Lei Huang

Xi Weng, Lei Huang, Lei Zhao et al.

A desirable objective in self-supervised learning (SSL) is to avoid feature collapse. Whitening loss guarantees collapse avoidance by minimizing the distance between embeddings of positive pairs under the conditioning that the embeddings from different views are whitened. In this paper, we propose a framework with an informative indicator to analyze whitening loss, which provides a clue to demystify several interesting phenomena as well as a pivoting point connecting to other SSL methods. We reveal that batch whitening (BW) based methods do not impose whitening constraints on the embedding, but they only require the embedding to be full-rank. This full-rank constraint is also sufficient to avoid dimensional collapse. Based on our analysis, we propose channel whitening with random group partition (CW-RGP), which exploits the advantages of BW-based methods in preventing collapse and avoids their disadvantages requiring large batch size. Experimental results on ImageNet classification and COCO object detection reveal that the proposed CW-RGP possesses a promising potential for learning good representations. The code is available at https://github.com/winci-ai/CW-RGP.

Lei Huang, Hui Shen, Kuan-Jui Su et al.

Genotype-based cis-expression prediction depends on accurately modeling local regulatory architecture. We present block-sparse Bayesian sparse linear mixed model (bsBSLMM), an extension of Bayesian sparse linear mixed model (BSLMM) that incorporates linkage disequilibrium (LD)-block spike-and-slab sparsity and a transcription start site (TSS)-informed SNP inclusion prior. Across 23,098 genes from GEUVADIS European-ancestry lymphoblastoid cell lines, bsBSLMM retained more predictable genes than BSLMM, LASSO, BLUP, TIGAR elastic net, and TIGAR Dirichlet-process regression under matched evaluation criteria. Compared with BSLMM, bsBSLMM improved held-out prediction performance for most shared genes, with gains driven primarily by LD-block sparsity and further enhanced by the TSS-informed prior. Variants selected by bsBSLMM showed stronger enrichment in GM12878 DNase and H3K27ac regulatory regions than variants selected by BSLMM. In transcriptome-wide association study (TWAS) analysis, bsBSLMM recovered established inflammatory bowel disease signals, including IL23R, and identified additional genome-wide significant genes not detected by BSLMM. Independent validation in the Louisiana Osteoporosis Study reproduced the increased prediction yield across ancestries and recovered biologically relevant bone mineral density pathways in downstream TWAS and gene set enrichment analyses. These results demonstrate that incorporating LD-block structure and biologically informed SNP priors improves cis-expression prediction and enhances downstream TWAS discovery.

Wang Lu, Jindong Wang, Han Yu et al.

Domain generalization (DG) aims to learn a generalizable model from multiple training domains such that it can perform well on unseen target domains. A popular strategy is to augment training data to benefit generalization through methods such as Mixup~\cite{zhang2018mixup}. While the vanilla Mixup can be directly applied, theoretical and empirical investigations uncover several shortcomings that limit its performance. Firstly, Mixup cannot effectively identify the domain and class information that can be used for learning invariant representations. Secondly, Mixup may introduce synthetic noisy data points via random interpolation, which lowers its discrimination capability. Based on the analysis, we propose a simple yet effective enhancement for Mixup-based DG, namely domain-invariant Feature mIXup (FIX). It learns domain-invariant representations for Mixup. To further enhance discrimination, we leverage existing techniques to enlarge margins among classes to further propose the domain-invariant Feature MIXup with Enhanced Discrimination (FIXED) approach. We present theoretical insights about guarantees on its effectiveness. Extensive experiments on seven public datasets across two modalities including image classification (Digits-DG, PACS, Office-Home) and time series (DSADS, PAMAP2, UCI-HAR, and USC-HAD) demonstrate that our approach significantly outperforms nine state-of-the-art related methods, beating the best performing baseline by 6.5\% on average in terms of test accuracy. Code is available at: https://github.com/jindongwang/transferlearning/tree/master/code/deep/fixed.

Lei Huang, Weijiang Yu, Weitao Ma et al.

The emergence of large language models (LLMs) has marked a significant breakthrough in natural language processing (NLP), fueling a paradigm shift in information acquisition. Nevertheless, LLMs are prone to hallucination, generating plausible yet nonfactual content. This phenomenon raises significant concerns over the reliability of LLMs in real-world information retrieval (IR) systems and has attracted intensive research to detect and mitigate such hallucinations. Given the open-ended general-purpose attributes inherent to LLMs, LLM hallucinations present distinct challenges that diverge from prior task-specific models. This divergence highlights the urgency for a nuanced understanding and comprehensive overview of recent advances in LLM hallucinations. In this survey, we begin with an innovative taxonomy of hallucination in the era of LLM and then delve into the factors contributing to hallucinations. Subsequently, we present a thorough overview of hallucination detection methods and benchmarks. Our discussion then transfers to representative methodologies for mitigating LLM hallucinations. Additionally, we delve into the current limitations faced by retrieval-augmented LLMs in combating hallucinations, offering insights for developing more robust IR systems. Finally, we highlight the promising research directions on LLM hallucinations, including hallucination in large vision-language models and understanding of knowledge boundaries in LLM hallucinations.

Jiawei Zhang, Xiang Wang, Xiao Bai et al.

Despite recent stereo matching networks achieving impressive performance given sufficient training data, they suffer from domain shifts and generalize poorly to unseen domains. We argue that maintaining feature consistency between matching pixels is a vital factor for promoting the generalization capability of stereo matching networks, which has not been adequately considered. Here we address this issue by proposing a simple pixel-wise contrastive learning across the viewpoints. The stereo contrastive feature loss function explicitly constrains the consistency between learned features of matching pixel pairs which are observations of the same 3D points. A stereo selective whitening loss is further introduced to better preserve the stereo feature consistency across domains, which decorrelates stereo features from stereo viewpoint-specific style information. Counter-intuitively, the generalization of feature consistency between two viewpoints in the same scene translates to the generalization of stereo matching performance to unseen domains. Our method is generic in nature as it can be easily embedded into existing stereo networks and does not require access to the samples in the target domain. When trained on synthetic data and generalized to four real-world testing sets, our method achieves superior performance over several state-of-the-art networks.

Jiaxi Wang, Ji Wu, Lei Huang

Batch Normalization (BN) is a core and prevalent technique in accelerating the training of deep neural networks and improving the generalization on Computer Vision (CV) tasks. However, it fails to defend its position in Natural Language Processing (NLP), which is dominated by Layer Normalization (LN). In this paper, we are trying to answer why BN usually performs worse than LN in NLP tasks with Transformer models. We find that the inconsistency between training and inference of BN is the leading cause that results in the failure of BN in NLP. We define Training Inference Discrepancy (TID) to quantitatively measure this inconsistency and reveal that TID can indicate BN's performance, supported by extensive experiments, including image classification, neural machine translation, language modeling, sequence labeling, and text classification tasks. We find that BN can obtain much better test performance than LN when TID keeps small through training. To suppress the explosion of TID, we propose Regularized BN (RBN) that adds a simple regularization term to narrow the gap between batch statistics and population statistics of BN. RBN improves the performance of BN consistently and outperforms or is on par with LN on 17 out of 20 settings, involving ten datasets and two common variants of Transformer Our code is available at https://github.com/wjxts/RegularizedBN.

