Guoyin Wang

Shengyu Zhang, Linfeng Dong, Xiaoya Li et al.

This paper surveys research works in the quickly advancing field of instruction tuning (IT), which can also be referred to as supervised fine-tuning (SFT)\footnote{In this paper, unless specified otherwise, supervised fine-tuning (SFT) and instruction tuning (IT) are used interchangeably.}, a crucial technique to enhance the capabilities and controllability of large language models (LLMs). Instruction tuning refers to the process of further training LLMs on a dataset consisting of \textsc{(instruction, output)} pairs in a supervised fashion, which bridges the gap between the next-word prediction objective of LLMs and the users' objective of having LLMs adhere to human instructions. In this work, we make a systematic review of the literature, including the general methodology of SFT, the construction of SFT datasets, the training of SFT models, and applications to different modalities, domains and application, along with analysis on aspects that influence the outcome of SFT (e.g., generation of instruction outputs, size of the instruction dataset, etc). We also review the potential pitfalls of SFT along with criticism against it, along with efforts pointing out current deficiencies of existing strategies and suggest some avenues for fruitful research. Project Page: github.com/xiaoya-li/Instruction-Tuning-Survey

Li Liu, Penggang Chen, Xin Li et al.

Aligning users across networks using graph representation learning has been found effective where the alignment is accomplished in a low-dimensional embedding space. Yet, achieving highly precise alignment is still challenging, especially when nodes with long-range connectivity to the labeled anchors are encountered. To alleviate this limitation, we purposefully designed WL-Align which adopts a regularized representation learning framework to learn distinctive node representations. It extends the Weisfeiler-Lehman Isormorphism Test and learns the alignment in alternating phases of "across-network Weisfeiler-Lehman relabeling" and "proximity-preserving representation learning". The across-network Weisfeiler-Lehman relabeling is achieved through iterating the anchor-based label propagation and a similarity-based hashing to exploit the known anchors' connectivity to different nodes in an efficient and robust manner. The representation learning module preserves the second-order proximity within individual networks and is regularized by the across-network Weisfeiler-Lehman hash labels. Extensive experiments on real-world and synthetic datasets have demonstrated that our proposed WL-Align outperforms the state-of-the-art methods, achieving significant performance improvements in the "exact matching" scenario. Data and code of WL-Align are available at https://github.com/ChenPengGang/WLAlignCode.

Rui Li, Guoyin Wang, Jiwei Li · stanford

Despite the promising few-shot ability of large language models (LLMs), the standard paradigm of In-context Learning (ICL) suffers the disadvantages of susceptibility to selected demonstrations and the intricacy to generate these demonstrations. In this paper, we raise the fundamental question that whether human-generated demonstrations are necessary for ICL. To answer this question, we propose self-contemplation prompting strategy (SEC), a paradigm free from human-crafted demonstrations. The key point of SEC is that, instead of using hand-crafted examples as demonstrations in ICL, SEC asks LLMs to first create demonstrations on their own, based on which the final output is generated. SEC is a flexible framework and can be adapted to both the vanilla ICL and the chain-of-thought (CoT), but with greater ease: as the manual-generation process of both examples and rationale can be saved. Extensive experiments in arithmetic reasoning, commonsense reasoning, multi-task language understanding, and code generation benchmarks, show that SEC, which does not require hand-crafted demonstrations, significantly outperforms the zero-shot learning strategy, and achieves comparable results to ICL with hand-crafted demonstrations. This demonstrates that, for many tasks, contemporary LLMs possess a sufficient level of competence to exclusively depend on their own capacity for decision making, removing the need for external training data. Code is available at https://github.com/ruili33/SEC.

Bytedance-Seed-Foundation-Code-Team, Yao Cheng, Jianfeng Chen et al. · bytedance

As the capabilities of code large language models (LLMs) continue to expand, their applications across diverse code intelligence domains are rapidly increasing. However, most existing datasets only evaluate limited application domains. To address this gap, we have developed a comprehensive code evaluation dataset FullStack Bench focusing on full-stack programming, which encompasses a wide range of application domains (e.g., basic programming, data analysis, software engineering, mathematics, and machine learning). Besides, to assess multilingual programming capabilities, in FullStack Bench, we design real-world instructions and corresponding unit test cases from 16 widely-used programming languages to reflect real-world usage scenarios rather than simple translations. Moreover, we also release an effective code sandbox execution tool (i.e., SandboxFusion) supporting various programming languages and packages to evaluate the performance of our FullStack Bench efficiently. Comprehensive experimental results on our FullStack Bench demonstrate the necessity and effectiveness of our FullStack Bench and SandboxFusion.

