论文检索

Computer Vision · Vision Models & Multimodal

Constantin Venhoff, Ashkan Khakzar, Sonia Joseph, Philip Torr, Neel Nanda

Training vision language models (VLMs) aims to align visual representations from a vision encoder with the textual representations of a pretrained large language model (LLM). However, many VLMs exhibit reduced factual recall performance compared to their LLM backbones, raising the question of how effective multimodal fine-tuning is at extending existing mechanisms within the LLM to visual inputs. We argue that factual recall based on visual inputs requires VLMs to solve a two-hop problem: (1) forming entity representations from visual inputs, and (2) recalling associated factual knowledge based on these entity representations. By benchmarking 14 VLMs with various architectures (LLaVA, Native, Cross-Attention), sizes (7B-124B parameters), and training setups on factual recall tasks against their original LLM backbone models, we find that 11 of 14 models exhibit factual recall degradation. We select three models exhibiting high- and two models with low performance degradation, and use attribution patching, activation patching, and probing to show that degraded VLMs struggle to use the existing factual recall circuit of their LLM backbone, because they resolve the first hop too late in the computation. In contrast, high-performing VLMs resolve entity representations early enough to reuse the existing factual recall mechanism. Finally, we demonstrate two methods to recover performance: patching entity representations from the LLM backbone into the VLM, and prompting with chain-of-thought reasoning. Our results highlight that the speed of early entity resolution critically determines how effective VLMs are in using preexisting LLM mechanisms. More broadly, our work illustrates how mechanistic analysis can explain and unveil systematic failures in multimodal alignment.

Optimization · Stochastic & Convex Optimization

Alina Ene, Ta Duy Nguyen, Adrian Vladu

In this paper, we study the problem $\min_{x\in R^{d},Nx=v}\sum\_{i=1}^{n}f((Ax-b)_{i})$ for a quasi-self-concordant function $f: R\to R$, where $A,N$ are $n\times d$ and $m\times d$ matrices, $b,v$ are vectors of length $n$ and $m$ with $n\ge d.$ We show an algorithm based on a trust-region method with an oracle that can be implemented using $\widetilde{O}(d^{1/3})$ linear system solves, improving the $\widetilde{O}(n^{1/3})$ oracle by [Adil-Bullins-Sachdeva, NeurIPS 2021]. Our implementation of the oracle relies on solving the overdetermined $\ell\_{\infty}$-regression problem $\min\_{x\in R^{d},Nx=v}\|Ax-b\|\_{\infty}$. We provide an algorithm that finds a $(1+\epsilon)$-approximate solution to this problem using $O((d^{1/3}/\epsilon+1/\epsilon^{2})\log(n/\epsilon))$ linear system solves. This algorithm leverages $\ell\_{\infty}$ Lewis weight overestimates and achieves this iteration complexity via a simple lightweight IRLS approach, inspired by the work of [Ene-Vladu, ICML 2019]. Experimentally, we demonstrate that our algorithm significantly improves the runtime of the standard CVX solver.

General Machine Learning · Transfer, Multitask and Meta-learning

Yuyang Deng, Samory Kpotufe

Theoretical works on supervised transfer learning (STL)---where the learner has access to labeled samples from both source and target distributions---have for the most part focused on statistical aspects of the problem, while efficient optimization has received less attention. We consider the problem of designing an SGD procedure for STL that alternates sampling between source and target data, while maintaining statistical transfer guarantees without prior knowledge of the quality of the source data. A main algorithmic difficulty is in understanding how to design such an adaptive sub-sampling mechanism at each SGD step, to automatically gain from the source when it is informative, or bias towards the target and avoid negative transfer when the source is less informative. We show that, such a mixed-sample SGD procedure is feasible for general prediction tasks with convex losses, rooted in tracking an abstract sequence of constrained convex programs that serve to maintain the desired transfer guarantees. We instantiate these results in the concrete setting of linear regression with square loss, and show that the procedure converges, with $1/\sqrt{T}$ rate, to a solution whose statistical performance on the target is adaptive to the a priori unknown quality of the source. Experiments with synthetic and real datasets support the theory.