Lei Huang, Hengtong Zhang, Tingyang Xu et al.

Molecule generation, especially generating 3D molecular geometries from scratch (i.e., 3D \textit{de novo} generation), has become a fundamental task in drug designs. Existing diffusion-based 3D molecule generation methods could suffer from unsatisfactory performances, especially when generating large molecules. At the same time, the generated molecules lack enough diversity. This paper proposes a novel diffusion model to address those two challenges. First, interatomic relations are not in molecules' 3D point cloud representations. Thus, it is difficult for existing generative models to capture the potential interatomic forces and abundant local constraints. To tackle this challenge, we propose to augment the potential interatomic forces and further involve dual equivariant encoders to encode interatomic forces of different strengths. Second, existing diffusion-based models essentially shift elements in geometry along the gradient of data density. Such a process lacks enough exploration in the intermediate steps of the Langevin dynamics. To address this issue, we introduce a distributional controlling variable in each diffusion/reverse step to enforce thorough explorations and further improve generation diversity. Extensive experiments on multiple benchmarks demonstrate that the proposed model significantly outperforms existing methods for both unconditional and conditional generation tasks. We also conduct case studies to help understand the physicochemical properties of the generated molecules.

Cheng Qian, Nicholas D. Sidiropoulos, Kejun Huang et al.

Reconstructing a signal from squared linear (rank-one quadratic) measurements is a challenging problem with important applications in optics and imaging, where it is known as phase retrieval. This paper proposes two new phase retrieval algorithms based on non-convex quadratically constrained quadratic programming (QCQP) formulations, and a recently proposed approximation technique dubbed feasible point pursuit (FPP). The first is designed for uniformly distributed bounded measurement errors, such as those arising from high-rate quantization (B-FPP). The second is designed for Gaussian measurement errors, using a least squares criterion (LS-FPP). Their performance is measured against state-of-the-art algorithms and the Cramér-Rao bound (CRB), which is also derived here. Simulations show that LS-FPP outperforms the state-of-art and operates close to the CRB. Compact CRB expressions, properties, and insights are obtained by explicitly computing the CRB in various special cases -- including when the signal of interest admits a sparse parametrization, using harmonic retrieval as an example.

Yu Zhou, Yu Chen, Xiao Zhang et al.

Recently, deep learning-based compressed sensing (CS) has achieved great success in reducing the sampling and computational cost of sensing systems and improving the reconstruction quality. These approaches, however, largely overlook the issue of the computational cost; they rely on complex structures and task-specific operator designs, resulting in extensive storage and high energy consumption in CS imaging systems. In this paper, we propose a lightweight but effective deep neural network based on recurrent learning to achieve a sustainable CS system; it requires a smaller number of parameters but obtains high-quality reconstructions. Specifically, our proposed network consists of an initial reconstruction sub-network and a residual reconstruction sub-network. While the initial reconstruction sub-network has a hierarchical structure to progressively recover the image, reducing the number of parameters, the residual reconstruction sub-network facilitates recurrent residual feature extraction via recurrent learning to perform both feature fusion and deep reconstructions across different scales. In addition, we also demonstrate that, after the initial reconstruction, feature maps with reduced sizes are sufficient to recover the residual information, and thus we achieved a significant reduction in the amount of memory required. Extensive experiments illustrate that our proposed model can achieve a better reconstruction quality than existing state-of-the-art CS algorithms, and it also has a smaller number of network parameters than these algorithms. Our source codes are available at: https://github.com/C66YU/CSRN.

Lei Huang, Yi Zhou, Tian Wang et al.

Batch normalization (BN) is a milestone technique in deep learning. It normalizes the activation using mini-batch statistics during training but the estimated population statistics during inference. This paper focuses on investigating the estimation of population statistics. We define the estimation shift magnitude of BN to quantitatively measure the difference between its estimated population statistics and expected ones. Our primary observation is that the estimation shift can be accumulated due to the stack of BN in a network, which has detriment effects for the test performance. We further find a batch-free normalization (BFN) can block such an accumulation of estimation shift. These observations motivate our design of XBNBlock that replace one BN with BFN in the bottleneck block of residual-style networks. Experiments on the ImageNet and COCO benchmarks show that XBNBlock consistently improves the performance of different architectures, including ResNet and ResNeXt, by a significant margin and seems to be more robust to distribution shift.

Shaowu Chen, Weize Sun, Lei Huang et al.

In recent years, two competitive time series classification models, namely, ROCKET and MINIROCKET, have garnered considerable attention due to their low training cost and high accuracy. However, they rely on a large number of random 1-D convolutional kernels to comprehensively capture features, which is incompatible with resource-constrained devices. Despite the development of heuristic algorithms designed to recognize and prune redundant kernels, the inherent time-consuming nature of evolutionary algorithms hinders efficient evaluation. To efficiently prune models, this paper eliminates feature groups contributing minimally to the classifier, thereby discarding the associated random kernels without direct evaluation. To this end, we incorporate both group-level ($l_{2,1}$-norm) and element-level ($l_2$-norm) regularizations to the classifier, formulating the pruning challenge as a group elastic net classification problem. An ADMM-based algorithm is initially introduced to solve the problem, but it is computationally intensive. Building on the ADMM-based algorithm, we then propose our core algorithm, POCKET, which significantly speeds up the process by dividing the task into two sequential stages. In Stage 1, POCKET utilizes dynamically varying penalties to efficiently achieve group sparsity within the classifier, removing features associated with zero weights and their corresponding kernels. In Stage 2, the remaining kernels and features are used to refit a $l_2$-regularized classifier for enhanced performance. Experimental results on diverse time series datasets show that POCKET prunes up to 60% of kernels without a significant reduction in accuracy and performs 11$\times$ faster than its counterparts. Our code is publicly available at https://github.com/ShaowuChen/POCKET.