Chiyu Ma, Shuo Yang, Kexin Huang et al.

We present Future-KL Influenced Policy Optimization (FIPO), a reinforcement learning algorithm designed to overcome reasoning bottlenecks in large language models. While GRPO style training scales effectively, it typically relies on outcome-based rewards (ORM) that distribute a global advantage uniformly across every token in a trajectory. We argue that this coarse-grained credit assignment imposes a performance ceiling by failing to distinguish critical logical pivots from trivial tokens. FIPO addresses this by incorporating discounted future-KL divergence into the policy update, creating a dense advantage formulation that re-weights tokens based on their influence on subsequent trajectory behavior. Empirically, FIPO enables models to break through the length stagnation seen in standard baselines. Evaluated on Qwen2.5-32B, FIPO extends the average chain-of-thought length from roughly 4,000 to over 10,000 tokens and increases AIME 2024 Pass@1 accuracy from 50.0% to a peak of 58.0% (converging at approximately 56.0\%). This outperforms both DeepSeek-R1-Zero-Math-32B (around 47.0%) and o1-mini (approximately 56.0%). Our results suggest that establishing dense advantage formulations is a vital path for evolving ORM-based algorithms to unlock the full reasoning potential of base models. We open-source our training system, built on the verl framework.

Bowen Jin, Hansi Zeng, Guoyin Wang et al.

Semantic identifier (ID) is an important concept in information retrieval that aims to preserve the semantics of objects such as documents and items inside their IDs. Previous studies typically adopt a two-stage pipeline to learn semantic IDs by first procuring embeddings using off-the-shelf text encoders and then deriving IDs based on the embeddings. However, each step introduces potential information loss, and there is usually an inherent mismatch between the distribution of embeddings within the latent space produced by text encoders and the anticipated distribution required for semantic indexing. It is non-trivial to design a method that can learn the document's semantic representations and its hierarchical structure simultaneously, given that semantic IDs are discrete and sequentially structured, and the semantic supervision is deficient. In this paper, we introduce LMIndexer, a self-supervised framework to learn semantic IDs with a generative language model. We tackle the challenge of sequential discrete ID by introducing a semantic indexer capable of generating neural sequential discrete representations with progressive training and contrastive learning. In response to the semantic supervision deficiency, we propose to train the model with a self-supervised document reconstruction objective. We show the high quality of the learned IDs and demonstrate their effectiveness on three tasks including recommendation, product search, and document retrieval on five datasets from various domains. Code is available at https://github.com/PeterGriffinJin/LMIndexer.

Yeon Seonwoo, Guoyin Wang, Changmin Seo et al. · pku

Unsupervised sentence representation learning has progressed through contrastive learning and data augmentation methods such as dropout masking. Despite this progress, sentence encoders are still limited to using only an input sentence when predicting its semantic vector. In this work, we show that the semantic meaning of a sentence is also determined by nearest-neighbor sentences that are similar to the input sentence. Based on this finding, we propose a novel unsupervised sentence encoder, RankEncoder. RankEncoder predicts the semantic vector of an input sentence by leveraging its relationship with other sentences in an external corpus, as well as the input sentence itself. We evaluate RankEncoder on semantic textual benchmark datasets. From the experimental results, we verify that 1) RankEncoder achieves 80.07% Spearman's correlation, a 1.1% absolute improvement compared to the previous state-of-the-art performance, 2) RankEncoder is universally applicable to existing unsupervised sentence embedding methods, and 3) RankEncoder is specifically effective for predicting the similarity scores of similar sentence pairs.

Shuyin Xia, Xiaoyu Lian, Guoyin Wang et al.

