Yichao Wu

Jun Guo, Aishan Liu, Xingyu Zheng et al.

Despite the broad application of Machine Learning models as a Service (MLaaS), they are vulnerable to model stealing attacks. These attacks can replicate the model functionality by using the black-box query process without any prior knowledge of the target victim model. Existing stealing defenses add deceptive perturbations to the victim's posterior probabilities to mislead the attackers. However, these defenses are now suffering problems of high inference computational overheads and unfavorable trade-offs between benign accuracy and stealing robustness, which challenges the feasibility of deployed models in practice. To address the problems, this paper proposes Isolation and Induction (InI), a novel and effective training framework for model stealing defenses. Instead of deploying auxiliary defense modules that introduce redundant inference time, InI directly trains a defensive model by isolating the adversary's training gradient from the expected gradient, which can effectively reduce the inference computational cost. In contrast to adding perturbations over model predictions that harm the benign accuracy, we train models to produce uninformative outputs against stealing queries, which can induce the adversary to extract little useful knowledge from victim models with minimal impact on the benign performance. Extensive experiments on several visual classification datasets (e.g., MNIST and CIFAR10) demonstrate the superior robustness (up to 48% reduction on stealing accuracy) and speed (up to 25.4x faster) of our InI over other state-of-the-art methods. Our codes can be found in https://github.com/DIG-Beihang/InI-Model-Stealing-Defense.

Chunyu Sun, Chenye Xu, Chengyuan Yao et al.

Adversarial training (AT) methods are effective against adversarial attacks, yet they introduce severe disparity of accuracy and robustness between different classes, known as the robust fairness problem. Previously proposed Fair Robust Learning (FRL) adaptively reweights different classes to improve fairness. However, the performance of the better-performed classes decreases, leading to a strong performance drop. In this paper, we observed two unfair phenomena during adversarial training: different difficulties in generating adversarial examples from each class (source-class fairness) and disparate target class tendencies when generating adversarial examples (target-class fairness). From the observations, we propose Balance Adversarial Training (BAT) to address the robust fairness problem. Regarding source-class fairness, we adjust the attack strength and difficulties of each class to generate samples near the decision boundary for easier and fairer model learning; considering target-class fairness, by introducing a uniform distribution constraint, we encourage the adversarial example generation process for each class with a fair tendency. Extensive experiments conducted on multiple datasets (CIFAR-10, CIFAR-100, and ImageNette) demonstrate that our method can significantly outperform other baselines in mitigating the robust fairness problem (+5-10\% on the worst class accuracy)

Xianyou Li, Weiran Yan, Yichao Wu et al.

Tool-using multi-agent large language model (LLM) systems spend computation through model tokens, tool calls, retries, and code execution before producing an answer. When a run fails, final-answer evaluation reveals the endpoint but usually not the point at which the trajectory stopped making recoverable progress. This paper introduces a failure-aware observability framework for diagnosing wasted computation in multi-agent LLM traces. The framework maps recurring failure modes to online trace signals, including tool reliability, execution recovery, orchestration loops, evidence availability, information change, and budget pressure. We instantiate the framework in a three- agent question-answering system and evaluate it on 165 GAIA validation traces under identical execution caps. Operational failures remain common: 22/53 level-1 runs, 33/86 level-2 runs, and 12/26 level-3 runs fail to produce a usable final answer. The traces expose different mechanisms behind these outcomes, including insufficient evidence, repeated-action loops, max-step termination, tool-failure streaks, and execution calls that succeed without useful output. Mean token use rises from 8,152 tokens at level 1 to 16,389 tokens at level 3, while evidence availability and sentence-level support diverge. A cached 10-trace LLM-judge grounding audit shows that cheap online signals and deeper semantic metrics capture complementary layers of failure. The results position failure-aware observability as a diagnostic layer between raw execution logs and final-answer accuracy.

Gang Li, Xiang Li, Yujie Wang et al.

The Mean-Teacher (MT) scheme is widely adopted in semi-supervised object detection (SSOD). In MT, the sparse pseudo labels, offered by the final predictions of the teacher (e.g., after Non Maximum Suppression (NMS) post-processing), are adopted for the dense supervision for the student via hand-crafted label assignment. However, the sparse-to-dense paradigm complicates the pipeline of SSOD, and simultaneously neglects the powerful direct, dense teacher supervision. In this paper, we attempt to directly leverage the dense guidance of teacher to supervise student training, i.e., the dense-to-dense paradigm. Specifically, we propose the Inverse NMS Clustering (INC) and Rank Matching (RM) to instantiate the dense supervision, without the widely used, conventional sparse pseudo labels. INC leads the student to group candidate boxes into clusters in NMS as the teacher does, which is implemented by learning grouping information revealed in NMS procedure of the teacher. After obtaining the same grouping scheme as the teacher via INC, the student further imitates the rank distribution of the teacher over clustered candidates through Rank Matching. With the proposed INC and RM, we integrate Dense Teacher Guidance into Semi-Supervised Object Detection (termed DTG-SSOD), successfully abandoning sparse pseudo labels and enabling more informative learning on unlabeled data. On COCO benchmark, our DTG-SSOD achieves state-of-the-art performance under various labelling ratios. For example, under 10% labelling ratio, DTG-SSOD improves the supervised baseline from 26.9 to 35.9 mAP, outperforming the previous best method Soft Teacher by 1.9 points.

