论文检索

Deep Learning · Attention Mechanisms

Peng Jin, Bo Zhu, Li Yuan, Shuicheng YAN

In this work, we upgrade the multi-head attention mechanism, the core of the Transformer model, to reduce computational costs while maintaining or surpassing the previous accuracy level. We show that multi-head attention can be expressed in the summation form. Drawing on the insight that not all attention heads hold equal significance, we propose Mixture-of-Head attention (MoH), a new architecture that treats attention heads as experts in the Mixture-of-Experts (MoE) mechanism. MoH has two significant advantages: First, MoH enables each token to select the appropriate attention heads, enhancing inference efficiency without compromising accuracy or increasing the number of parameters. Second, MoH replaces the standard summation in multi-head attention with a weighted summation, introducing flexibility to the attention mechanism and unlocking extra performance potential. Extensive experiments on ViT, DiT, and LLMs demonstrate that MoH outperforms multi-head attention by using only 50%$\sim$90% of the attention heads. Moreover, we demonstrate that pre-trained multi-head attention models, such as LLaMA3-8B, can be further continue-tuned into our MoH models. Notably, MoH-LLaMA3-8B achieves an average accuracy of 64.0% across 14 benchmarks, outperforming LLaMA3-8B by 2.4% by utilizing only 75% of the attention heads. We believe the proposed MoH is a promising alternative to multi-head attention and provides a strong foundation for developing advanced and efficient attention-based models.

Theory · Online Learning and Bandits

Su Jia, Peter Frazier, Nathan Kallus

Experimentation with interference poses a significant challenge in contemporary online platforms. Prior research on experimentation with interference has concentrated on the final output of a policy. Cumulative performance, while equally important, is less well understood. To address this gap, we introduce the problem of Multi-armed Bandits with Interference (MABI), where the learner assigns an arm to each of $N$ experimental units over $T$ rounds. The reward of each unit in each round depends on the treatments of all units, where the interference between two units decays in their distance. The reward functions are chosen by an adversary and may vary arbitrarily over time and across different units. We first show that the optimal expected regret (against the best fixed-arm policy) is $\tilde O(\sqrt T)$, and can be achieved by a switchback policy. However, the regret (as a random variable) for any switchback policy suffers a high variance, since it does not account for $N$. We propose a policy based on a novel clustered randomization scheme, whose regret (i) is optimal in expectation and (ii) admits a high probability bound that vanishes in $N$.

Deep Learning · Theory

Zixiang Chen, Greg Yang, Qingyue Zhao, Quanquan Gu

Despite deep neural networks' powerful representation learning capabilities, theoretical understanding of how networks can simultaneously achieve meaningful feature learning and global convergence remains elusive. Existing approaches like the neural tangent kernel (NTK) are limited because features stay close to their initialization in this parametrization, leaving open questions about feature properties during substantial evolution. In this paper, we investigate the training dynamics of infinitely wide, $L$-layer neural networks using the tensor program (TP) framework. Specifically, we show that, when trained with stochastic gradient descent (SGD) under the Maximal Update parametrization ($\mu$P) and mild conditions on the activation function, SGD enables these networks to learn linearly independent features that substantially deviate from their initial values. This rich feature space captures relevant data information and ensures that any convergent point of the training process is a global minimum. Our analysis leverages both the interactions among features across layers and the properties of Gaussian random variables, providing new insights into deep representation learning. We further validate our theoretical findings through experiments on real-world datasets.

Optimization

Tao Li, Zhengbao He, Yujun Li, Yasheng Wang, Lifeng Shang, Xiaolin Huang

Fine-tuning large-scale pre-trained models is prohibitively expensive in terms of computation and memory costs. Low-Rank Adaptation (LoRA), a popular Parameter-Efficient Fine-Tuning (PEFT) method, offers an efficient solution by optimizing only low-rank matrices. Despite recent progress in improving LoRA's performance, the relationship between the LoRA optimization space and the full parameter space is often overlooked. A solution that appears flat in the loss landscape of the LoRA space may still exhibit sharp directions in the full parameter space, potentially compromising generalization.We introduce Flat-LoRA, which aims to identify a low-rank adaptation situated in a flat region of the full parameter space.Instead of adopting the well-established sharpness-aware minimization approach, which incurs significant computation and memory overheads, we employ a Bayesian expectation loss objective to preserve training efficiency. Further, we design a refined strategy for generating random perturbations to enhance performance and carefully manage memory overhead using random seeds.Experiments across diverse tasks—including mathematical reasoning, coding abilities, dialogue generation, instruction following, and text-to-image generation—demonstrate that Flat-LoRA improves both in-domain and out-of-domain generalization.Code is available at https://github.com/nblt/Flat-LoRA.