Lei Huang, Zheng Yuan, Huihui Yan et al. · tsinghua

Advances in deep generative models shed light on de novo molecule generation with desired properties. However, molecule generation targeted for dual protein targets still faces formidable challenges including protein 3D structure data requisition for model training, auto-regressive sampling, and model generalization for unseen targets. Here, we proposed DiffDTM, a novel conditional structure-free deep generative model based on a diffusion model for dual targets based molecule generation to address the above issues. Specifically, DiffDTM receives protein sequences and molecular graphs as inputs instead of protein and molecular conformations and incorporates an information fusion module to achieve conditional generation in a one-shot manner. We have conducted comprehensive multi-view experiments to demonstrate that DiffDTM can generate drug-like, synthesis-accessible, novel, and high-binding affinity molecules targeting specific dual proteins, outperforming the state-of-the-art (SOTA) models in terms of multiple evaluation metrics. Furthermore, we utilized DiffDTM to generate molecules towards dopamine receptor D2 and 5-hydroxytryptamine receptor 1A as new antipsychotics. The experimental results indicate that DiffDTM can be easily plugged into unseen dual targets to generate bioactive molecules, addressing the issues of requiring insufficient active molecule data for training as well as the need to retrain when encountering new targets.

Zhangyin Feng, Weitao Ma, Weijiang Yu et al.

Large language models (LLMs) exhibit superior performance on various natural language tasks, but they are susceptible to issues stemming from outdated data and domain-specific limitations. In order to address these challenges, researchers have pursued two primary strategies, knowledge editing and retrieval augmentation, to enhance LLMs by incorporating external information from different aspects. Nevertheless, there is still a notable absence of a comprehensive survey. In this paper, we propose a review to discuss the trends in integration of knowledge and large language models, including taxonomy of methods, benchmarks, and applications. In addition, we conduct an in-depth analysis of different methods and point out potential research directions in the future. We hope this survey offers the community quick access and a comprehensive overview of this research area, with the intention of inspiring future research endeavors.

Ge Kan, Jinhu Lü, Tian Wang et al.

Energy-based latent variable models (EBLVMs) are more expressive than conventional energy-based models. However, its potential on visual tasks are limited by its training process based on maximum likelihood estimate that requires sampling from two intractable distributions. In this paper, we propose Bi-level doubly variational learning (BiDVL), which is based on a new bi-level optimization framework and two tractable variational distributions to facilitate learning EBLVMs. Particularly, we lead a decoupled EBLVM consisting of a marginal energy-based distribution and a structural posterior to handle the difficulties when learning deep EBLVMs on images. By choosing a symmetric KL divergence in the lower level of our framework, a compact BiDVL for visual tasks can be obtained. Our model achieves impressive image generation performance over related works. It also demonstrates the significant capacity of testing image reconstruction and out-of-distribution detection.

Lei Huang, Xiaocheng Feng, Weitao Ma et al.

Despite the impressive performance on information-seeking tasks, large language models (LLMs) still struggle with hallucinations. Attributed LLMs, which augment generated text with in-line citations, have shown potential in mitigating hallucinations and improving verifiability. However, current approaches suffer from suboptimal citation quality due to their reliance on in-context learning. Furthermore, the practice of citing only coarse document identifiers makes it challenging for users to perform fine-grained verification. In this work, we introduce FRONT, a training framework designed to teach LLMs to generate Fine-Grained Grounded Citations. By grounding model outputs in fine-grained supporting quotes, these quotes guide the generation of grounded and consistent responses, not only improving citation quality but also facilitating fine-grained verification. Experiments on the ALCE benchmark demonstrate the efficacy of FRONT in generating superior grounded responses and highly supportive citations. With LLaMA-2-7B, the framework significantly outperforms all the baselines, achieving an average of 14.21% improvement in citation quality across all datasets, even surpassing ChatGPT.

Chengyu Zheng, Ning song, Ruoyu Zhang et al.

Image-text retrieval in remote sensing aims to provide flexible information for data analysis and application. In recent years, state-of-the-art methods are dedicated to ``scale decoupling'' and ``semantic decoupling'' strategies to further enhance the capability of representation. However, these previous approaches focus on either the disentangling scale or semantics but ignore merging these two ideas in a union model, which extremely limits the performance of cross-modal retrieval models. To address these issues, we propose a novel Scale-Semantic Joint Decoupling Network (SSJDN) for remote sensing image-text retrieval. Specifically, we design the Bidirectional Scale Decoupling (BSD) module, which exploits Salience Feature Extraction (SFE) and Salience-Guided Suppression (SGS) units to adaptively extract potential features and suppress cumbersome features at other scales in a bidirectional pattern to yield different scale clues. Besides, we design the Label-supervised Semantic Decoupling (LSD) module by leveraging the category semantic labels as prior knowledge to supervise images and texts probing significant semantic-related information. Finally, we design a Semantic-guided Triple Loss (STL), which adaptively generates a constant to adjust the loss function to improve the probability of matching the same semantic image and text and shorten the convergence time of the retrieval model. Our proposed SSJDN outperforms state-of-the-art approaches in numerical experiments conducted on four benchmark remote sensing datasets.

Qiming Li, Zekai Ye, Xiaocheng Feng et al.

Although Large Vision-Language Models (LVLMs) have demonstrated remarkable performance on downstream tasks, they frequently produce contents that deviate from visual information, leading to object hallucination. To tackle this, recent works mostly depend on expensive manual annotations and training cost, or decoding strategies which significantly increase inference time. In this work, we observe that LVLMs' attention to visual information is significantly enhanced when answering caption queries compared to non-caption queries. Inspired by this phenomenon, we propose Caption-guided Visual Attention Steering (CAST), a training-free, plug-and-play hallucination mitigation method that leverages the attention activation pattern corresponding to caption queries to enhance LVLMs' visual perception capability. Specifically, we use probing techniques to identify attention heads that are highly sensitive to caption queries and estimate optimized steering directions for their outputs. This steering strengthens LVLM's fine-grained visual perception capabilities, thereby effectively mitigating object hallucination. CAST reduced object hallucination by an average of 6.03% across five widely used LVLMs and five benchmarks including both discriminative and generative tasks, demonstrating state-of-the-art performance while adding little inference cost and preserving other foundational capabilities.

Shaowu Chen, Weize Sun, Lei Huang

Filter pruning has attracted increasing attention in recent years for its capacity in compressing and accelerating convolutional neural networks. Various data-independent criteria, including norm-based and relationship-based ones, were proposed to prune the most unimportant filters. However, these state-of-the-art criteria fail to fully consider the dissimilarity of filters, and thus might lead to performance degradation. In this paper, we first analyze the limitation of relationship-based criteria with examples, and then introduce a new data-independent criterion, Weighted Hybrid Criterion (WHC), to tackle the problems of both norm-based and relationship-based criteria. By taking the magnitude of each filter and the linear dependence between filters into consideration, WHC can robustly recognize the most redundant filters, which can be safely pruned without introducing severe performance degradation to networks. Extensive pruning experiments in a simple one-shot manner demonstrate the effectiveness of the proposed WHC. In particular, WHC can prune ResNet-50 on ImageNet with more than 42% of floating point operations reduced without any performance loss in top-5 accuracy.

Qin Liao, Jieyu Liu, Anqi Huang et al.