GBSVM (Granular-ball Support Vector Machine) is a significant attempt to construct a classifier using the coarse-to-fine granularity of a granular-ball as input, rather than a single data point. It is the first classifier whose input contains no points. However, the existing model has some errors, and its dual model has not been derived. As a result, the current algorithm cannot be implemented or applied. To address these problems, this paper has fixed the errors of the original model of the existing GBSVM, and derived its dual model. Furthermore, a particle swarm optimization algorithm is designed to solve the dual model. The sequential minimal optimization algorithm is also carefully designed to solve the dual model. The solution is faster and more stable than the particle swarm optimization based version. The experimental results on the UCI benchmark datasets demonstrate that GBSVM has good robustness and efficiency. All codes have been released in the open source library at http://www.cquptshuyinxia.com/GBSVM.html or https://github.com/syxiaa/GBSVM.

Ji Xu, Gang Ren, Yao Xiao et al.

The centrality and diversity of the labeled data are very influential to the performance of semi-supervised learning (SSL), but most SSL models select the labeled data randomly. This study first construct a leading forest that forms a partially ordered topological space in an unsupervised way, and select a group of most representative samples to label with one shot (differs from active learning essentially) using property of homeomorphism. Then a kernelized large margin metric is efficiently learned for the selected data to classify the remaining unlabeled sample. Optimal leading forest (OLF) has been observed to have the advantage of revealing the difference evolution along a path within a subtree. Therefore, we formulate an optimization problem based on OLF to select the samples. Also with OLF, the multiple local metrics learning is facilitated to address multi-modal and mix-modal problem in SSL, especially when the number of class is large. Attribute to this novel design, stableness and accuracy of the performance is significantly improved when compared with the state-of-the-art graph SSL methods. The extensive experimental studies have shown that the proposed method achieved encouraging accuracy and efficiency. Code has been made available at https://github.com/alanxuji/DeLaLA.

Shuhe Wang, Xiaofei Sun, Xiaoya Li et al.

Despite the fact that large-scale Language Models (LLM) have achieved SOTA performances on a variety of NLP tasks, its performance on NER is still significantly below supervised baselines. This is due to the gap between the two tasks the NER and LLMs: the former is a sequence labeling task in nature while the latter is a text-generation model. In this paper, we propose GPT-NER to resolve this issue. GPT-NER bridges the gap by transforming the sequence labeling task to a generation task that can be easily adapted by LLMs e.g., the task of finding location entities in the input text "Columbus is a city" is transformed to generate the text sequence "@@Columbus## is a city", where special tokens @@## marks the entity to extract. To efficiently address the "hallucination" issue of LLMs, where LLMs have a strong inclination to over-confidently label NULL inputs as entities, we propose a self-verification strategy by prompting LLMs to ask itself whether the extracted entities belong to a labeled entity tag. We conduct experiments on five widely adopted NER datasets, and GPT-NER achieves comparable performances to fully supervised baselines, which is the first time as far as we are concerned. More importantly, we find that GPT-NER exhibits a greater ability in the low-resource and few-shot setups, when the amount of training data is extremely scarce, GPT-NER performs significantly better than supervised models. This demonstrates the capabilities of GPT-NER in real-world NER applications where the number of labeled examples is limited.

Yifan Song, Guoyin Wang, Sujian Li et al. · pku

Current evaluations of large language models (LLMs) often overlook non-determinism, typically focusing on a single output per example. This limits our understanding of LLM performance variability in real-world applications. Our study addresses this issue by exploring key questions about the performance differences between greedy decoding and sampling, identifying benchmarks' consistency regarding non-determinism, and examining unique model behaviors. Through extensive experiments, we observe that greedy decoding generally outperforms sampling methods for most evaluated tasks. We also observe consistent performance across different LLM sizes and alignment methods, noting that alignment can reduce sampling variance. Moreover, our best-of-N sampling approach demonstrates that smaller LLMs can match or surpass larger models such as GPT-4-Turbo, highlighting the untapped potential of smaller LLMs. This research shows the importance of considering non-determinism in LLM evaluations and provides insights for future LLM development and evaluation.

Guanlin Li, Yifei Chen, Jie Zhang et al.