Deep Learning · Other Representation Learning

Yupeng Hou, Jianmo Ni, Zhankui He, Noveen Sachdeva, Wang-Cheng Kang, Ed Chi, Julian McAuley, Derek Cheng

Generative recommendation (GR) is an emerging paradigm where user actions are tokenized into discrete token patterns and autoregressively generated as predictions. However, existing GR models tokenize each action independently, assigning the same fixed tokens to identical actions across all sequences without considering contextual relationships. This lack of context-awareness can lead to suboptimal performance, as the same action may hold different meanings depending on its surrounding context. To address this issue, we propose ActionPiece to explicitly incorporate context when tokenizing action sequences. In ActionPiece, each action is represented as a *set* of item features. Given the action sequence corpora, we construct the vocabulary by merging feature patterns as new tokens, based on their co-occurrence frequency both within individual sets and across adjacent sets. Considering the unordered nature of feature sets, we further introduce set permutation regularization, which produces multiple segmentations of action sequences with the same semantics. Our code is available at: https://github.com/google-deepmind/action_piece.

Theory · Online Learning and Bandits

Sho Takemori, Yuhei Umeda, Aditya Gopalan

This paper studies a pure exploration problem with linear bandit feedback on continuous arm sets, aiming to identify an $\epsilon$-optimal arm with high probability. Previous approaches for continuous arm sets have employed instance-independent methods due to technical challenges such as the infinite dimensionality of the space of probability measures and the non-smoothness of the objective function. This paper proposes a novel, tractable algorithm that addresses these challenges by leveraging a reparametrization of the sampling distribution and projected subgradient descent. However, this approach introduces new challenges related to the projection and reconstruction of the distribution from the reparametrization. We address these by focusing on the connection to the approximate Carath\'eodory problem. Compared to the original optimization problem on the infinite-dimensional space, our method is tractable, requiring only the solution of quadratic and fractional quadratic problems on the arm set. We establish an instance-dependent optimality for our method, and empirical results on synthetic data demonstrate its superiority over existing instance-independent baselines.

Deep Learning · Foundation Models

Chunyu Xie, Bin Wang, Fanjing Kong, Jincheng Li, Dawei Liang, Gengshen Zhang, Dawei Leng, Yuhui Yin

Contrastive Language-Image Pre-training (CLIP) excels in multimodal tasks such as image-text retrieval and zero-shot classification but struggles with fine-grained understanding due to its focus on coarse-grained short captions. To address this, we propose Fine-Grained CLIP (FG-CLIP), which enhances fine-grained understanding through three key innovations. First, we leverage large multimodal models to generate 1.6 billion long caption-image pairs for capturing global-level semantic details. Second, a high-quality dataset is constructed with 12 million images and 40 million region-specific bounding boxes aligned with detailed captions to ensure precise, context-rich representations. Third, 10 million hard fine-grained negative samples are incorporated to improve the model's ability to distinguish subtle semantic differences. We construct a comprehensive dataset, termed FineHARD, by integrating high-quality region-specific annotations with challenging fine-grained negative samples. Corresponding training methods are meticulously designed for these data. Extensive experiments demonstrate that FG-CLIP outperforms the original CLIP and other state-of-the-art methods across various downstream tasks, including fine-grained understanding, open-vocabulary object detection, image-text retrieval, and general multimodal benchmarks. These results highlight FG-CLIP's effectiveness in capturing fine-grained image details and improving overall model performance. The data, code, and models are available at https://github.com/360CVGroup/FG-CLIP.

Theory · Learning Theory

Jingwen Fu, Nanning Zheng

Distributional Compositional Generalization (DCG) refers to the ability to tackle tasks from new distributions by leveraging the knowledge of concepts learned from supporting distributions. In this work, we aim to explore the statistical mechanisms of DCG, which have been largely overlooked in previous studies. By statistically formulating the problem, this paper seeks to address two key research questions: 1) Can a method to one DCG problem be applicable to another? 2) What statistical properties can indicate a learning algorithm's capacity for knowledge composition in DCG tasks? \textbf{To address the first question}, an invariant measure is proposed to provide a dimension where all different methods converge. This measure underscores the critical role of data in enabling improvements without trade-offs. \textbf{As for the second question}, we reveal that by decoupling the impacts of insufficient data and knowledge composition, the ability of the learning algorithm to compose knowledge relies on the compatibility and sensitivity between the learning algorithm and the composition rule. In summary, the statistical analysis of the generalization mechanisms provided in this paper deepens our understanding of compositional generalization, offering a complementary evidence on the importance of data in DCG task.