We propose a high-rate scheme for discretely-modulated continuous-variable quantum key distribution (DM CVQKD) using quantum machine learning technologies, which divides the whole CVQKD system into three parts, i.e., the initialization part that is used for training and estimating quantum classifier, the prediction part that is used for generating highly correlated raw keys, and the data-postprocessing part that generates the final secret key string shared by Alice and Bob. To this end, a low-complexity quantum k-nearest neighbor (QkNN) classifier is designed for predicting the lossy discretely-modulated coherent states (DMCSs) at Bob's side. The performance of the proposed QkNN-based CVQKD especially in terms of machine learning metrics and complexity is analyzed, and its theoretical security is proved by using semi-definite program (SDP) method. Numerical simulation shows that the secret key rate of our proposed scheme is explicitly superior to the existing DM CVQKD protocols, and it can be further enhanced with the increase of modulation variance.

Shania Mitra, Lei Huang, Manolis Kellis

Protein function prediction is a crucial task in bioinformatics, with significant implications for understanding biological processes and disease mechanisms. While the relationship between sequence and function has been extensively explored, translating protein structure to function continues to present substantial challenges. Various models, particularly, CNN and graph-based deep learning approaches that integrate structural and functional data, have been proposed to address these challenges. However, these methods often fall short in elucidating the functional significance of key residues essential for protein functionality, as they predominantly adopt a retrospective perspective, leading to suboptimal performance. Inspired by region proposal networks in computer vision, we introduce the Protein Region Proposal Network (ProteinRPN) for accurate protein function prediction. Specifically, the region proposal module component of ProteinRPN identifies potential functional regions (anchors) which are refined through the hierarchy-aware node drop pooling layer favoring nodes with defined secondary structures and spatial proximity. The representations of the predicted functional nodes are enriched using attention mechanisms and subsequently fed into a Graph Multiset Transformer, which is trained with supervised contrastive (SupCon) and InfoNCE losses on perturbed protein structures. Our model demonstrates significant improvements in predicting Gene Ontology (GO) terms, effectively localizing functional residues within protein structures. The proposed framework provides a robust, scalable solution for protein function annotation, advancing the understanding of protein structure-function relationships in computational biology.

Lei Huang, Jiaming Guo, Guanhua He et al.

Generating long-term texts such as novels using artificial intelligence has always been a challenge. A common approach is to use large language models (LLMs) to construct a hierarchical framework that first plans and then writes. Despite the fact that the generated novels reach a sufficient length, they exhibit poor logical coherence and appeal in their plots and deficiencies in character and event depiction, ultimately compromising the overall narrative quality. In this paper, we propose a method named Extracting Excelsior and Expanding. Ex3 initially extracts structure information from raw novel data. By combining this structure information with the novel data, an instruction-following dataset is meticulously crafted. This dataset is then utilized to fine-tune the LLM, aiming for excelsior generation performance. In the final stage, a tree-like expansion method is deployed to facilitate the generation of arbitrarily long novels. Evaluation against previous methods showcases Ex3's ability to produce higher-quality long-form novels.

Siwei Wen, Zhangcheng Wang, Xingjian Zhang et al.

Online video understanding requires models to perform continuous perception and long-range reasoning within potentially infinite visual streams. Its fundamental challenge lies in the conflict between the unbounded nature of streaming media input and the limited context window of Multimodal Large Language Models (MLLMs). Current methods primarily rely on passive processing, which often face a trade-off between maintaining long-range context and capturing the fine-grained details necessary for complex tasks. To address this, we introduce EventMemAgent, an active online video agent framework based on a hierarchical memory module. Our framework employs a dual-layer strategy for online videos: short-term memory detects event boundaries and utilizes event-granular reservoir sampling to process streaming video frames within a fixed-length buffer dynamically; long-term memory structuredly archives past observations on an event-by-event basis. Furthermore, we integrate a multi-granular perception toolkit for active, iterative evidence capture and employ Agentic Reinforcement Learning (Agentic RL) to end-to-end internalize reasoning and tool-use strategies into the agent's intrinsic capabilities. Experiments show that EventMemAgent achieves competitive results on online video benchmarks. The code will be released here: https://github.com/lingcco/EventMemAgent.

Weitao Ma, Xiaocheng Feng, Lei Huang et al.

Effective memory management is essential for large language model agents to navigate long-horizon tasks. Recent research has explored using Reinforcement Learning to develop specialized memory manager agents. However, existing approaches rely on final task performance as the primary reward, which results in severe reward sparsity and ineffective credit assignment, providing insufficient guidance for individual memory operations. To this end, we propose Fine-Mem, a unified framework designed for fine-grained feedback alignment. First, we introduce a Chunk-level Step Reward to provide immediate step-level supervision via auxiliary chunk-specific question answering tasks. Second, we devise Evidence-Anchored Reward Attribution to redistribute global rewards by anchoring credit to key memory operations, based on the specific memory items utilized as evidence in reasoning. Together, these components enable stable policy optimization and align local memory operations with the long-term utility of memory. Experiments on Memalpha and MemoryAgentBench demonstrate that Fine-Mem consistently outperforms strong baselines, achieving superior success rates across various sub-tasks. Further analysis reveals its adaptability and strong generalization capabilities across diverse model configurations and backbones.

Lei Zhao, Zihao Ma, Boyu Lin et al.

We present an RL-central framework for Language and Vision Assistants (RLLaVA) with its formulation of Markov decision process (MDP). RLLaVA decouples RL algorithmic logic from model architecture and distributed execution, supporting researchers in implementing new RL algorithms with minimal code, and to plug in a broad family of RL methods and vision-language models (VLMs) while remaining agnostic to specific training and inference engines. RLLaVA makes resource-efficient training of 1B--7B models feasible on common GPUs; notably, 4B-scale models can be trained end-to-end with full-parameter updates on a single 24GB GPU. Experiments on multi-modal and agentic tasks demonstrate that RLLaVA has task extensibility, and the models trained with it consistently improve performance over base models, competitive with other specially engineered RL frameworks. The code is available at https://github.com/TinyLoopX/RLLaVA.

Yangfan Ye, Xiaocheng Feng, Xiachong Feng et al.

Joint multilingual instruction tuning is a widely adopted approach to improve the multilingual instruction-following ability and downstream performance of large language models (LLMs), but the resulting multilingual capability remains highly sensitive to the composition and selection of the training data. Existing selection methods, often based on features like text quality, diversity, or task relevance, typically overlook the intrinsic linguistic structure of multilingual data. In this paper, we propose LangGPS, a lightweight two-stage pre-selection framework guided by language separability which quantifies how well samples in different languages can be distinguished in the model's representation space. LangGPS first filters training data based on separability scores and then refines the subset using existing selection methods. Extensive experiments across six benchmarks and 22 languages demonstrate that applying LangGPS on top of existing selection methods improves their effectiveness and generalizability in multilingual training, especially for understanding tasks and low-resource languages. Further analysis reveals that highly separable samples facilitate the formation of clearer language boundaries and support faster adaptation, while low-separability samples tend to function as bridges for cross-lingual alignment. Besides, we also find that language separability can serve as an effective signal for multilingual curriculum learning, where interleaving samples with diverse separability levels yields stable and generalizable gains. Together, we hope our work offers a new perspective on data utility in multilingual contexts and support the development of more linguistically informed LLMs.