AI-Generated Content (AIGC) is rapidly expanding, with services using advanced generative models to create realistic images and fluent text. Regulating such content is crucial to prevent policy violations, such as unauthorized commercialization or unsafe content distribution. Watermarking is a promising solution for content attribution and verification, but we demonstrate its vulnerability to two key attacks: (1) Watermark removal, where adversaries erase embedded marks to evade regulation, and (2) Watermark forging, where they generate illicit content with forged watermarks, leading to misattribution. We propose Warfare, a unified attack framework leveraging a pre-trained diffusion model for content processing and a generative adversarial network for watermark manipulation. Evaluations across datasets and embedding setups show that Warfare achieves high success rates while preserving content quality. We further introduce Warfare-Plus, which enhances efficiency without compromising effectiveness. The code can be found in https://github.com/GuanlinLee/warfare.

Dawei Dai, Yuanhui Zhang, Long Xu et al.

The previous advancements in pathology image understanding primarily involved developing models tailored to specific tasks. Recent studies has demonstrated that the large vision-language model can enhance the performance of various downstream tasks in medical image understanding. In this study, we developed a domain-specific large language-vision assistant (PA-LLaVA) for pathology image understanding. Specifically, (1) we first construct a human pathology image-text dataset by cleaning the public medical image-text data for domain-specific alignment; (2) Using the proposed image-text data, we first train a pathology language-image pretraining (PLIP) model as the specialized visual encoder for pathology image, and then we developed scale-invariant connector to avoid the information loss caused by image scaling; (3) We adopt two-stage learning to train PA-LLaVA, first stage for domain alignment, and second stage for end to end visual question \& answering (VQA) task. In experiments, we evaluate our PA-LLaVA on both supervised and zero-shot VQA datasets, our model achieved the best overall performance among multimodal models of similar scale. The ablation experiments also confirmed the effectiveness of our design. We posit that our PA-LLaVA model and the datasets presented in this work can promote research in field of computational pathology. All codes are available at: https://github.com/ddw2AIGROUP2CQUPT/PA-LLaVA}{https://github.com/ddw2AIGROUP2CQUPT/PA-LLaVA

Dawei Dai, YuTang Li, YingGe Liu et al.

Currently, image-text-driven multi-modal deep learning models have demonstrated their outstanding potential in many fields. In practice, tasks centered around facial images have broad application prospects. This paper presents \textbf{FaceCaption-15M}, a large-scale, diverse, and high-quality dataset of facial images accompanied by their natural language descriptions (facial image-to-text). This dataset aims to facilitate a study on face-centered tasks. FaceCaption-15M comprises over 15 million pairs of facial images and their corresponding natural language descriptions of facial features, making it the largest facial image-caption dataset to date. We conducted a comprehensive analysis of image quality, text naturalness, text complexity, and text-image relevance to demonstrate the superiority of FaceCaption-15M. To validate the effectiveness of FaceCaption-15M, we first trained a facial language-image pre-training model (FLIP, similar to CLIP) to align facial image with its corresponding captions in feature space. Subsequently, using both image and text encoders and fine-tuning only the linear layer, our FLIP-based models achieved state-of-the-art results on two challenging face-centered tasks. The purpose is to promote research in the field of face-related tasks through the availability of the proposed FaceCaption-15M dataset. All data, codes, and models are publicly available. https://huggingface.co/datasets/OpenFace-CQUPT/FaceCaption-15M

Jiang Xie, Shuyin Xia, Guoyin Wang et al.

Most of the existing clustering methods are based on a single granularity of information, such as the distance and density of each data. This most fine-grained based approach is usually inefficient and susceptible to noise. Therefore, we propose a clustering algorithm that combines multi-granularity Granular-Ball and minimum spanning tree (MST). We construct coarsegrained granular-balls, and then use granular-balls and MST to implement the clustering method based on "large-scale priority", which can greatly avoid the influence of outliers and accelerate the construction process of MST. Experimental results on several data sets demonstrate the power of the algorithm. All codes have been released at https://github.com/xjnine/GBMST.

Dhanasekar Sundararaman, Vivek Subramanian, Guoyin Wang et al.

Numbers are essential components of text, like any other word tokens, from which natural language processing (NLP) models are built and deployed. Though numbers are typically not accounted for distinctly in most NLP tasks, there is still an underlying amount of numeracy already exhibited by NLP models. In this work, we attempt to tap this potential of state-of-the-art NLP models and transfer their ability to boost performance in related tasks. Our proposed classification of numbers into entities helps NLP models perform well on several tasks, including a handcrafted Fill-In-The-Blank (FITB) task and on question answering using joint embeddings, outperforming the BERT and RoBERTa baseline classification.