Jiaheng Liu, Haoyu Qin, Yichao Wu et al.

Knowledge distillation is an effective method to improve the performance of a lightweight neural network (i.e., student model) by transferring the knowledge of a well-performed neural network (i.e., teacher model), which has been widely applied in many computer vision tasks, including face recognition. Nevertheless, the current face recognition distillation methods usually utilize the Feature Consistency Distillation (FCD) (e.g., L2 distance) on the learned embeddings extracted by the teacher and student models for each sample, which is not able to fully transfer the knowledge from the teacher to the student for face recognition. In this work, we observe that mutual relation knowledge between samples is also important to improve the discriminative ability of the learned representation of the student model, and propose an effective face recognition distillation method called CoupleFace by additionally introducing the Mutual Relation Distillation (MRD) into existing distillation framework. Specifically, in MRD, we first propose to mine the informative mutual relations, and then introduce the Relation-Aware Distillation (RAD) loss to transfer the mutual relation knowledge of the teacher model to the student model. Extensive experimental results on multiple benchmark datasets demonstrate the effectiveness of our proposed CoupleFace for face recognition. Moreover, based on our proposed CoupleFace, we have won the first place in the ICCV21 Masked Face Recognition Challenge (MS1M track).

Yuhang Wang, Huafeng Shi, Rui Min et al.

Extensive evidence has demonstrated that deep neural networks (DNNs) are vulnerable to backdoor attacks, which motivates the development of backdoor attacks detection. Most detection methods are designed to verify whether a model is infected with presumed types of backdoor attacks, yet the adversary is likely to generate diverse backdoor attacks in practice that are unforeseen to defenders, which challenge current detection strategies. In this paper, we focus on this more challenging scenario and propose a universal backdoor attacks detection method named Adaptive Adversarial Probe (A2P). Specifically, we posit that the challenge of universal backdoor attacks detection lies in the fact that different backdoor attacks often exhibit diverse characteristics in trigger patterns (i.e., sizes and transparencies). Therefore, our A2P adopts a global-to-local probing framework, which adversarially probes images with adaptive regions/budgets to fit various backdoor triggers of different sizes/transparencies. Regarding the probing region, we propose the attention-guided region generation strategy that generates region proposals with different sizes/locations based on the attention of the target model, since trigger regions often manifest higher model activation. Considering the attack budget, we introduce the box-to-sparsity scheduling that iteratively increases the perturbation budget from box to sparse constraint, so that we could better activate different latent backdoors with different transparencies. Extensive experiments on multiple datasets (CIFAR-10, GTSRB, Tiny-ImageNet) demonstrate that our method outperforms state-of-the-art baselines by large margins (+12%).

Jing Yang, Yichao Wu, Jianan Liu et al.

Recursive Multi-Agent Trading System (RMATS) integrates four specialized agents -- Sentiment, Report, Analysis, and Risk -- coordinated through a recursive Manager Agent with iterative feedback loops. Experimental evaluation over a 561-trading-day period (January 2023 to March 2025) across a 24-asset multi-class universe demonstrates that RMATS achieves a maximum drawdown of 9.62%, lower than MVO (15.49%) and FinBERT Sentiment (15.28%), and exhibits the lowest event-period drawdown in 3 of 5 geopolitical stress scenarios tested. While RMATS underperforms return-maximizing baselines in a sustained bull market environment, ablation studies confirm the individual contribution of each agent component to downside protection. These results position RMATS as a risk-control-oriented architecture suitable for institutions prioritizing capital preservation under geopolitical uncertainty.

Haoyu Wang, Guozheng Ma, Shugang Cui et al.

While Large Language Models (LLMs) excel in language-based agentic tasks, their applicability to unseen, nonlinguistic environments (e.g., symbolic or spatial tasks) remains limited. Previous work attributes this performance gap to the mismatch between the pretraining distribution and the testing distribution. In this work, we demonstrate the primary bottleneck is the prohibitive cost of exploration: mastering these tasks requires extensive trial-and-error, which is computationally unsustainable for parameter-heavy LLMs operating in a high dimensional semantic space. To address this, we propose SCOUT (Sub-Scale Collaboration On Unseen Tasks), a novel framework that decouples exploration from exploitation. We employ lightweight "scouts" (e.g., small MLPs) to probe environmental dynamics at a speed and scale far exceeding LLMs. The collected trajectories are utilized to bootstrap the LLM via Supervised Fine-Tuning (SFT), followed by multi-turn Reinforcement Learning (RL) to activate its latent world knowledge. Empirically, SCOUT enables a Qwen2.5-3B-Instruct model to achieve an average score of 0.86, significantly outperforming proprietary models, including Gemini-2.5-Pro (0.60), while saving about 60% GPU hours consumption.

Yuanhan Zhang, Yichao Wu, Zhenfei Yin et al.