Probabilistic Methods · Monte Carlo and Sampling Methods

Kijung Jeon, Michael Muehlebach, Molei Tao

Sampling from constrained statistical distributions is a fundamental task in various fields including Bayesian statistics, computational chemistry, and statistical physics. This article considers the cases where the constrained distribution is described by an unconstrained density, as well as additional equality and/or inequality constraints, which often make the constraint set nonconvex. Existing methods for nonconvex constraint set $\Sigma \subset \mathbb{R}^d$ defined by equality or inequality constraints commonly rely on costly projection steps. Moreover, they cannot handle equality and inequality constraints simultaneously as each method only specialized in one case. In addition, rigorous and quantitative convergence guarantee is often lacking. In this paper, we introduce Overdamped Langevin with LAnding (OLLA), a new framework that can design overdamped Langevin dynamics accommodating both equality and inequality constraints. The proposed dynamics also deterministically corrects trajectories along the normal direction of the constraint surface, thus obviating the need for explicit projections. We show that, under suitable regularity conditions on the target density and $\Sigma$, OLLA converges exponentially fast in $W_2$ distance to the constrained target density $\rho_\Sigma(x) \propto \exp(-f(x))d\sigma_\Sigma$. Lastly, through experiments, we demonstrate the efficiency of OLLA compared to projection-based constrained Langevin algorithms and their slack variable variants, highlighting its favorable computational cost and reasonable empirical mixing.

Zhongjian Wang, Peng Zhang, Jinwei Qi, wang yuan, Sheng Xu, Bang Zhang

Although significant progress has been made in audio-driven talking head generation, text-driven methods remain underexplored. In this work, we present OmniTalker, a unified framework that jointly generates synchronized talking audio-video content from input text while emulating the target identity's speaking and facial movement styles, including speech characteristics, head motion, and facial dynamics. Our framework adopts a dual-branch diffusion transformer (DiT) architecture, with one branch dedicated to audio generation and the other to video synthesis. At the shallow layers, cross-modal fusion modules are introduced to integrate information between the two modalities. In deeper layers, each modality is processed independently, with the generated audio decoded by a vocoder and the video rendered using a GAN-based high-quality visual renderer. Leveraging DiT’s in-context learning capability through a masked-infilling strategy, our model can simultaneously capture both audio and visual styles without requiring explicit style extraction modules. Thanks to the efficiency of the DiT backbone and the optimized visual renderer, OmniTalker achieves real-time inference at 25 FPS. To the best of our knowledge, OmniTalker is the first one-shot framework capable of jointly modeling speech and facial styles in real time. Extensive experiments demonstrate its superiority over existing methods in terms of generation quality, particularly in preserving style consistency and ensuring precise audio-video synchronization, all while maintaining efficient inference.

Computer Vision · Classification and Understanding

Zhiwei Yang, Chen Gao, Mike Zheng Shou

Video anomaly detection (VAD) is a critical yet challenging task due to the complex and diverse nature of real-world scenarios. Previous methods typically rely on domain-specific training data and manual adjustments when applying to new scenarios and unseen anomaly types, suffering from high labor costs and limited generalization. Therefore, we aim to achieve generalist VAD, \ie, automatically handle any scene and any anomaly types without training data or human involvement. In this work, we propose PANDA, an agentic AI engineer based on MLLMs. Specifically, we achieve PANDA by comprehensively devising four key capabilities: (1) self-adaptive scene-aware strategy planning, (2) goal-driven heuristic reasoning, (3) tool-augmented self-reflection, and (4) self-improving chain-of-memory. Concretely, we develop a self-adaptive scene-aware RAG mechanism, enabling PANDA to retrieve anomaly-specific knowledge for anomaly detection strategy planning. Next, we introduce a latent anomaly-guided heuristic prompt strategy to enhance reasoning precision. Furthermore, PANDA employs a progressive reflection mechanism alongside a suite of context-aware tools to iteratively refine decision-making in complex scenarios. Finally, a chain-of-memory mechanism enables PANDA to leverage historical experiences for continual performance improvement. Extensive experiments demonstrate that PANDA achieves state-of-the-art performance in multi-scenario, open-set, and complex scenario settings without training and manual involvement, validating its generalizable and robust anomaly detection capability. Code is released at https://github.com/showlab/PANDA.