Applications · Chemistry, Physics, and Earth Sciences

Dong-Hee Shin, Young-Han Son, Hyun Jung Lee, Deok-Joong Lee, Tae-Eui Kam

Molecular discovery has attracted significant attention in scientific fields for its ability to generate novel molecules with desirable properties. Although numerous methods have been developed to tackle this problem, most rely on an online setting that requires repeated online evaluation of candidate molecules using the oracle. However, in real-world molecular discovery, the oracle is often represented by wet-lab experiments, making this online setting impractical due to the significant time and resource demands. To fill this gap, we propose the Molecular Stitching (MolStitch) framework, which utilizes a fixed offline dataset to explore and optimize molecules without the need for repeated oracle evaluations. Specifically, MolStitch leverages existing molecules from the offline dataset to generate novel `stitched molecules' that combine their desirable properties. These stitched molecules are then used as training samples to fine-tune the generative model using preference optimization techniques. Experimental results on various offline multi-objective molecular optimization problems validate the effectiveness of MolStitch. The source code is available online.

Theory · Everything Else

Raphael Meyer, William Swartworth, David Woodruff

We study the computational model where we can access a matrix $\mathbf{A}$ only by computing matrix-vector products $\mathbf{A}\mathrm{x}$ for vectors of the form $\mathrm{x} = \mathrm{x}_1 \otimes \cdots \otimes \mathrm{x}_q$. We prove exponential lower bounds on the number of queries needed to estimate various properties, including the trace and the top eigenvalue of $\mathbf{A}$. Our proofs hold for all adaptive algorithms, modulo a mild conditioning assumption on the algorithm's queries. We further prove that algorithms whose queries come from a small alphabet (e.g., $\mathrm{x}_i \in \\{\pm1\\}^n$) cannot test if $\mathbf{A}$ is identically zero with polynomial complexity, despite the fact that a single query using Gaussian vectors solves the problem with probability 1. In steep contrast to the non-Kronecker case, this shows that sketching $\mathbf{A}$ with different distributions of the same subguassian norm can yield exponentially different query complexities. Our proofs follow from the observation that random vectors with Kronecker structure have exponentially smaller inner products than their non-Kronecker counterparts.

Applications · Time Series

Haotian Si, Changhua Pei, Jianhui LI, Dan Pei, Gaogang Xie

Recent advances in lightweight time series forecasting models suggest the inherent simplicity of time series forecasting tasks. In this paper, we present CMoS, a super-lightweight time series forecasting model. Instead of learning the embedding of the shapes, CMoS directly models the spatial correlations between different time series chunks. Additionally, we introduce a Correlation Mixing technique that enables the model to capture diverse spatial correlations with minimal parameters, and an optional Periodicity Injection technique to ensure faster convergence. Despite utilizing as low as 1% of the lightweight model DLinear's parameters count, experimental results demonstrate that CMoS outperforms existing state-of-the-art models across multiple datasets. Furthermore, the learned weights of CMoS exhibit great interpretability, providing practitioners with valuable insights into temporal structures within specific application scenarios.

Deep Learning · Large Language Models

Zongyu Lin, Yao Tang, Xingcheng Yao, Da Yin, ziniu hu, Yizhou Sun, Kai-Wei Chang

Language agents have become a promising solution to complex interactive tasks. One of the key ingredients to the success of language agents is the reward model on the trajectory of the agentic workflow, which provides valuable guidance during training or inference. However, due to the lack of annotations of intermediate interactions, most existing works use an outcome reward model to optimize policies across entire trajectories. This may lead to sub-optimal policies and hinder the overall performance. To address this, we propose QLASS (Q-guided Language Agent Stepwise Search), to automatically generate annotations by estimating Q-values in a stepwise manner for open language agents. By introducing a reasoning tree and performing process reward modeling, QLASS provides effective intermediate guidance for each step. With the stepwise guidance, we propose a Q-guided generation strategy to enable language agents to better adapt to long-term value, resulting in significant performance improvement during model inference on complex interactive agent tasks. Notably, even with almost half the annotated data, QLASS retains strong performance, demonstrating its efficiency in handling limited supervision. We also empirically demonstrate that QLASS can lead to more effective decision making through qualitative analysis.

