing (weak) spectral bias of transformers in the NTK regime, and improved robustness due to the lower sensitivity.

Low-Rank Adaptation (LoRA) has emerged as a widely adopted technique in text-to-image models, enabling precise rendering of multiple distinct elements, such as characters and styles, in multi-concept image generation. However, current approaches face significant challenges when composing these LoRAs for multi-concept image generation, particularly as the number of LoRAs increases, resulting in diminished generated image quality. In this paper, we initially investigate the role of LoRAs in the denoising process through the lens of the Fourier frequency domain. Based on the hypothesis that applying multiple LoRAs could lead to "semantic conflicts", we have conducted empirical experiments and find that certain LoRAs amplify high-frequency features such as edges and textures, whereas others mainly focus on low-frequency elements, including the overall structure and smooth color gradients. Building on these insights, we devise a frequency domain based sequencing strategy to determine the optimal order in which LoRAs should be integrated during inference. This strategy offers a methodical and generalizable solution compared to the naive integration commonly found in existing LoRA fusion techniques. To fully leverage our proposed LoRA order sequence determination method in multi-LoRA composition tasks, we introduce a novel, training-free framework, Cached Multi-LoRA (CMLoRA), designed to efficiently integrate multiple LoRAs while maintaining cohesive image generation. With its flexible backbone for multi-LoRA fusion and a non-uniform caching strategy tailored to individual LoRAs, CMLoRA has the potential to reduce semantic conflicts in LoRA composition and improve computational efficiency. Our experimental evaluations demonstrate that CMLoRA outperforms state-of-the-art training-free LoRA fusion methods by a significant margin -- it achieves an average improvement of $2.19$% in CLIPScore, and $11.25%$% in MLLM win rate compared to LoraHub, LoRA Composite, and LoRA Switch.

As the quality of large language models has improved, there has been increased interest in using them to model non-linguistic tokens. For example, the Decision Transformer recasts agentic decision making as a sequence modeling problem, using a decoder-only LLM to model the distribution over the discrete action space for an Atari agent. However, when adapting LLMs to non-linguistic domains, it remains unclear if softmax over discrete bins captures the continuous structure of the tokens and the potentially complex distributions needed for high quality token generation. We introduce a neural network layer, constructed using Fourier series, which we can easily substitute for any linear layer if we want the outputs to have a more continuous structure. We perform extensive analysis on synthetic datasets, as well as on large-scale decision making and time series forecasting tasks. We also provide theoretical evidence that this layer can better learn signal from data while ignoring high-frequency noise. All of our results support the effectiveness of our proposed Fourier head in scenarios where the underlying data distribution has a natural continuous structure. For example, the Fourier head improves a Decision Transformer agent's returns across four benchmark Atari games by as much as 377\%, and increases a state-of-the-art times series foundation model's forecasting performance by 3.5\% across 20 benchmarks unseen during training. We release our implementation at https://nategillman.com/fourier-head

Data selection can reduce the amount of training data needed to finetune LLMs; however, the efficacy of data selection scales directly with its compute. Motivated by the practical challenge of compute-constrained finetuning, we consider the setting in which both the cost of selecting data and training are budgeted for. We first formalize the problem of data selection with a cost-aware utility function, and model the data selection problem as trading off initial-selection cost for training gain. We run a comprehensive sweep of experiments across multiple tasks, varying compute budget by scaling finetuning tokens, model sizes, and data selection compute. Interestingly we find that many powerful data selection methods are almost never compute-optimal, and that cheaper data selection alternatives dominate both from a theoretical and empirical perspective. For compute-optimal training, we find that perplexity and gradient data selection require training-to-selection model size ratios of 5x and 10x, respectively.