Optimization · Discrete Optimization

Shiyue Wang, Ziao Guo, Changhong Lu, Junchi Yan

Langevin Dynamics (LD) and its discrete proposal have been widely applied in the field of Combinatorial Optimization (CO). Both sampling-based and data-driven approaches have benefited significantly from these methods. However, LD's reliance on Gaussian noise limits its ability to escape narrow local optima, requires costly parallel chains, and performs poorly in rugged landscapes or with non-strict constraints. These challenges have impeded the development of more advanced approaches. To address these issues, we introduce Fractional Langevin Dynamics (FLD) for CO, replacing Gaussian noise with $\alpha$-stable L\'evy noise. FLD can escape from local optima more readily via L\'evy flights, and in multiple-peak CO problems with high potential barriers it exhibits a polynomial escape time that outperforms the exponential escape time of LD. Moreover, FLD coincides with LD when $\alpha = 2$, and by tuning $\alpha$ it can be adapted to a wider range of complex scenarios in the CO fields. We provide theoretical proof that our method offers enhanced exploration capabilities and improved convergence. Experimental results on the Maximum Independent Set, Maximum Clique, and Maximum Cut problems demonstrate that incorporating FLD advances both sampling-based and data-driven approaches, achieving state-of-the-art (SOTA) performance in most of the experiments.

Deep Learning · Algorithms

Shihan Dou, Shichun Liu, Yuming Yang, Yicheng Zou, Yunhua Zhou, Shuhao Xing, Chenhao Huang, Qiming Ge, haijun Lv, Demin Song 等

We offer a novel perspective on reward modeling by formulating it as a policy discriminator, which quantifies the difference between two policies to generate a reward signal, guiding the training policy towards a target policy with desired behaviors. Based on this conceptual insight, we propose a scalable pre-training method named POLicy DiscriminAtive LeaRning (POLAR), which trains a reward model (RM) to discern identical policies and discriminate different ones. Unlike traditional reward modeling methods relying on absolute preferences, POLAR captures the relative difference between one policy and an arbitrary target policy, which is a scalable, high-level optimization objective suitable for modeling generic ranking relationships. Leveraging the POLAR pre-training paradigm, we present a series of RMs with parameter scales from 1.8B to 7B. Empirical results show that POLAR substantially outperforms traditional non-pre-trained methods, significantly enhancing RM performance. For instance, POLAR-7B could improve preference accuracy from 54.8% to 81.0% on STEM tasks and from 57.9% to 85.5% on creative writing tasks compared to SOTA baselines. POLAR also shows robust generalization capabilities in RLHF using Reinforcement Fine-tuning (RFT), providing reliable reward signals and markedly enhancing policy performance—improving LLaMa3.1-8B from an average of 47.36% to 56.33% and Qwen2.5-32B from 64.49% to 70.47% on 20 benchmarks. Moreover, scaling experiments reveal a clear power-law relationship between computation and performance, supported by linear correlation coefficients approaching 0.99. The impressive performance, strong generalization, and scaling properties suggest that POLAR is a promising direction for developing general and strong reward models.

Probabilistic Methods · Bayesian Models and Methods

Roberto Gheda, Abele Mălan, Thiago Guzella, Carlo Lancia, Robert Birke, Lydia Chen

Bayesian Network models are a powerful tool to collaboratively optimize production processes in various manufacturing industries. When interacting, collaborating parties must preserve their business secrets by maintaining the confidentiality of their model structures and parameters. While most realistic industry scenarios involve hybrid settings, handling both discrete and continuous data, current state-of-the-art methods for collaborative and confidential inference only support discrete data and have high communication costs. In a centralized setting, Junction Trees enable efficient inference even in hybrid scenarios without discretizing continuous variables, but no extension for collaborative and confidential scenarios exists. To address this research gap, we introduce Hybrid CCJT, the first framework for confidential multiparty inference in hybrid domains with semi-honest, non-colluding adversaries, comprising: (i) a method to construct a strongly-rooted Junction Tree across collaborating parties through a novel construct of interface cliques; and, (ii) a protocol for confidential inference built upon multiparty computation primitives comprising a one-time alignment phase and a belief propagation system for combining the inference results across the Junction Tree cliques. Extensive evaluation on nine datasets shows that Hybrid CCJT improves the predictive accuracy of continuous target variables by 32% on average compared to the state-of-the-art, while reducing communication costs by a median 10.4x under purely discrete scenarios.