Xiaofei Sun, Linfeng Dong, Xiaoya Li et al.

Despite the success of ChatGPT, its performances on most NLP tasks are still well below the supervised baselines. In this work, we looked into the causes, and discovered that its subpar performance was caused by the following factors: (1) token limit in the prompt does not allow for the full utilization of the supervised datasets; (2) mismatch between the generation nature of ChatGPT and NLP tasks; (3) intrinsic pitfalls of LLMs models, e.g., hallucination, overly focus on certain keywords, etc. In this work, we propose a collection of general modules to address these issues, in an attempt to push the limits of ChatGPT on NLP tasks. Our proposed modules include (1) a one-input-multiple-prompts strategy that employs multiple prompts for one input to accommodate more demonstrations; (2) using fine-tuned models for better demonstration retrieval; (3) transforming tasks to formats that are more tailored to the generation nature; (4) employing reasoning strategies that are tailored to addressing the task-specific complexity; (5) the self-verification strategy to address the hallucination issue of LLMs; (6) the paraphrase strategy to improve the robustness of model predictions. We conduct experiments on 21 datasets of 10 representative NLP tasks, including question answering, commonsense reasoning, natural language inference, sentiment analysis, named entity recognition, entity-relation extraction, event extraction, dependency parsing, semantic role labeling, and part-of-speech tagging. Using the proposed assemble of techniques, we are able to significantly boost the performance of ChatGPT on the selected NLP tasks, achieving performances comparable to or better than supervised baselines, or even existing SOTA performances.

Shuhe Wang, Yuxian Meng, Rongbin Ouyang et al.

To better handle long-tail cases in the sequence labeling (SL) task, in this work, we introduce graph neural networks sequence labeling (GNN-SL), which augments the vanilla SL model output with similar tagging examples retrieved from the whole training set. Since not all the retrieved tagging examples benefit the model prediction, we construct a heterogeneous graph, and leverage graph neural networks (GNNs) to transfer information between the retrieved tagging examples and the input word sequence. The augmented node which aggregates information from neighbors is used to do prediction. This strategy enables the model to directly acquire similar tagging examples and improves the general quality of predictions. We conduct a variety of experiments on three typical sequence labeling tasks: Named Entity Recognition (NER), Part of Speech Tagging (POS), and Chinese Word Segmentation (CWS) to show the significant performance of our GNN-SL. Notably, GNN-SL achieves SOTA results of 96.9 (+0.2) on PKU, 98.3 (+0.4) on CITYU, 98.5 (+0.2) on MSR, and 96.9 (+0.2) on AS for the CWS task, and results comparable to SOTA performances on NER datasets, and POS datasets.

Jinye Yang, Ji Xu, Di Wu et al.

Long-tailed (LT) classification is an unavoidable and challenging problem in the real world. Most existing long-tailed classification methods focus only on solving the class-wise imbalance while ignoring the attribute-wise imbalance. The deviation of a classification model is caused by both class-wise and attribute-wise imbalance. Due to the fact that attributes are implicit in most datasets and the combination of attributes is complex, attribute-wise imbalance is more difficult to handle. For this purpose, we proposed a novel long-tailed classification framework, aiming to build a multi-granularity classification model by means of invariant feature learning. This method first unsupervisedly constructs Coarse-Grained forest (CLF) to better characterize the distribution of attributes within a class. Depending on the distribution of attributes, one can customize suitable sampling strategies to construct different imbalanced datasets. We then introduce multi-center loss (MCL) that aims to gradually eliminate confusing attributes during feature learning process. The proposed framework does not necessarily couple to a specific LT classification model structure and can be integrated with any existing LT method as an independent component. Extensive experiments show that our approach achieves state-of-the-art performance on both existing benchmarks ImageNet-GLT and MSCOCO-GLT and can improve the performance of existing LT methods. Our codes are available on GitHub: \url{https://github.com/jinyery/cognisance}