Yangfan Ye, Xiaocheng Feng, Xiachong Feng et al.

Languages encode distinct abstractions and inductive priors, yet most large language models (LLMs) overlook this diversity by reasoning in a single dominant language. In this work, we introduce x1, a family of reasoning models that can adaptively reason in an advantageous language on a per-instance basis. To isolate the effect of reasoning-language choice, x1 is constructed without expanding the model's knowledge boundaries and is trained by contrasting linguistically distinct reasoning trajectories for the same input. Our extensive experiments demonstrate the benefits of adaptive multilingual reasoning across multilingual mathematical reasoning and culturally grounded tasks. Moreover, our results challenge a simplistic view of scaling laws: while scaling reduces cross-lingual disparities in procedural domains such as math reasoning, it does not eliminate the advantages of culture-associated languages in culturally grounded tasks, as we empirically show that such reasoning enables more efficient and accurate cultural knowledge recall. Overall, our findings establish language choice as a functional component of reasoning, with implications for building more generalist and globally competent reasoning models.

Xingjian Zhang, Siwei Wen, Wenjun Wu et al.

Recently, improving the reasoning ability of large multimodal models (LMMs) through reinforcement learning has made great progress. However, most existing works are based on highly reasoning-intensive datasets such as mathematics and code, and researchers generally choose large-scale models as the foundation. We argue that exploring small-scale models' reasoning capabilities remains valuable for researchers with limited computational resources. Moreover, enabling models to explain their reasoning processes on general question-answering datasets is equally meaningful. Therefore, we present the small-scale video reasoning model TinyLLaVA-Video-R1. Based on TinyLLaVA-Video, a traceably trained video understanding model with no more than 4B parameters, it not only demonstrates significantly improved reasoning and thinking capabilities after using reinforcement learning on general Video-QA datasets, but also exhibits the emergent characteristic of "aha moments". Furthermore, we share a series of experimental findings, aiming to provide practical insights for future exploration of video reasoning (thinking) abilities in small-scale models. It is available at https://github.com/ZhangXJ199/TinyLLaVA-Video-R1.

Xiachong Feng, Yi Jiang, Xiaocheng Feng et al.

Social intelligence, the ability to navigate complex interpersonal interactions, presents a fundamental challenge for language agents. Training such agents via reinforcement learning requires solving the credit assignment problem: determining how individual utterances contribute to multi-turn dialogue outcomes. Existing approaches directly employ language models to distribute episode-level rewards, yielding attributions that are retrospective and lack theoretical grounding. We propose SAVOIR (ShApley Value fOr SocIal RL), a novel principled framework grounded in cooperative game theory. Our approach combines two complementary principles: expected utility shifts evaluation from retrospective attribution to prospective valuation, capturing an utterance's strategic potential for enabling favorable future trajectories; Shapley values ensure fair credit distribution with axiomatic guarantees of efficiency, symmetry, and marginality. Experiments on the SOTOPIA benchmark demonstrate that SAVOIR achieves new state-of-the-art performance across all evaluation settings, with our 7B model matching or exceeding proprietary models including GPT-4o and Claude-3.5-Sonnet. Notably, even large reasoning models consistently underperform, suggesting social intelligence requires qualitatively different capabilities than analytical reasoning.

Xiachong Feng, Deyi Yin, Xiaocheng Feng et al.

Games offer a compelling paradigm for developing general reasoning capabilities in language models, as they naturally demand strategic planning, probabilistic inference, and adaptive decision-making. However, existing self-play approaches rely solely on terminal game outcomes, providing no mechanism to distinguish transferable reasoning patterns from game-specific heuristics. We present STRATAGEM, which addresses two fundamental barriers to reasoning transfer: domain specificity, where learned patterns remain anchored in game semantics, and contextual stasis, where static game contexts fail to cultivate progressive reasoning. STRATAGEM selectively reinforces trajectories exhibiting abstract, domain-agnostic reasoning through a Reasoning Transferability Coefficient, while incentivizing adaptive reasoning development via a Reasoning Evolution Reward. Experiments across mathematical reasoning, general reasoning, and code generation benchmarks demonstrate substantial improvements, with particularly strong gains on competition-level mathematics where multi-step reasoning is critical. Ablation studies and human evaluation confirm that both components contribute to transferable reasoning.

Lei Huang, Weitao Li, Chenrui Zhang et al.

Cross-domain recommendation has attracted substantial interest in industrial apps such as Meituan, which serves multiple business domains via knowledge transfer and meets the diverse interests of users. However, existing methods typically follow an implicit modeling paradigm that blends the knowledge from both the source and target domains, and design intricate network structures to share learned embeddings or patterns between domains to improve recommendation accuracy. Since the transfer of interest signals is unsupervised, these implicit paradigms often struggle with the negative transfer resulting from differences in service functions and presentation forms across different domains. In this paper, we propose a simple and effective EXplicit Interest Transfer framework named EXIT to address the stated challenge. Specifically, we propose a novel label combination approach that enables the model to directly learn beneficial source domain interests through supervised learning, while excluding inappropriate interest signals. Moreover, we introduce a scene selector network to model the interest transfer intensity under fine-grained scenes. Offline experiments conducted on the industrial production dataset and online A/B tests validate the superiority and effectiveness of our proposed framework. Without complex network structures or training processes, EXIT can be easily deployed in the industrial recommendation system. EXIT has been successfully deployed in the online homepage recommendation system of Meituan App, serving the main traffic.

Xuhua Li, Weize Sun, Lei Huang et al.

Filter pruning is a common method to achieve model compression and acceleration in deep neural networks (DNNs).Some research regarded filter pruning as a combinatorial optimization problem and thus used evolutionary algorithms (EA) to prune filters of DNNs. However, it is difficult to find a satisfactory compromise solution in a reasonable time due to the complexity of solution space searching. To solve this problem, we first formulate a multi-objective optimization problem based on a sub-network of the full model and propose a Sub-network Multiobjective Evolutionary Algorithm (SMOEA) for filter pruning. By progressively pruning the convolutional layers in groups, SMOEA can obtain a lightweight pruned result with better performance.Experiments on VGG-14 model for CIFAR-10 verify the effectiveness of the proposed SMOEA. Specifically, the accuracy of the pruned model with 16.56% parameters decreases by 0.28% only, which is better than the widely used popular filter pruning criteria.

Wei Shao, Lei Huang, Shuqi Liu et al.