Applications · Chemistry, Physics, and Earth Sciences

Ye Liu, Yuntian Chen

Automotive drag coefficient ($C_d$) is pivotal to energy efficiency, fuel consumption, and aerodynamic performance. However, costly computational fluid dynamics (CFD) simulations and wind tunnel tests struggle to meet the rapid-iteration demands of automotive design. We present DragSolver, a Transformer-based framework for rapid and accurate $C_d$ estimation from large-scale, diverse 3D vehicle models.DragSolver tackles four key real-world challenges: (1) multi-scale feature extraction to capture both global shape and fine local geometry; (2) heterogeneous scale normalization to handle meshes with varying sizes and densities;(3) surface-guided gating to suppress internal structures irrelevant to external aerodynamics;and (4) epistemic uncertainty estimation via Monte Carlo dropout for risk-aware design. Extensive evaluations on three industrial-scale datasets (DrivaerNet, DrivaerNet++, and DrivaerML) show that DragSolver outperforms existing approaches in accuracy and generalization, achieving an average reduction of relative $L_2$ error by 58.7% across real-world datasets. Crucially, DragSolver is the first to achieve reliable, real-time $C_d$ inference on production-level automotive geometries.

Deep Learning · Foundation Models

Konstantin Donhauser, Kristina Ulicna, Gemma Moran, Aditya Ravuri, Kian Kenyon-Dean, Cian Eastwood, Jason Hartford

Sparse dictionary learning (DL) has emerged as a powerful approach to extract semantically meaningful concepts from the internals of large language models (LLMs) trained mainly in the text domain. In this work, we explore whether DL can extract meaningful concepts from less human-interpretable scientific data, such as vision foundation models trained on cell microscopy images, where limited prior knowledge exists about which high-level concepts should arise. We propose a novel combination of a sparse DL algorithm, Iterative Codebook Feature Learning (ICFL), with a PCA whitening pre-processing step derived from control data. Using this combined approach, we successfully retrieve biologically meaningful concepts, such as cell types and genetic perturbations. Moreover, we demonstrate how our method reveals subtle morphological changes arising from human-interpretable interventions, offering a promising new direction for scientific discovery via mechanistic interpretability in bioimaging.

Deep Learning · Large Language Models

Wen Wang, Ruibing Hou, Hong Chang, Shiguang Shan, Xilin Chen

Large audio-language models (LALMs), built upon powerful Large Language Models (LLMs), have exhibited remarkable audio comprehension and reasoning capabilities. However, the training of LALMs demands a large corpus of audio-language pairs, which requires substantial costs in both data collection and training resources. In this paper, we propose **MATS**, an audio-language multimodal LLM designed to handle **M**ultiple **A**udio task using solely **T**ext-only **S**upervision. By leveraging pre-trained audio-language alignment models such as CLAP, we develop a text-only training strategy that projects the shared audio-language latent space into LLM latent space, endowing the LLM with audio comprehension capabilities without relying on audio data during training. To further bridge the modality gap between audio and language embeddings within CLAP, we propose the **S**trongly-rel**a**ted **n**oisy **t**ext with **a**udio (**Santa**) mechanism. Santa maps audio embeddings into CLAP language embedding space while preserving essential information from the audio input. Extensive experiments demonstrate that MATS, despite being trained exclusively on text data, achieves competitive performance compared to recent LALMs trained on large-scale audio-language pairs. The code is publicly available in [https://github.com/wangwen-banban/MATS](https://github.com/wangwen-banban/MATS)

Reinforcement Learning · Multi-agent

Laixi Shi, Jingchu Gai, Eric Mazumdar, Yuejie Chi, Adam Wierman

Standard multi-agent reinforcement learning (MARL) algorithms are vulnerable to sim-to-real gaps. To address this, distributionally robust Markov games (RMGs) have been proposed to enhance robustness in MARL by optimizing the worst-case performance when game dynamics shift within a prescribed uncertainty set. RMGs remains under-explored, from reasonable problem formulation to the development of sample-efficient algorithms. Two notorious and open challenges are the formulation of the uncertainty set and whether the corresponding RMGs can overcome the curse of multiagency, where the sample complexity scales exponentially with the number of agents. In this work, we propose a natural class of RMGs inspired by behavioral economics, where each agent's uncertainty set is shaped by both the environment and the integrated behavior of other agents. We first establish the well-posedness of this class of RMGs by proving the existence of game-theoretic solutions such as robust Nash equilibria and coarse correlated equilibria (CCE). Assuming access to a generative model, we then introduce a sample-efficient algorithm for learning the CCE whose sample complexity scales polynomially with all relevant parameters. To the best of our knowledge, this is the first algorithm to break the curse of multiagency for RMGs, regardless of the uncertainty set formulation.