Computer Vision · Image and Video Generation

Kiet Nguyen, Chanhyuk Lee, Donggyun Kim, Dong Hoon Lee, Seunghoon Hong

Spatial conditioning in pretrained text-to-image diffusion models has significantly improved fine-grained control over the structure of generated images. However, existing control adapters exhibit limited adaptability and incur high training costs when encountering novel spatial control conditions that differ substantially from the training tasks. To address this limitation, we propose Universal Few-Shot Control (UFC), a versatile few-shot control adapter capable of generalizing to novel spatial conditions. Given a few image-condition pairs of an unseen task and a query condition, UFC leverages the analogy between query and support conditions to construct task-specific control features, instantiated by a matching mechanism and an update on a small set of task-specific parameters. Experiments on six novel spatial control tasks show that UFC, fine-tuned with only 30 annotated examples of novel tasks, achieves fine-grained control consistent with the spatial conditions. Notably, when fine-tuned with 0.1\% of the full training data, UFC achieves competitive performance with the fully supervised baselines in various control tasks. We also show that UFC is applicable agnostically to various diffusion backbones and demonstrate its effectiveness on both UNet and DiT architectures. Code is available at https://github.com/kietngt00/UFC.

Deep Learning · Algorithms

Fei Wang, Li Shen, Liang Ding, Chao Xue, Ye Liu, Changxing Ding

Large Language Models (LLMs) excel at natural language processing tasks, but their massive size leads to high computational and storage demands. Recent works have sought to reduce their model size through layer-wise structured pruning. However, they tend to ignore retaining the capabilities in the pruned part. In this work, we re-examine structured pruning paradigms and uncover several key limitations: 1) notable performance degradation due to direct layer removal, 2) incompetent linear weighted layer aggregation, and 3) the lack of effective post-training recovery mechanisms. To address these limitations, we propose CoMe, including a progressive layer pruning framework with a Concatenation-based Merging technology and a hierarchical distillation post-training process. Specifically, we introduce a channel sensitivity metric that utilizes activation intensity and weight norms for fine-grained channel selection. Subsequently, we employ a concatenation-based layer merging method to fuse the most critical channels in the adjacent layers, enabling a progressive model size reduction. Finally, we propose a hierarchical distillation protocol, which leverages the correspondences between the original and pruned model layers established during pruning, enabling efficient knowledge transfer. Experiments on seven benchmarks show that CoMe achieves state-of-the-art performance; when pruning 30% of LLaMA-2-7b's parameters, the pruned model retains 83% of its original average accuracy.

Applications · Physics

Fabian J. Roth, Dominik K. Klein, Maximilian Kannapinn, Jan Peters, Oliver Weeger

In recent years, nonlinear dynamic system identification using artificial neural networks has garnered attention due to its broad potential applications across science and engineering. However, purely data-driven approaches often struggle with extrapolation and may yield physically implausible forecasts. Furthermore, the learned dynamics can exhibit instabilities, making it difficult to apply such models safely and robustly. This article introduces stable port-Hamiltonian neural networks, a machine learning architecture that incorporates physical biases of energy conservation and dissipation while ensuring global Lyapunov stability of the learned dynamics. Through illustrative and real-world examples, we demonstrate that these strong inductive biases facilitate robust learning of stable dynamics from sparse data, while avoiding instability and surpassing purely data-driven approaches in accuracy and physically meaningful generalization. Furthermore, the model's applicability and potential for data-driven surrogate modeling are showcased on multi-physics simulation data.

Rui Jiao, Hanlin Wu, Wenbing Huang, Yuxuan Song, Yawen Ouyang, Yu Rong, Tingyang Xu, Pengju Wang, Hao Zhou, Wei-Ying Ma 等

Metal-Organic Frameworks (MOFs) have attracted considerable attention due to their unique properties including high surface area and tunable porosity, and promising applications in catalysis, gas storage, and drug delivery. Structure prediction for MOFs is a challenging task, as these frameworks are intrinsically periodic and hierarchically organized, where the entire structure is assembled from building blocks like metal nodes and organic linkers. To address this, we introduce MOF-BFN, a novel generative model for MOF structure prediction based on Bayesian Flow Networks (BFNs). Given the local geometry of building blocks, MOF-BFN jointly predicts the lattice parameters, as well as the positions and orientations of all building blocks within the unit cell. In particular, the positions are modelled in the fractional coordinate system to naturally incorporate the periodicity. Meanwhile, the orientations are modeled as unit quaternions sampled from learned Bingham distributions via the proposed Bingham BFN, enabling effective orientation generation on the 4D unit hypersphere. Experimental results demonstrate that MOF-BFN achieves state-of-the-art performance across multiple tasks, including structure prediction, geometric property evaluation, and de novo generation, offering a promising tool for designing complex MOF materials.