The field of face anti-spoofing (FAS) has witnessed great progress with the surge of deep learning. Due to its data-driven nature, existing FAS methods are sensitive to the noise in the dataset, which will hurdle the learning process. However, very few works consider noise modeling in FAS. In this work, we attempt to fill this gap by automatically addressing the noise problem from both label and data perspectives in a probabilistic manner. Specifically, we propose a unified framework called Dual Probabilistic Modeling (DPM), with two dedicated modules, DPM-LQ (Label Quality aware learning) and DPM-DQ (Data Quality aware learning). Both modules are designed based on the assumption that data and label should form coherent probabilistic distributions. DPM-LQ is able to produce robust feature representations without overfitting to the distribution of noisy semantic labels. DPM-DQ can eliminate data noise from `False Reject' and `False Accept' during inference by correcting the prediction confidence of noisy data based on its quality distribution. Both modules can be incorporated into existing deep networks seamlessly and efficiently. Furthermore, we propose the generalized DPM to address the noise problem in practical usage without the need of semantic annotations. Extensive experiments demonstrate that this probabilistic modeling can 1) significantly improve the accuracy, and 2) make the model robust to the noise in real-world datasets. Without bells and whistles, our proposed DPM achieves state-of-the-art performance on multiple standard FAS benchmarks.

Ziliang Wang, Xuhui Zheng, Kang An et al.

Efficient multi-hop reasoning requires Large Language Models (LLMs) based agents to acquire high-value external knowledge iteratively. Previous work has explored reinforcement learning (RL) to train LLMs to perform search-based document retrieval, achieving notable improvements in QA performance, but underperform on complex, multi-hop QA resulting from the sparse rewards from global signal only. To address this gap in existing research, we introduce StepSearch, a framework for search LLMs that trained with step-wise proximal policy optimization method. It consists of richer and more detailed intermediate search rewards and token-level process supervision based on information gain and redundancy penalties to better guide each search step. We constructed a fine-grained question-answering dataset containing sub-question-level search trajectories based on open source datasets through a set of data pipeline method. On standard multi-hop QA benchmarks, it significantly outperforms global-reward baselines, achieving 11.2% and 4.2% absolute improvements for 3B and 7B models over various search with RL baselines using only 19k training data, demonstrating the effectiveness of fine-grained, stepwise supervision in optimizing deep search LLMs. Our code will be released on https://github.com/Zillwang/StepSearch.

Jianan Liu, Jing Yang, Xianyou Li et al.

The rapid adoption of artificial intelligence (AI) and large language models (LLMs) is transforming financial analytics by enabling natural language interfaces for reporting, decision support, and automated reasoning. However, limited empirical understanding exists regarding how different LLM-based reasoning architectures perform across realistic financial workflows, particularly under the cost, accuracy, and compliance constraints faced by small and medium-sized enterprises (SMEs). SMEs typically operate within severe infrastructure constraints, lacking cloud GPU budgets, dedicated AI teams, and API-scale inference capacity, making architectural efficiency a first-class concern. To ensure practical relevance, we introduce an explicit SME-constrained evaluation setting in which all experiments are conducted using a locally hosted 8B-parameter instruction-tuned model without cloud-scale infrastructure. This design isolates the impact of architectural choices within a realistic deployment environment. We systematically compare four reasoning architectures: baseline LLM, retrieval-augmented generation (RAG), structured long-term memory, and memory-augmented conversational reasoning across both FinQA and ConvFinQA benchmarks. Results reveal a consistent architectural inversion: structured memory improves precision in deterministic, operand-explicit tasks, while retrieval-based approaches outperform memory-centric methods in conversational, reference-implicit settings. Based on these findings, we propose a hybrid deployment framework that dynamically selects reasoning strategies to balance numerical accuracy, auditability, and infrastructure efficiency, providing a practical pathway for financial AI adoption in resource-constrained environments.

Junqin Huang, Zhongjie Hu, Zihao Jing et al.

In this report, we introduce Piccolo2, an embedding model that surpasses other models in the comprehensive evaluation over 6 tasks on CMTEB benchmark, setting a new state-of-the-art. Piccolo2 primarily leverages an efficient multi-task hybrid loss training approach, effectively harnessing textual data and labels from diverse downstream tasks. In addition, Piccolo2 scales up the embedding dimension and uses MRL training to support more flexible vector dimensions. The latest information of piccolo models can be accessed via: https://huggingface.co/sensenova/

Yichao Wu, Penghao Liang, Yafei Xiang et al.

Retrieval-Augmented Generation (RAG) grounds Large Language Models (LLMs) to mitigate factual hallucinations. Recent paradigms shift from static pipelines to Modular and Agentic RAG frameworks, granting models autonomy for multi-hop reasoning or self-correction. However, current reflective RAG heavily relies on massive LLMs as universal evaluators. In high-throughput systems, executing complete forward passes for billion-parameter models merely for binary routing introduces severe computational redundancy. Furthermore, in autonomous agent scenarios, inaccurate retrieval causes models to expend excessive tokens on spurious reasoning and redundant tool calls, inflating Time-to-First-Token (TTFT) and costs. We propose Tiny-Critic RAG, decoupling evaluation by deploying a parameter-efficient Small Language Model (SLM) via Low-Rank Adaptation (LoRA). Acting as a deterministic gatekeeper, Tiny-Critic employs constrained decoding and non-thinking inference modes for ultra-low latency binary routing. Evaluations on noise-injected datasets demonstrate Tiny-Critic RAG achieves routing accuracy comparable to GPT-4o-mini while reducing latency by an order of magnitude, establishing a highly cost-effective paradigm for agent deployment.

Mengwei Yuan, Jianan Liu, Jing Yang et al.