Recent advances in diffusion models have shown remarkable potential in the conditional generation of novel molecules. These models can be guided in two ways: (i) explicitly, through additional features representing the condition, or (ii) implicitly, using a property predictor. However, training property predictors or conditional diffusion models requires an abundance of labeled data and is inherently challenging in real-world applications. We propose a novel approach that attenuates the limitations of acquiring large labeled datasets by leveraging domain knowledge from quantum chemistry as a non-differentiable oracle to guide an unconditional diffusion model. Instead of relying on neural networks, the oracle provides accurate guidance in the form of estimated gradients, allowing the diffusion process to sample from a conditional distribution specified by quantum chemistry. We show that this results in more precise conditional generation of novel and stable molecular structures. Our experiments demonstrate that our method: (1) significantly reduces atomic forces, enhancing the validity of generated molecules when used for stability optimization; (2) is compatible with both explicit and implicit guidance in diffusion models, enabling joint optimization of molecular properties and stability; and (3) generalizes effectively to molecular optimization tasks beyond stability optimization. Our implementation is available at https://github.com/A-Chicharito-S/ChemGuide.

Point processes model the distribution of random point sets in mathematical spaces, such as spatial and temporal domains, with applications in fields like seismology, neuroscience, and economics. Existing statistical and machine learning models for point processes are predominantly constrained by their reliance on the characteristic intensity function, introducing an inherent trade-off between efficiency and flexibility. In this paper, we introduce Point Set Diffusion, a diffusion-based latent variable model that can represent arbitrary point processes on general metric spaces without relying on the intensity function. By directly learning to stochastically interpolate between noise and data point sets, our approach effectively captures the distribution of point processes and enables efficient, parallel sampling and flexible generation for complex conditional tasks. Experiments on synthetic and real-world datasets demonstrate that Point Set Diffusion achieves state-of-the-art performance in unconditional and conditional generation of spatial and spatiotemporal point processes while providing up to orders of magnitude faster sampling.

Deep learning has proven to be effective in a wide variety of loss minimization problems. However, many applications of interest, like minimizing projected Bellman error and min-max optimization, cannot be modelled as minimizing a scalar loss function but instead correspond to solving a variational inequality (VI) problem. This difference in setting has caused many practical challenges as naive gradient-based approaches from supervised learning tend to diverge and cycle in the VI case. In this work, we propose a principled surrogate-based approach compatible with deep learning to solve VIs. We show that our surrogate-based approach has three main benefits: (1) under assumptions that are realistic in practice (when hidden monotone structure is present, interpolation, and sufficient optimization of the surrogates), it guarantees convergence, (2) it provides a unifying perspective of existing methods, and (3) is amenable to existing deep learning optimizers like ADAM. Experimentally, we demonstrate our surrogate-based approach is effective in min-max optimization and minimizing projected Bellman error. Furthermore, in the deep reinforcement learning case, we propose a novel variant of TD(0) which is more compute and sample efficient.

Face recognition datasets are often collected by crawling Internet and without individuals' consents, raising ethical and privacy concerns. Generating synthetic datasets for training face recognition models has emerged as a promising alternative. However, the generation of synthetic datasets remains challenging as it entails adequate inter-class and intra-class variations. While advances in generative models have made it easier to increase intra-class variations in face datasets (such as pose, illumination, etc.), generating sufficient inter-class variation is still a difficult task. In this paper, we formulate the dataset generation as a packing problem on the embedding space (represented on a hypersphere) of a face recognition model and propose a new synthetic dataset generation approach, called HyperFace. We formalize our packing problem as an optimization problem and solve it with a gradient descent-based approach. Then, we use a conditional face generator model to synthesize face images from the optimized embeddings. We use our generated datasets to train face recognition models and evaluate the trained models on several benchmarking real datasets. Our experimental results show that models trained with HyperFace achieve state-of-the-art performance in training face recognition using synthetic datasets. Project page: https://www.idiap.ch/paper/hyperface

The tokenization of audio with neural audio codec models is a vital part of modern AI pipelines for the generation or understanding of speech, alone or in a multimodal context. Traditionally such tokenization models have concentrated on low parameter-count architectures using only components with strong inductive biases. In this work we show that by applying a transformer architecture with large parameter count to this problem, and applying a flexible Finite Scalar Quantization (FSQ) based bottleneck, it is possible to reach state-of-the-art speech quality at extremely low bit-rates of $400$ or $700$ bits-per-second. The trained models strongly out-perform existing baselines in both objective and subjective tests.