Applications · Neuroscience, Cognitive Science

Alexander Ratzan, Sidharth Goel, Junhao Wen, Christos Davatzikos, Erdem Varol

Spatial location and molecular interactions have long been linked to the connectivity patterns of neural circuits. Yet, at the macroscale of human brain networks, the interplay between spatial position, gene expression, and connectivity remains incompletely understood. Recent efforts to map the human transcriptome and connectome have yielded spatially resolved brain atlases, however modeling the relationship between high-dimensional transcriptomic data and connectivity while accounting for inherent spatial confounds presents a significant challenge. In this paper, we present the first deep learning approaches for predicting whole-brain functional connectivity from gene expression and regional spatial coordinates, including our proposed Spatiomolecular Transformer (SMT). SMT explicitly models biological context by tokenizing genes based on their transcription start site (TSS) order to capture multi-scale genomic organization, and incorporating regional 3D spatial location via a dedicated context [CLS] token within its multi-head self-attention mechanism. We rigorously benchmark context-aware neural networks, including SMT and a single-gene resolution Multilayer-Perceptron (MLP), to established rules-based and bilinear methods. Crucially, to ensure that learned relationships in any model are not mere artifacts of spatial proximity, we introduce novel spatiomolecular null maps, preserving both spatial and transcriptomic autocorrelation. Context-aware neural networks outperform linear methods, significantly exceed our stringent null shuffle models, and generalize across diverse connectomic datasets and parcellation resolutions. Together, these findings demonstrate a strong, predictable link between the spatial distributions of gene expression and functional brain network architecture, and establish a rigorously validated deep learning framework for decoding this relationship. Code to reproduce our results is available at: github.com/neuroinfolab/GeneEx2Conn.

Applications · Neuroscience, Cognitive Science

Eray Erturk, Maryam Shanechi

Real-time decoding of target variables from multiple simultaneously recorded neural time-series modalities, such as discrete spiking activity and continuous field potentials, is important across various neuroscience applications. However, a major challenge for doing so is that different neural modalities can have different timescales (i.e., sampling rates) and different probabilistic distributions, or can even be missing at some time-steps. Existing nonlinear models of multimodal neural activity do not address different timescales or missing samples across modalities. Further, some of these models do not allow for real-time decoding. Here, we develop a learning framework that can enable real-time recursive decoding while nonlinearly aggregating information across multiple modalities with different timescales and distributions and with missing samples. This framework consists of 1) a multiscale encoder that nonlinearly aggregates information after learning within-modality dynamics to handle different timescales and missing samples in real time, 2) a multiscale dynamical backbone that extracts multimodal temporal dynamics and enables real-time recursive decoding, and 3) modality-specific decoders to account for different probabilistic distributions across modalities. In both simulations and three distinct multiscale brain datasets, we show that our model can aggregate information across modalities with different timescales and distributions and missing samples to improve real-time target decoding. Further, our method outperforms various linear and nonlinear multimodal benchmarks in doing so.

Applications · Neuroscience, Cognitive Science

Lan Pan, Hanbo Xie, Robert Wilson

Large Language Models (LLMs) have emerged with many intellectual capacities. While numerous benchmarks assess their intelligence, limited attention has been given to their ability to explore—an essential capacity for discovering new information and adapting to novel environments in both natural and artificial systems. The extent to which LLMs can effectively explore, particularly in open-ended tasks, remains unclear. This study investigates whether LLMs can surpass humans in exploration during an open-ended task, using Little Alchemy 2 as a paradigm, where agents combine elements to discover new ones. Results show most LLMs underperform compared to humans, except for the o1 model, with those traditional LLMs relying primarily on uncertainty-driven strategies, unlike humans who balance uncertainty and empowerment. Results indicate that traditional reasoning-focused LLMs, such as GPT-4o, exhibit a significantly faster and less detailed reasoning process, limiting their exploratory performance. In contrast, the DeepSeek reasoning model demonstrates prolonged, iterative thought processes marked by repetitive analysis of combinations and past trials, reflecting a more thorough and human-like exploration strategy. Representational analysis of the models with Sparse Autoencoders (SAE) revealed that uncertainty and choices are represented at earlier transformer blocks, while empowerment values are processed later, causing LLMs to think too fast and make premature decisions, hindering effective exploration. These findings shed light on the limitations of LLM exploration and suggest directions for improving their adaptability.