Large Language Model (LLM) has exhibited strong reasoning ability in text-based contexts across various domains, yet the limitation of context window poses challenges for the model on long-range inference tasks and necessitates a memory storage system. While many current storage approaches have been proposed with episodic notes and graph representations of memory, retrieval methods still primarily rely on predefined workflows or static similarity top-k over embeddings. To address this inflexibility, we introduced a novel tool-augmented autonomous memory retrieval framework (TA-Mem), which contains: (1) a memory extraction LLM agent which is prompted to adaptively chuck an input into sub-context based on semantic correlation, and extract information into structured notes, (2) a multi-indexed memory database designed for different types of query methods including both key-based lookup and similarity-based retrieval, (3) a tool-augmented memory retrieval agent which explores the memory autonomously by selecting appropriate tools provided by the database based on the user input, and decides whether to proceed to the next iteration or finalizing the response after reasoning on the fetched memories. The TA-Mem is evaluated on the LoCoMo dataset, achieving significant performance improvements over existing baseline approaches. In addition, an analysis of tool use across different question types also demonstrates the adaptivity of the proposed method.

Penghao Liang, Mengwei Yuan, Jianan Liu et al.

We present DynaRAG, a retrieval-augmented generation (RAG) framework designed to handle both static and time-sensitive information needs through dynamic knowledge integration. Unlike traditional RAG pipelines that rely solely on static corpora, DynaRAG selectively invokes external APIs when retrieved documents are insufficient for answering a query. The system employs an LLM-based reranker to assess document relevance, a sufficiency classifier to determine when fallback is necessary, and Gorilla v2 -- a state-of-the-art API calling model -- for accurate tool invocation. We further enhance robustness by incorporating schema filtering via FAISS to guide API selection. Evaluations on the CRAG benchmark demonstrate that DynaRAG significantly improves accuracy on dynamic questions, while also reducing hallucinations. Our results highlight the importance of dynamic-aware routing and selective tool use in building reliable, real-world question-answering systems.

Peiqin Zhuang, Lei Bai, Yichao Wu et al.

Recently, action recognition has been dominated by transformer-based methods, thanks to their spatiotemporal contextual aggregation capacities. However, despite the significant progress achieved on scene-related datasets, they do not perform well on motion-sensitive datasets due to the lack of elaborate motion modeling designs. Meanwhile, we observe that the widely-used cost volume in traditional action recognition is highly similar to the affinity matrix defined in self-attention, but equipped with powerful motion modeling capacities. In light of this, we propose to integrate those effective motion modeling properties into the existing transformer in a unified and neat way, with the proposal of the Explicit Motion Information Mining module (EMIM). In EMIM, we propose to construct the desirable affinity matrix in a cost volume style, where the set of key candidate tokens is sampled from the query-based neighboring area in the next frame in a sliding-window manner. Then, the constructed affinity matrix is used to aggregate contextual information for appearance modeling and is converted into motion features for motion modeling as well. We validate the motion modeling capacities of our method on four widely-used datasets, and our method performs better than existing state-of-the-art approaches, especially on motion-sensitive datasets, i.e., Something-Something V1 & V2. Our project is available at https://github.com/PeiqinZhuang/EMIM .

Ruijia Wu, Yuhang Wang, Huafeng Shi et al.

Denoising diffusion models have shown remarkable potential in various generation tasks. The open-source large-scale text-to-image model, Stable Diffusion, becomes prevalent as it can generate realistic artistic or facial images with personalization through fine-tuning on a limited number of new samples. However, this has raised privacy concerns as adversaries can acquire facial images online and fine-tune text-to-image models for malicious editing, leading to baseless scandals, defamation, and disruption to victims' lives. Prior research efforts have focused on deriving adversarial loss from conventional training processes for facial privacy protection through adversarial perturbations. However, existing algorithms face two issues: 1) they neglect the image-text fusion module, which is the vital module of text-to-image diffusion models, and 2) their defensive performance is unstable against different attacker prompts. In this paper, we propose the Adversarial Decoupling Augmentation Framework (ADAF), addressing these issues by targeting the image-text fusion module to enhance the defensive performance of facial privacy protection algorithms. ADAF introduces multi-level text-related augmentations for defense stability against various attacker prompts. Concretely, considering the vision, text, and common unit space, we propose Vision-Adversarial Loss, Prompt-Robust Augmentation, and Attention-Decoupling Loss. Extensive experiments on CelebA-HQ and VGGFace2 demonstrate ADAF's promising performance, surpassing existing algorithms.

Gang Li, Xiang Li, Yujie Wang et al.

In this paper, we delve into two key techniques in Semi-Supervised Object Detection (SSOD), namely pseudo labeling and consistency training. We observe that these two techniques currently neglect some important properties of object detection, hindering efficient learning on unlabeled data. Specifically, for pseudo labeling, existing works only focus on the classification score yet fail to guarantee the localization precision of pseudo boxes; For consistency training, the widely adopted random-resize training only considers the label-level consistency but misses the feature-level one, which also plays an important role in ensuring the scale invariance. To address the problems incurred by noisy pseudo boxes, we design Noisy Pseudo box Learning (NPL) that includes Prediction-guided Label Assignment (PLA) and Positive-proposal Consistency Voting (PCV). PLA relies on model predictions to assign labels and makes it robust to even coarse pseudo boxes; while PCV leverages the regression consistency of positive proposals to reflect the localization quality of pseudo boxes. Furthermore, in consistency training, we propose Multi-view Scale-invariant Learning (MSL) that includes mechanisms of both label- and feature-level consistency, where feature consistency is achieved by aligning shifted feature pyramids between two images with identical content but varied scales. On COCO benchmark, our method, termed PSEudo labeling and COnsistency training (PseCo), outperforms the SOTA (Soft Teacher) by 2.0, 1.8, 2.0 points under 1%, 5%, and 10% labelling ratios, respectively. It also significantly improves the learning efficiency for SSOD, e.g., PseCo halves the training time of the SOTA approach but achieves even better performance. Code is available at https://github.com/ligang-cs/PseCo.