Deep Learning · Large Language Models

Shang Liu, Yu Pan, Guanting Chen, Xiaocheng Li

The canonical setup of learning a reward model (RM) from human preferences with binary feedback discards potentially useful samples (such as "tied" between the two responses) and loses fine-grained information (such as "slightly better'"). This paper proposes a framework for learning RMs under *ordinal feedback*, generalizing the binary feedback to arbitrary granularity. We first identify a marginal unbiasedness condition, which generalizes the existing assumption of the binary feedback. The condition is validated via the sociological concept called "wisdom of the crowd". Under this condition, we develop a natural probability model and prove the benefits of fine-grained feedback in terms of reducing the Rademacher complexity, which may be of independent interest to another problem: the bias-variance trade-off in knowledge distillation. The framework also sheds light on designing guidelines for human annotators. Our numerical experiments validate that: (1) fine-grained feedback leads to better RM learning for both in- and out-of-distribution settings; (2) incorporating a certain proportion of tied samples boosts RM learning.

Probabilistic Methods · Variational Inference

Chuanhui Liu, Xiao Wang

This paper provides a comprehensive analysis of variational inference in latent variable models for survival analysis, emphasizing the distinctive challenges associated with applying variational methods to survival data. We identify a critical weakness in the existing methodology, demonstrating how a poorly designed variational distribution may hinder the objective of survival analysis tasks—modeling time-to-event distributions. We prove that the optimal variational distribution, which perfectly bounds the log-likelihood, may depend on the censoring mechanism. To address this issue, we propose censor-dependent variational inference (CDVI), tailored for latent variable models in survival analysis. More practically, we introduce CD-CVAE, a V-structure Variational Autoencoder (VAE) designed for the scalable implementation of CDVI. Further discussion extends some existing theories and training techniques to survival analysis. Extensive experiments validate our analysis and demonstrate significant improvements in the estimation of individual survival distributions.

Applications · Health / Medicine

Lea Bogensperger, Dominik Narnhofer, Ahmed Allam, Konrad Schindler, Michael Krauthammer

The goal of protein fitness optimization is to discover new protein variants with enhanced fitness for a given use. The vast search space and the sparsely populated fitness landscape, along with the discrete nature of protein sequences, pose significant challenges when trying to determine the gradient towards configurations with higher fitness. We introduce *Variational Latent Generative Protein Optimization* (VLGPO), a variational perspective on fitness optimization. Our method embeds protein sequences in a continuous latent space to enable efficient sampling from the fitness distribution and combines a (learned) flow matching prior over sequence mutations with a fitness predictor to guide optimization towards sequences with high fitness. VLGPO achieves state-of-the-art results on two different protein benchmarks of varying complexity. Moreover, the variational design with explicit prior and likelihood functions offers a flexible plug-and-play framework that can be easily customized to suit various protein design tasks.

Applications · Everything Else

Joan Serrà, Recep Oguz Araz, Dmitry Bogdanov, Yuki Mitsufuji

Detecting musical versions (different renditions of the same piece) is a challenging task with important applications. Because of the ground truth nature, existing approaches match musical versions at the track level (e.g., whole song). However, most applications require to match them at the segment level (e.g., 20s chunks). In addition, existing approaches resort to classification and triplet losses, disregarding more recent losses that could bring meaningful improvements. In this paper, we propose a method to learn from weakly annotated segments, together with a contrastive loss variant that outperforms well-studied alternatives. The former is based on pairwise segment distance reductions, while the latter modifies an existing loss following decoupling, hyper-parameter, and geometric considerations. With these two elements, we do not only achieve state-of-the-art results in the standard track-level evaluation, but we also obtain a breakthrough performance in a segment-level evaluation. We believe that, due to the generality of the challenges addressed here, the proposed methods may find utility in domains beyond audio or musical version matching.