Deep Learning · Algorithms

Kai Lion, Liang Zhang, Bingcong Li, Niao He

We show that low-rank adaptation of large-scale models suffers from a low stable rank that is well below the linear algebraic rank of the subspace, degrading fine-tuning performance. To mitigate the underutilization of the allocated subspace, we propose PoLAR, a parameterization inspired by the polar decomposition that factorizes the low-rank update into two direction matrices constrained to Stiefel manifolds and an unconstrained scale matrix. Our theory shows that PoLAR yields an exponentially faster convergence rate on a canonical low-rank adaptation problem. Pairing the parameterization with Riemannian optimization leads to consistent gains on three different benchmarks testing general language understanding, commonsense reasoning, and mathematical problem solving with base model sizes ranging from 350M to 27B.

Deep Learning · Theory

Etienne Boursier, Scott Pesme, Radu-Alexandru Dragomir

We study the dynamics of gradient flow with small weight decay on general training losses $F: \mathbb{R}^d \to \mathbb{R}$. Under mild regularity assumptions and assuming convergence of the unregularised gradient flow, we show that the trajectory with weight decay $\lambda$ exhibits a two-phase behaviour as $\lambda \to 0$. During the initial fast phase, the trajectory follows the unregularised gradient flow and converges to a manifold of critical points of $F$. Then, at time of order $1/\lambda$, the trajectory enters a slow drift phase and follows a Riemannian gradient flow minimising the $\ell_2$-norm of the parameters. This purely optimisation-based phenomenon offers a natural explanation for the \textit{grokking} effect observed in deep learning, where the training loss rapidly reaches zero while the test loss plateaus for an extended period before suddenly improving. We argue that this generalisation jump can be attributed to the slow norm reduction induced by weight decay, as explained by our analysis. We validate this mechanism empirically on several synthetic regression tasks.

Social Aspects · Trustworthy Machine Learning

Xuankun Rong, Wenke Huang, Jian Liang, Jinhe Bi, Xun Xiao, Yiming Li, Bo Du, Mang Ye

Multimodal Large Language Models (MLLMs) are increasingly deployed in fine-tuning-as-a-service (FTaaS) settings, where user-submitted datasets adapt general-purpose models to downstream tasks. This flexibility, however, introduces serious security risks, as malicious fine-tuning can implant backdoors into MLLMs with minimal effort. In this paper, we observe that backdoor triggers systematically disrupt cross-modal processing by causing abnormal attention concentration on non-semantic regions—a phenomenon we term **attention collapse**. Based on this insight, we propose **Believe Your Eyes (BYE)**, a data filtering framework that leverages attention entropy patterns as self-supervised signals to identify and filter backdoor samples. BYE operates via a three-stage pipeline: (1) extracting attention maps using the fine-tuned model, (2) computing entropy scores and profiling sensitive layers via bimodal separation, and (3) performing unsupervised clustering to remove suspicious samples. Unlike prior defenses, BYE equires no clean supervision, auxiliary labels, or model modifications. Extensive experiments across various datasets, models, and diverse trigger types validate BYE's effectiveness: it achieves near-zero attack success rates while maintaining clean-task performance, offering a robust and generalizable solution against backdoor threats in MLLMs.

Deep Learning · Graph Neural Networks

Ruyue Liu, Rong Yin, Xiangzhen Bo, Xiaoshuai Hao, Yong Liu, Jinwen Zhong, Can Ma, Weiping Wang

Large-scale pre-trained models have revolutionized Natural Language Processing (NLP) and Computer Vision (CV), showcasing remarkable cross-domain generalization abilities. However, in graph learning, models are typically trained on individual graph datasets, limiting their capacity to transfer knowledge across different graphs and tasks. This approach also heavily relies on large volumes of annotated data, which presents a significant challenge in resource-constrained settings. Unlike NLP and CV, graph-structured data presents unique challenges due to its inherent heterogeneity, including domain-specific feature spaces and structural diversity across various applications. To address these challenges, we propose a novel structure-aware self-supervised learning method for Text-Attributed Graphs (SSTAG). By leveraging text as a unified representation medium for graph learning, SSTAG bridges the gap between the semantic reasoning of Large Language Models (LLMs) and the structural modeling capabilities of Graph Neural Networks (GNNs). Our approach introduces a dual knowledge distillation framework that co-distills both LLMs and GNNs into structure-aware multilayer perceptrons (MLPs), enhancing the scalability of large-scale TAGs. Additionally, we introduce an in-memory mechanism that stores typical graph representations, aligning them with memory anchors in an in-memory repository to integrate invariant knowledge, thereby improving the model’s generalization ability. Extensive experiments demonstrate that SSTAG outperforms state-of-the-art models on cross-domain transfer learning tasks, achieves exceptional scalability, and reduces inference costs while maintaining competitive performance.