Shaowei Zhang, Faqiang Qian, Yan Chen et al.

Emotion Recognition in Conversation (ERC) has become a fundamental capability for large language models (LLMs) in human-centric interaction. Beyond accurate recognition, coherent emotional expression is also crucial, yet both are limited by the scarcity and static nature of high-quality annotated data. In this work, we propose SELF-EMO, a self-evolution framework grounded in the hypothesis that better emotion prediction leads to more consistent emotional responses. We introduce two auxiliary tasks, emotional understanding and emotional expression, and design a role-based self-play paradigm where the model acts as both an emotion recognizer and a dialogue responder. Through iterative interactions, the model generates diverse conversational trajectories, enabling scalable data generation. To ensure quality, we adopt a data flywheel mechanism that filters candidate predictions and responses using a smoothed IoU-based reward and feeds selected samples back for continuous self-improvement without external supervision. We further develop SELF-GRPO, a reinforcement learning algorithm that stabilizes optimization with multi-label alignment rewards and group-level consistency signals. Experiments on IEMOCAP, MELD, and EmoryNLP show that SELF-EMO achieves state-of-the-art performance, improving accuracy by +6.33% on Qwen3-4B and +8.54% on Qwen3-8B, demonstrating strong effectiveness and generalization.

Yichao Wu, Zhengyu Jin, Chenxi Shi et al.

This paper explores the application of deep learning techniques, particularly focusing on BERT models, in sentiment analysis. It begins by introducing the fundamental concept of sentiment analysis and how deep learning methods are utilized in this domain. Subsequently, it delves into the architecture and characteristics of BERT models. Through detailed explanation, it elucidates the application effects and optimization strategies of BERT models in sentiment analysis, supported by experimental validation. The experimental findings indicate that BERT models exhibit robust performance in sentiment analysis tasks, with notable enhancements post fine-tuning. Lastly, the paper concludes by summarizing the potential applications of BERT models in sentiment analysis and suggests directions for future research and practical implementations.

Xuhui Zheng, Kang An, Ziliang Wang et al.

Multimodal pre-training remains constrained by the descriptive bias of image-caption pairs, leading models to favor surface linguistic cues over grounded visual understanding. We introduce MMRPT, a masked multimodal reinforcement pre-training framework that strengthens visual reasoning in MLLMs. We are the first to incorporate reinforcement learning directly into the pre-training of large vision-language models, enabling learning signals that reward visual grounding rather than caption imitation. MMRPT constructs masked multimodal data by estimating sentence-level visual dependency via attention over visual tokens and masking highly vision-dependent segments; the model reconstructs these spans through vision-grounded reasoning guided by a semantic-visual reward. Experiments show consistent zero-shot gains across diverse benchmarks and substantially improved robustness under supervised fine-tuning, demonstrating that reinforcement-driven masked reasoning provides a more reliable and generalizable pre-training objective for multimodal models.

Chenxi Shi, Penghao Liang, Yichao Wu et al.

The integration of LLMOps into personalized recommendation systems marks a significant advancement in managing LLM-driven applications. This innovation presents both opportunities and challenges for enterprises, requiring specialized teams to navigate the complexity of engineering technology while prioritizing data security and model interpretability. By leveraging LLMOps, enterprises can enhance the efficiency and reliability of large-scale machine learning models, driving personalized recommendations aligned with user preferences. Despite ethical considerations, LLMOps is poised for widespread adoption, promising more efficient and secure machine learning services that elevate user experience and shape the future of personalized recommendation systems.

Yanlin Zhou, Tong Zhan, Yichao Wu et al.

The Human Genome Project has led to an exponential increase in data related to the sequence, structure, and function of biomolecules. Bioinformatics is an interdisciplinary research field that primarily uses computational methods to analyze large amounts of biological macromolecule data. Its goal is to discover hidden biological patterns and related information. Furthermore, analysing additional relevant information can enhance the study of biological operating mechanisms. This paper discusses the fundamental concepts of RNA, RNA secondary structure, and its prediction.Subsequently, the application of machine learning technologies in predicting the structure of biological macromolecules is explored. This chapter describes the relevant knowledge of algorithms and computational complexity and presents a RNA tertiary structure prediction algorithm based on ResNet. To address the issue of the current scoring function's unsuitability for long RNA, a scoring model based on ResNet is proposed, and a structure prediction algorithm is designed. The chapter concludes by presenting some open and interesting challenges in the field of RNA tertiary structure prediction.