Embedding-based neural topic models could explicitly represent words and topics by embedding them to a homogeneous feature space, which shows higher interpretability. However, there are no explicit constraints for the training of embeddings, leading to a larger optimization space. Also, a clear description of the changes in embeddings and the impact on model performance is still lacking. In this paper, we propose an embedding regularized neural topic model, which applies the specially designed training constraints on word embedding and topic embedding to reduce the optimization space of parameters. To reveal the changes and roles of embeddings, we introduce \textbf{uniformity} into the embedding-based neural topic model as the evaluation metric of embedding space. On this basis, we describe how embeddings tend to change during training via the changes in the uniformity of embeddings. Furthermore, we demonstrate the impact of changes in embeddings in embedding-based neural topic models through ablation studies. The results of experiments on two mainstream datasets indicate that our model significantly outperforms baseline models in terms of the harmony between topic quality and document modeling. This work is the first attempt to exploit uniformity to explore changes in embeddings of embedding-based neural topic models and their impact on model performance to the best of our knowledge.

Guobin Shen, Chenxiao Zhao, Xiang Cheng et al.

Training stability remains a central challenge in reinforcement learning (RL) for large language models (LLMs). Policy staleness, asynchronous training, and mismatches between training and inference engines all cause the behavior policy to diverge from the current policy, risking training collapse. Importance sampling provides a principled correction for this distribution shift but suffers from high variance; existing remedies such as token-level clipping and sequence-level normalization lack a unified theoretical foundation. We propose Variational sEquence-level Soft Policy Optimization (VESPO). By incorporating variance reduction into a variational formulation over proposal distributions, VESPO derives a closed-form reshaping kernel that operates directly on sequence-level importance weights without length normalization. Experiments on mathematical reasoning benchmarks show that VESPO maintains stable training under staleness ratios up to 64x and fully asynchronous execution, and delivers consistent gains across both dense and Mixture-of-Experts models. Code is available at https://github.com/FloyedShen/VESPO

Junlong Jia, Ying Hu, Xi Weng et al.

We present TinyLLaVA Factory, an open-source modular codebase for small-scale large multimodal models (LMMs) with a focus on simplicity of code implementations, extensibility of new features, and reproducibility of training results. Following the design philosophy of the factory pattern in software engineering, TinyLLaVA Factory modularizes the entire system into interchangeable components, with each component integrating a suite of cutting-edge models and methods, meanwhile leaving room for extensions to more features. In addition to allowing users to customize their own LMMs, TinyLLaVA Factory provides popular training recipes to let users pretrain and finetune their models with less coding effort. Empirical experiments validate the effectiveness of our codebase. The goal of TinyLLaVA Factory is to assist researchers and practitioners in exploring the wide landscape of designing and training small-scale LMMs with affordable computational resources.

Xingjian Zhang, Siwei Wen, Wenjun Wu et al.

Large Language Models (LLMs) have made remarkable progress in enhancing step-by-step reasoning through reinforcement learning. However, the Group Relative Policy Optimization (GRPO) algorithm, which relies on sparse reward rules, often encounters the issue of identical rewards within groups, leading to the advantage collapse problem. Existing works typically address this challenge from two perspectives: enforcing model reflection to enhance response diversity, and introducing internal feedback to augment the training signal (advantage). In this work, we begin by analyzing the limitations of model reflection and investigating the policy entropy of responses at the fine-grained sample level. Based on our experimental findings, we propose the EDGE-GRPO algorithm, which adopts \textbf{E}ntropy-\textbf{D}riven Advantage and \textbf{G}uided \textbf{E}rror Correction to effectively mitigate the problem of advantage collapse. Extensive experiments on several main reasoning benchmarks demonstrate the effectiveness and superiority of our approach. It is available at https://github.com/ZhangXJ199/EDGE-GRPO.

Yuxuan Gu, Lunjun Liu, Xiaocheng Feng et al.

An essential problem in artificial intelligence is whether LLMs can simulate human cognition or merely imitate surface-level behaviors, while existing datasets suffer from either synthetic reasoning traces or population-level aggregation, failing to capture authentic individual cognitive patterns. We introduce a benchmark grounded in the longitudinal research trajectories of 217 researchers across diverse domains of artificial intelligence, where each author's scientific publications serve as an externalized representation of their cognitive processes. To distinguish whether LLMs transfer cognitive patterns or merely imitate behaviors, our benchmark deliberately employs a cross-domain, temporal-shift generalization setting. A multidimensional cognitive alignment metric is further proposed to assess individual-level cognitive consistency. Through systematic evaluation of state-of-the-art LLMs and various enhancement techniques, we provide a first-stage empirical study on the questions: (1) How well do current LLMs simulate human cognition? and (2) How far can existing techniques enhance these capabilities?

Lei Huang

The finite-length performance of spatially-coupled low-density parity-check (SC-LDPC) codes is strongly affected by short cycle configurations and the harmful structures induced by them. This paper studies SC-LDPC code design directly at the protograph level, where the design variables are the edge-spreading assignments specified by the partition matrix. In contrast to CLLL/Moser--Tardos based constructive frameworks for QC-SC-LDPC codes, we focus on sharper nonconstructive existence and counting bounds. By encoding cycle-activation conditions as polynomial vanishing constraints over finite grids, we apply the Combinatorial Nullstellensatz to derive sufficient memory conditions for eliminating prescribed cycle-induced harmful structures. For fully connected $(γ,κ)$ base graphs, the resulting bounds explicitly characterize the memory required to destroy all $4$-cycles as well as all $4$- and $6$-cycles, and for fixed $γ$, they are asymptotically tight up to a constant factor compared with known lower bounds. We further apply the Alon--Füredi theorem to obtain lower bounds on the number of feasible edge-spreading assignments, including an explicit counting bound for assignments that eliminate all $4$-cycles and hence yield girth at least six. These results provide a refined algebraic-combinatorial characterization of the feasible design space for high-memory SC-LDPC codes, although no corresponding construction algorithm is provided.

Zekun Zhou, Xiaocheng Feng, Lei Huang et al.

Research is a fundamental process driving the advancement of human civilization, yet it demands substantial time and effort from researchers. In recent years, the rapid development of artificial intelligence (AI) technologies has inspired researchers to explore how AI can accelerate and enhance research. To monitor relevant advancements, this paper presents a systematic review of the progress in this domain. Specifically, we organize the relevant studies into three main categories: hypothesis formulation, hypothesis validation, and manuscript publication. Hypothesis formulation involves knowledge synthesis and hypothesis generation. Hypothesis validation includes the verification of scientific claims, theorem proving, and experiment validation. Manuscript publication encompasses manuscript writing and the peer review process. Furthermore, we identify and discuss the current challenges faced in these areas, as well as potential future directions for research. Finally, we also offer a comprehensive overview of existing benchmarks and tools across various domains that support the integration of AI into the research process. We hope this paper serves as an introduction for beginners and fosters future research. Resources have been made publicly available at https://github.com/zkzhou126/AI-for-Research.