Bo Song, Yuanhao Xu, Yichao Wu

Machine learning models have achieved significant milestones in various domains, for example, computer vision models have an exceptional result in object recognition, and in natural language processing, where Large Language Models (LLM) like GPT can start a conversation with human-like proficiency. However, abstract reasoning remains a challenge for these models, Can AI really thinking like a human? still be a question yet to be answered. Raven Progressive Matrices (RPM) is a metric designed to assess human reasoning capabilities. It presents a series of eight images as a problem set, where the participant should try to discover the underlying rules among these images and select the most appropriate image from eight possible options that best completes the sequence. This task always be used to test human reasoning abilities and IQ. Zhang et al proposed a dataset called RAVEN which can be used to test Machine Learning model abstract reasoning ability. In this paper, we purposed Vision Transformer Contrastive Network which build on previous work with the Contrastive Perceptual Inference network (CoPiNet), which set a new benchmark for permutationinvariant models Raven Progressive Matrices by incorporating contrast effects from psychology, cognition, and education, and extends this foundation by leveraging the cutting-edge Vision Transformer architecture. This integration aims to further refine the machine ability to process and reason about spatial-temporal information from pixel-level inputs and global wise features on RAVEN dataset.

Xiaonan Xu, Yichao Wu, Penghao Liang et al.

With the boom of e-commerce and web applications, recommender systems have become an important part of our daily lives, providing personalized recommendations based on the user's preferences. Although deep neural networks (DNNs) have made significant progress in improving recommendation systems by simulating the interaction between users and items and incorporating their textual information, these DNN-based approaches still have some limitations, such as the difficulty of effectively understanding users' interests and capturing textual information. It is not possible to generalize to different seen/unseen recommendation scenarios and reason about their predictions. At the same time, the emergence of large language models (LLMs), represented by ChatGPT and GPT-4, has revolutionized the fields of natural language processing (NLP) and artificial intelligence (AI) due to their superior capabilities in the basic tasks of language understanding and generation, and their impressive generalization and reasoning capabilities. As a result, recent research has sought to harness the power of LLM to improve recommendation systems. Given the rapid development of this research direction in the field of recommendation systems, there is an urgent need for a systematic review of existing LLM-driven recommendation systems for researchers and practitioners in related fields to gain insight into. More specifically, we first introduced a representative approach to learning user and item representations using LLM as a feature encoder. We then reviewed the latest advances in LLMs techniques for collaborative filtering enhanced recommendation systems from the three paradigms of pre-training, fine-tuning, and prompting. Finally, we had a comprehensive discussion on the future direction of this emerging field.

Yichao Wu, Yafei Xiang, Shuning Huo et al.

In addressing the computational and memory demands of fine-tuning Large Language Models(LLMs), we propose LoRA-SP(Streamlined Partial Parameter Adaptation), a novel approach utilizing randomized half-selective parameter freezing within the Low-Rank Adaptation(LoRA)framework. This method efficiently balances pre-trained knowledge retention and adaptability for task-specific optimizations. Through a randomized mechanism, LoRA-SP determines which parameters to update or freeze, significantly reducing computational and memory requirements without compromising model performance. We evaluated LoRA-SP across several benchmark NLP tasks, demonstrating its ability to achieve competitive performance with substantially lower resource consumption compared to traditional full-parameter fine-tuning and other parameter-efficient techniques. LoRA-SP innovative approach not only facilitates the deployment of advanced NLP models in resource-limited settings but also opens new research avenues into effective and efficient model adaptation strategies.

FaQiang Qian, WeiKun Zhang, Ziliang Wang et al.

Shaping powerful LLMs to be beneficial and safe is central to AI alignment. We argue that post-training alignment is fundamentally a unified Preference Learning problem, involving two modalities: demonstrated preferences (e.g., Supervised Fine-Tuning, SFT) and comparative preferences (e.g., Reinforcement Learning, RL).The standard sequential pipeline-SFT followed by RL-is flawed due to a critical distributional mismatch: SFT uses static expert data, but as the policy evolves, its generation distribution drifts, making SFT knowledge brittle. Subsequent RL then explores without direct access to the rich, ground-truth knowledge in expert demonstrations, leading to inefficient, ungrounded updates. This separation prevents mutual regularization between data sources. To address this, we reframe alignment as a constrained optimization problem and propose Unified Adversarial Preference Learning (UniAPL),a novel framework that dynamically aligns the policy's distribution with the expert's. UniAPL implements a single-stage unified training objective, jointly learning from mixed batches of SFT and preference data. In every gradient step, dense expert demonstrations directly ground and regularize online exploration, inherently resolving distributional mismatch and maximizing data synergy.We evaluate UniAPL on instruction-following tasks using Qwen3-235B-Instruct-2507 as the teacher. Our models match or exceed strong GRPO baselines: +5.77% on Qwen3-0.6B (matching a 32B model) and +3.75% on Qwen3-4B,even outperforming the teacher. Analyses of response length and log-probability distributions confirm that UniAPL outputs closely mimic expert demonstrations, achieving both stronger performance and better behavioral alignment.

Ziliang Wang, Kang An, Xuhui Zheng et al.

While search-augmented large language models (LLMs) exhibit impressive capabilities, their reliability in complex multi-hop reasoning remains limited. This limitation arises from three fundamental challenges: decomposition errors, where tasks are incorrectly broken down; retrieval missing, where key evidence fails to be retrieved; and reasoning errors, where flawed logic propagates through the reasoning chain. A single failure in any of these stages can derail the final answer. We propose Erasable Reinforcement Learning (ERL), a novel framework that transforms fragile reasoning into a robust process. ERL explicitly identifies faulty steps, erases them, and regenerates reasoning in place, preventing defective logic from propagating through the reasoning chain. This targeted correction mechanism turns brittle reasoning into a more resilient process. Models trained with ERL, termed ESearch, achieve substantial improvements on HotpotQA, MuSiQue, 2Wiki, and Bamboogle, with the 3B model achieving +8.48% EM and +11.56% F1, and the 7B model achieving +5.38% EM and +7.22% F1 over previous state-of-the-art(SOTA) results. These findings suggest that erasable reinforcement learning provides a powerful paradigm shift for robust multi-step reasoning in LLMs.

Qinghong Sun, Zhenfei Yin, Yichao Wu et al.

With the development of deep learning, the field of face anti-spoofing (FAS) has witnessed great progress. FAS is usually considered a classification problem, where each class is assumed to contain a single cluster optimized by softmax loss. In practical deployment, one class can contain several local clusters, and a single-center is insufficient to capture the inherent structure of the FAS data. However, few approaches consider large distribution discrepancies in the field of FAS. In this work, we propose a unified framework called Latent Distribution Adjusting (LDA) with properties of latent, discriminative, adaptive, generic to improve the robustness of the FAS model by adjusting complex data distribution with multiple prototypes. 1) Latent. LDA attempts to model the data of each class as a Gaussian mixture distribution, and acquire a flexible number of centers for each class in the last fully connected layer implicitly. 2) Discriminative. To enhance the intra-class compactness and inter-class discrepancy, we propose a margin-based loss for providing distribution constrains for prototype learning. 3) Adaptive. To make LDA more efficient and decrease redundant parameters, we propose Adaptive Prototype Selection (APS) by selecting the appropriate number of centers adaptively according to different distributions. 4) Generic. Furthermore, LDA can adapt to unseen distribution by utilizing very few training data without re-training. Extensive experiments demonstrate that our framework can 1) make the final representation space both intra-class compact and inter-class separable, 2) outperform the state-of-the-art methods on multiple standard FAS benchmarks.

Gang Li, Xiang Li, Yujie Wang et al.

Knowledge Distillation (KD) is a widely-used technology to inherit information from cumbersome teacher models to compact student models, consequently realizing model compression and acceleration. Compared with image classification, object detection is a more complex task, and designing specific KD methods for object detection is non-trivial. In this work, we elaborately study the behaviour difference between the teacher and student detection models, and obtain two intriguing observations: First, the teacher and student rank their detected candidate boxes quite differently, which results in their precision discrepancy. Second, there is a considerable gap between the feature response differences and prediction differences between teacher and student, indicating that equally imitating all the feature maps of the teacher is the sub-optimal choice for improving the student's accuracy. Based on the two observations, we propose Rank Mimicking (RM) and Prediction-guided Feature Imitation (PFI) for distilling one-stage detectors, respectively. RM takes the rank of candidate boxes from teachers as a new form of knowledge to distill, which consistently outperforms the traditional soft label distillation. PFI attempts to correlate feature differences with prediction differences, making feature imitation directly help to improve the student's accuracy. On MS COCO and PASCAL VOC benchmarks, extensive experiments are conducted on various detectors with different backbones to validate the effectiveness of our method. Specifically, RetinaNet with ResNet50 achieves 40.4% mAP in MS COCO, which is 3.5% higher than its baseline, and also outperforms previous KD methods.

Yujie Wang, Junqin Huang, Mengya Gao et al.

The foundation model is not the last chapter of the model production pipeline. Transferring with few data in a general way to thousands of downstream tasks is becoming a trend of the foundation model's application. In this paper, we proposed a universal transfer framework: One to Transfer All (OTA) to transfer any Vision Foundation Model (VFM) to any downstream tasks with few downstream data. We first transfer a VFM to a task-specific model by Image Re-representation Fine-tuning (IRF) then distilling knowledge from a task-specific model to a deployed model with data produced by Downstream Image-Guided Generation (DIGG). OTA has no dependency on upstream data, VFM, and downstream tasks when transferring. It also provides a way for VFM researchers to release their upstream information for better transferring but not leaking data due to privacy requirements. Massive experiments validate the effectiveness and superiority of our methods in few data setting. Our code will be released.

Jing Shao, Siyu Chen, Yangguang Li et al.

Enormous waves of technological innovations over the past several years, marked by the advances in AI technologies, are profoundly reshaping the industry and the society. However, down the road, a key challenge awaits us, that is, our capability of meeting rapidly-growing scenario-specific demands is severely limited by the cost of acquiring a commensurate amount of training data. This difficult situation is in essence due to limitations of the mainstream learning paradigm: we need to train a new model for each new scenario, based on a large quantity of well-annotated data and commonly from scratch. In tackling this fundamental problem, we move beyond and develop a new learning paradigm named INTERN. By learning with supervisory signals from multiple sources in multiple stages, the model being trained will develop strong generalizability. We evaluate our model on 26 well-known datasets that cover four categories of tasks in computer vision. In most cases, our models, adapted with only 10% of the training data in the target domain, outperform the counterparts trained with the full set of data, often by a significant margin. This is an important step towards a promising prospect where such a model with general vision capability can dramatically reduce our reliance on data, thus expediting the adoption of AI technologies. Furthermore, revolving around our new paradigm, we also introduce a new data system, a new architecture, and a new benchmark, which, together, form a general vision ecosystem to support its future development in an open and inclusive manner. See project website at https://opengvlab.shlab.org.cn .

Jiaheng Liu, Yudong Wu, Yichao Wu et al.