Yuanshuai Wang, Xingjian Zhang, Jinkun Zhao et al.

Large Language Models (LLMs) are key technologies driving intelligent systems to handle multiple tasks. To meet the demands of various tasks, an increasing number of LLMs-driven experts with diverse capabilities have been developed, accompanied by corresponding benchmarks to evaluate their performance. This paper proposes the Bench-CoE framework, which enables Collaboration of Experts (CoE) by effectively leveraging benchmark evaluations to achieve optimal performance across various tasks. Bench-CoE includes a set of expert models, a router for assigning tasks to corresponding experts, and a benchmark dataset for training the router. Moreover, we formulate Query-Level and Subject-Level approaches based on our framework, and analyze the merits and drawbacks of these two approaches. Finally, we conduct a series of experiments with vary data distributions on both language and multimodal tasks to validate that our proposed Bench-CoE outperforms any single model in terms of overall performance. We hope this method serves as a baseline for further research in this area. The code is available at \url{https://github.com/ZhangXJ199/Bench-CoE}.

Hao Guo, Erpeng Xue, Lei Huang et al.

Deep Learning Recommendation Models (DLRMs) often rely on extensive manual feature engineering to improve accuracy and user experience, which increases system complexity and limits scalability of model performance with respect to computational resources. Recently, Meta introduced a generative ranking paradigm based on HSTU block that enables end-to-end learning from raw user behavior sequences and demonstrates scaling law on large datasets that can be regarded as the state-of-the-art (SOTA). However, splitting user behaviors into interleaved item and action information significantly increases the input sequence length, which adversely affects both training and inference efficiency. To address this issue, we propose the Dual-Flow Generative Ranking Network (DFGR), that employs a dual-flow mechanism to optimize interaction modeling, ensuring efficient training and inference through end-to-end token processing. DFGR duplicates the original user behavior sequence into a real flow and a fake flow based on the authenticity of the action information, and then defines a novel interaction method between the real flow and the fake flow within the QKV module of the self-attention mechanism. This design reduces computational overhead and improves both training efficiency and inference performance compared to Meta's HSTU-based model. Experiments on both open-source and real industrial datasets show that DFGR outperforms DLRM, which serves as the industrial online baseline with extensive feature engineering, as well as Meta's HSTU and other common recommendation models such as DIN, DCN, DIEN, and DeepFM. Furthermore, we investigate optimal parameter allocation strategies under computational constraints, establishing DFGR as an efficient and effective next-generation generative ranking paradigm.

Yuxin Guo, Yihao Yue, Yunhao Ni et al.

Layer normalization (LN) is a fundamental component in modern deep learning, but its per-sample centering and scaling introduce non-negligible inference overhead. RMSNorm improves efficiency by removing the centering operation, yet this may discard benefits associated with centering. This paper propose a framework to determine whether an LN in an arbitrary DNN can be replaced by RMSNorm without changing the model function. The key idea is to fold LN's centering operation into upstream general linear layers by enforcing zero-mean outputs through the column-centered constraint (CCC) and column-based weight centering (CBWC). We extend the analysis to arbitrary DNNs, define such LNs as foldable LNs, and develop a graph-based detection algorithm. Our analysis shows that many LNs in widely used architectures are foldable, enabling exact inference-time conversion and end-to-end acceleration of 2% to 12% without changing model predictions. Experiments across multiple task families further show that, when exact equivalence is partially broken in practical training settings, our method remains competitive with vanilla LN while improving efficiency.

Xiang Shen, Yuhang Zhou, Yifan Wu et al.

Memory-augmented LLM agents have advanced personalized recommendation, yet existing approaches universally adopt flat memory representations that conflate ephemeral signals with stable preferences, and none provides a complete lifecycle governing how memory should evolve. We propose MARS (Memory-Augmented Agentic Recommender System), a framework that treats recommendation as a partially observable problem and maintains a structured belief state that progressively abstracts noisy behavioral observations into a compact estimate of user preferences. MARS organizes this belief state into three tiers: event memory buffers raw signals, preference memory maintains fine-grained mutable chunks with explicit strength and evidence tracking, and profile memory distills all preferences into a coherent natural language narrative. A complete lifecycle of six operations -- extraction, reinforcement, weakening, consolidation, forgetting, and resynthesis -- is adaptively scheduled by an LLM-based planner rather than fixed-interval heuristics. Experiments on four InstructRec benchmark domains show that \ours achieves state-of-the-art performance with average improvements of 26.4% in HR@1 and 10.3% in NDCG@10 over the strongest baselines with further gains from agentic scheduling in evolving settings.

Wenfeng Zhang, Jun Ni, Yue Meng et al.

Object detection in unmanned aerial vehicle (UAV) images remains a highly challenging task, primarily caused by the complexity of background noise and the imbalance of target scales. Traditional methods easily struggle to effectively separate objects from intricate backgrounds and fail to fully leverage the rich multi-scale information contained within images. To address these issues, we have developed a synergistic feature fusion network (SFFNet) with dual-domain edge enhancement specifically tailored for object detection in UAV images. Firstly, the multi-scale dynamic dual-domain coupling (MDDC) module is designed. This component introduces a dual-driven edge extraction architecture that operates in both the frequency and spatial domains, enabling effective decoupling of multi-scale object edges from background noise. Secondly, to further enhance the representation capability of the model's neck in terms of both geometric and semantic information, a synergistic feature pyramid network (SFPN) is proposed. SFPN leverages linear deformable convolutions to adaptively capture irregular object shapes and establishes long-range contextual associations around targets through the designed wide-area perception module (WPM). Moreover, to adapt to the various applications or resource-constrained scenarios, six detectors of different scales (N/S/M/B/L/X) are designed. Experiments on two challenging aerial datasets (VisDrone and UAVDT) demonstrate the outstanding performance of SFFNet-X, achieving 36.8 AP and 20.6 AP, respectively. The lightweight models (N/S) also maintain a balance between detection accuracy and parameter efficiency. The code will be available at https://github.com/CQNU-ZhangLab/SFFNet.

Xingjian Zhang, Xi Weng, Yihao Yue et al.