The field of face recognition (FR) has witnessed great progress with the surge of deep learning. Existing methods mainly focus on extracting discriminative features, and directly compute the cosine or L2 distance by the point-to-point way without considering the context information. In this study, we make a key observation that the local con-text represented by the similarities between the instance and its inter-class neighbors1plays an important role forFR. Specifically, we attempt to incorporate the local in-formation in the feature space into the metric, and pro-pose a unified framework calledInter-class DiscrepancyAlignment(IDA), with two dedicated modules, Discrepancy Alignment Operator(IDA-DAO) andSupport Set Estimation(IDA-SSE). IDA-DAO is used to align the similarity scores considering the discrepancy between the images and its neighbors, which is defined by adaptive support sets on the hypersphere. For practical inference, it is difficult to acquire support set during online inference. IDA-SSE can provide convincing inter-class neighbors by introducing virtual candidate images generated with GAN. Further-more, we propose the learnable IDA-SSE, which can implicitly give estimation without the need of any other images in the evaluation process. The proposed IDA can be incorporated into existing FR systems seamlessly and efficiently. Extensive experiments demonstrate that this frame-work can 1) significantly improve the accuracy, and 2) make the model robust to the face images of various distributions.Without bells and whistles, our method achieves state-of-the-art performance on multiple standard FR benchmarks.

Zhenmao Li, Yichao Wu, Ken Chen et al.

Example weighting algorithm is an effective solution to the training bias problem, however, most previous typical methods are usually limited to human knowledge and require laborious tuning of hyperparameters. In this paper, we propose a novel example weighting framework called Learning to Auto Weight (LAW). The proposed framework finds step-dependent weighting policies adaptively, and can be jointly trained with target networks without any assumptions or prior knowledge about the dataset. It consists of three key components: Stage-based Searching Strategy (3SM) is adopted to shrink the huge searching space in a complete training process; Duplicate Network Reward (DNR) gives more accurate supervision by removing randomness during the searching process; Full Data Update (FDU) further improves the updating efficiency. Experimental results demonstrate the superiority of weighting policy explored by LAW over standard training pipeline. Compared with baselines, LAW can find a better weighting schedule which achieves much more superior accuracy on both biased CIFAR and ImageNet.

Xiao Jin, Baoyun Peng, Yichao Wu et al.

Distillation-based learning boosts the performance of the miniaturized neural network based on the hypothesis that the representation of a teacher model can be used as structured and relatively weak supervision, and thus would be easily learned by a miniaturized model. However, we find that the representation of a converged heavy model is still a strong constraint for training a small student model, which leads to a high lower bound of congruence loss. In this work, inspired by curriculum learning we consider the knowledge distillation from the perspective of curriculum learning by routing. Instead of supervising the student model with a converged teacher model, we supervised it with some anchor points selected from the route in parameter space that the teacher model passed by, as we called route constrained optimization (RCO). We experimentally demonstrate this simple operation greatly reduces the lower bound of congruence loss for knowledge distillation, hint and mimicking learning. On close-set classification tasks like CIFAR100 and ImageNet, RCO improves knowledge distillation by 2.14% and 1.5% respectively. For the sake of evaluating the generalization, we also test RCO on the open-set face recognition task MegaFace.

Baoyun Peng, Xiao Jin, Jiaheng Liu et al.

Most teacher-student frameworks based on knowledge distillation (KD) depend on a strong congruent constraint on instance level. However, they usually ignore the correlation between multiple instances, which is also valuable for knowledge transfer. In this work, we propose a new framework named correlation congruence for knowledge distillation (CCKD), which transfers not only the instance-level information, but also the correlation between instances. Furthermore, a generalized kernel method based on Taylor series expansion is proposed to better capture the correlation between instances. Empirical experiments and ablation studies on image classification tasks (including CIFAR-100, ImageNet-1K) and metric learning tasks (including ReID and Face Recognition) show that the proposed CCKD substantially outperforms the original KD and achieves state-of-the-art accuracy compared with other SOTA KD-based methods. The CCKD can be easily deployed in the majority of the teacher-student framework such as KD and hint-based learning methods.

Yingcheng Su, Shunfeng Zhou, Yichao Wu et al.

Although deeper and larger neural networks have achieved better performance, the complex network structure and increasing computational cost cannot meet the demands of many resource-constrained applications. Existing methods usually choose to execute or skip an entire specific layer, which can only alter the depth of the network. In this paper, we propose a novel method called Dynamic Multi-path Neural Network (DMNN), which provides more path selection choices in terms of network width and depth during inference. The inference path of the network is determined by a controller, which takes into account both previous state and object category information. The proposed method can be easily incorporated into most modern network architectures. Experimental results on ImageNet and CIFAR-100 demonstrate the superiority of our method on both efficiency and overall classification accuracy. To be specific, DMNN-101 significantly outperforms ResNet-101 with an encouraging 45.1% FLOPs reduction, and DMNN-50 performs comparably to ResNet-101 while saving 42.1% parameters.

Yang Feng, Yichao Wu, Leonard Stefanski

We propose a flexible nonparametric regression method for ultrahigh-dimensional data. As a first step, we propose a fast screening method based on the favored smoothing bandwidth of the marginal local constant regression. Then, an iterative procedure is developed to recover both the important covariates and the regression function. Theoretically, we prove that the favored smoothing bandwidth based screening possesses the model selection consistency property. Simulation studies as well as real data analysis show the competitive performance of the new procedure.