Video behavior recognition and scene understanding are fundamental tasks in multimodal intelligence, serving as critical building blocks for numerous real-world applications. Through large multimodal models (LMMs) have achieved remarkable progress in video understanding, most existing open-source models rely on over 7B parameters and require large-scale datasets for training, making them resource-intensive and inaccessible to many researchers. Furthermore, lightweight models face persistent challenges in effectively processing long visual sequences and temporal understanding. In this work, we introduce TinyLLaVA-Video, a lightweight yet powerful video understanding model with approximately 3.6B parameters. The cornerstone of our design is the video-level group resampler, a novel mechanism that significantly reduces and controls the number of visual tokens at the video level. Unlike traditional image-level resampler, our approach effectively mitigates redundancy while enhancing temporal comprehension, leading to improved performance on video-based tasks. In addition, TinyLLaVA-Video demonstrates exceptional efficiency, requiring only one day of training on 8 A100-40G GPUs. It surpasses several existing 7B-parameter models on multiple benchmarks. We believe this work provides a valuable foundation for future research on lightweight video understanding models. The code and weights is available at https://github.com/ZhangXJ199/TinyLLaVA-Video.

Ze Zhang, Enyuan Zhao, Di Niu et al.

Driven by practical demands in land resource monitoring and national defense security, this paper introduces the Remote Sensing Copy-Move Question Answering (RSCMQA) task. Unlike traditional Remote Sensing Visual Question Answering (RSVQA), RSCMQA focuses on interpreting complex tampering scenarios and inferring relationships between objects. We present a suite of global RSCMQA datasets, comprising images from 29 different regions across 14 countries. Specifically, we propose five distinct datasets, including the basic dataset RS-CMQA, the category-balanced dataset RS-CMQA-B, the high-authenticity dataset Real-RSCM, the extended dataset RS-TQA, and the extended category-balanced dataset RS-TQA-B. These datasets fill a critical gap in the field while ensuring comprehensiveness, balance, and challenge. Furthermore, we introduce a region-discrimination-guided multimodal copy-move forgery perception framework (CMFPF), which enhances the accuracy of answering questions about tampered images by leveraging prompt about the differences and connections between the source and tampered domains. Extensive experiments demonstrate that our method provides a stronger benchmark for RSCMQA compared to general VQA and RSVQA models. Our datasets and code are publicly available at https://github.com/shenyedepisa/RSCMQA.

Lei Huang, Xiang Cheng, Chenxiao Zhao et al.

Large language models (LLMs) typically receive diverse natural language (NL) feedback through interaction with the environment. However, current reinforcement learning (RL) algorithms rely solely on scalar rewards, leaving the rich information in NL feedback underutilized and leading to inefficient exploration. In this work, we propose GOLF, an RL framework that explicitly exploits group-level language feedback to guide targeted exploration through actionable refinements. GOLF aggregates two complementary feedback sources: (i) external critiques that pinpoint errors or propose targeted fixes, and (ii) intra-group attempts that supply alternative partial ideas and diverse failure patterns. These group-level feedbacks are aggregated to produce high-quality refinements, which are adaptively injected into training as off-policy scaffolds to provide targeted guidance in sparse-reward regions. Meanwhile, GOLF jointly optimizes generation and refinement within a unified RL loop, creating a virtuous cycle that continuously improves both capabilities. Experiments on both verifiable and non-verifiable benchmarks show that GOLF achieves superior performance and exploration efficiency, achieving 2.2$\times$ improvements in sample efficiency compared to RL methods trained solely on scalar rewards. Code is available at https://github.com/LuckyyySTA/GOLF.

Wei Ma, Shaowu Chen, Junjie Ye et al.

Existing video inference (VI) enhancement methods typically aim to improve performance by scaling up model sizes and employing sophisticated network architectures. While these approaches demonstrated state-of-the-art performance, they often overlooked the trade-off of resource efficiency and inference effectiveness, leading to inefficient resource utilization and suboptimal inference performance. To address this problem, a fuzzy controller (FC-r) is developed based on key system parameters and inference-related metrics. Guided by the FC-r, a VI enhancement framework is proposed, where the spatiotemporal correlation of targets across adjacent video frames is leveraged. Given the real-time resource conditions of the target device, the framework can dynamically switch between models of varying scales during VI. Experimental results demonstrate that the proposed method effectively achieves a balance between resource utilization and inference performance.

Guobin Shen, Xiang Cheng, Chenxiao Zhao et al.

On-policy self-distillation, where a student is pulled toward a copy of itself conditioned on privileged context (e.g., a verified solution or feedback), offers a promising direction for advancing reasoning capability without a stronger external teacher. Yet in math reasoning the gains are inconsistent, even when the same approach succeeds elsewhere. A pointwise mutual information analysis traces the failure to the privileged context itself: it inflates the teacher's confidence on tokens already implied by the solution (structural connectives, verifiable claims) and deflates it on deliberation tokens ("Wait", "Let", "Maybe") that drive multi-step search. We propose Anti-Self-Distillation (AntiSD), which ascends a divergence between student and teacher rather than descending it: this reverses the per-token sign and yields a naturally bounded advantage in one step. An entropy-triggered gate disables the term once the teacher entropy collapses, completing a drop-in replacement for default self-distillation. Across five models from 4B to 30B parameters on math reasoning benchmarks, AntiSD reaches the GRPO baseline's accuracy in 2 to 10x fewer training steps and improves final accuracy by up to 11.5 points. AntiSD opens a path to scalable self-improvement, where a language model bootstraps its own reasoning through its training signal.

Guobin Shen, Lei Huang, Xiang Cheng et al.

On-policy self-distillation has emerged as a promising paradigm for post-training language models, in which the model conditions on environment feedback to serve as its own teacher, providing dense token-level rewards without external teacher models or step-level annotations. Despite its empirical success, what this reward actually measures and what kind of credit it assigns remain unclear. Under a posterior-compatibility interpretation of feedback conditioning, standard in the implicit-reward literature, we show that the self-distillation token reward is a Bayesian filtering increment whose trajectory sum is exactly the pointwise mutual information between the response and the feedback given the input. This pMI can be raised by input-specific reasoning or by input-generic shortcuts, so we further decompose the teacher log-probability along the input axis. Based on this analysis, we propose CREDIT (Contrastive REward from DIsTillation), which isolates the input-specific component with a batch-contrastive baseline. At the sequence level, CREDIT is a teacher-side surrogate for a contrastive pMI objective that also penalizes responses remaining likely under unrelated inputs. Across coding, scientific reasoning, and tool-use benchmarks on two model families, CREDIT delivers the strongest aggregate performance at negligible additional compute.

Baichuan Zhou, Ying Hu, Xi Weng et al.

We present the TinyLLaVA framework that provides a unified perspective in designing and analyzing the small-scale Large Multimodal Models (LMMs). We empirically study the effects of different vision encoders, connection modules, language models, training data and training recipes. Our extensive experiments showed that better quality of data combined with better training recipes, smaller LMMs can consistently achieve on-par performances compared to bigger LMMs. Under our framework, we train a family of small-scale LMMs. Our best model, TinyLLaVA-3.1B, achieves better overall performance against existing 7B models such as LLaVA-1.5 and Qwen-VL. We hope our findings can serve as baselines for future research in terms of data scaling, training setups and model selections. Our model weights and codes will be made public.