Haoming Meng, Kexin Huang, Shaohang Wei et al. · pku

Reinforcement learning with verifiable rewards (RLVR) has significantly improved reasoning in large language models (LLMs), yet the token-level mechanisms underlying these improvements remain unclear. We present a systematic empirical study of RLVR's distributional effects organized around three main analyses: (1) token-level characterization of distributional shifts between base and RL models, (2) the impact of token-level distributional shifts on sequence-level reasoning performance through cross-sampling interventions, and (3) fine-grained mechanics of these shifts at the token level. We find that RL fine-tuning induces highly sparse and targeted changes, with only a small fraction of token distributions exhibiting meaningful divergence between the base and RL policies. We further characterize the structure and evolution of these shifts through analyses of token entropy, positional concentration, and reallocation of probability mass. To assess the functional importance of these sparse changes, we conduct cross-sampling experiments that selectively swap token choices between the base and RL models with varying intervention budgets. We show that inserting only a small fraction of RL-sampled tokens into base generations progressively recovers RL performance gains, while injecting a similarly small number of base token choices into otherwise RL-generated sequences collapses performance to base levels, isolating a small set of token-level decisions directly responsible for RLVR's performance gains. Finally, we explore divergence-weighted variants of the advantage signal as a diagnostic intervention, finding that they can yield improvements over baselines. Together, our results shed light on the distributional changes induced by RLVR and provide a fine-grained, token-level lens for understanding RLVR fine-tuning as a targeted refinement process.

Minsik Oh, Joosung Lee, Jiwei Li et al.

Identifying relevant persona or knowledge for conversational systems is critical to grounded dialogue response generation. However, each grounding has been mostly researched in isolation with more practical multi-context dialogue tasks introduced in recent works. We define Persona and Knowledge Dual Context Identification as the task to identify persona and knowledge jointly for a given dialogue, which could be of elevated importance in complex multi-context dialogue settings. We develop a novel grounding retrieval method that utilizes all contexts of dialogue simultaneously. Our method requires less computational power via utilizing neural QA retrieval models. We further introduce our novel null-positive rank test which measures ranking performance on semantically dissimilar samples (i.e. hard negatives) in relation to data augmentation.

Ke Bai, Guoyin Wang, Jiwei Li et al.

Open world classification is a task in natural language processing with key practical relevance and impact. Since the open or {\em unknown} category data only manifests in the inference phase, finding a model with a suitable decision boundary accommodating for the identification of known classes and discrimination of the open category is challenging. The performance of existing models is limited by the lack of effective open category data during the training stage or the lack of a good mechanism to learn appropriate decision boundaries. We propose an approach based on \underline{a}daptive \underline{n}egative \underline{s}amples (ANS) designed to generate effective synthetic open category samples in the training stage and without requiring any prior knowledge or external datasets. Empirically, we find a significant advantage in using auxiliary one-versus-rest binary classifiers, which effectively utilize the generated negative samples and avoid the complex threshold-seeking stage in previous works. Extensive experiments on three benchmark datasets show that ANS achieves significant improvements over state-of-the-art methods.

Guan Wang, Shuyin Xia, Lei Qian et al.

Graph Convolutional Network (GCN) is a model that can effectively handle graph data tasks and has been successfully applied. However, for large-scale graph datasets, GCN still faces the challenge of high computational overhead, especially when the number of convolutional layers in the graph is large. Currently, there are many advanced methods that use various sampling techniques or graph coarsening techniques to alleviate the inconvenience caused during training. However, among these methods, some ignore the multi-granularity information in the graph structure, and the time complexity of some coarsening methods is still relatively high. In response to these issues, based on our previous work, in this paper, we propose a new framework called Efficient and Scalable Granular-ball Graph Coarsening Method for Large-scale Graph Node Classification. Specifically, this method first uses a multi-granularity granular-ball graph coarsening algorithm to coarsen the original graph to obtain many subgraphs. The time complexity of this stage is linear and much lower than that of the exiting graph coarsening methods. Then, subgraphs composed of these granular-balls are randomly sampled to form minibatches for training GCN. Our algorithm can adaptively and significantly reduce the scale of the original graph, thereby enhancing the training efficiency and scalability of GCN. Ultimately, the experimental results of node classification on multiple datasets demonstrate that the method proposed in this paper exhibits superior performance. The code is available at https://anonymous.4open.science/r/1-141D/.

Ye Wang, Jingbo Liao, Hong Yu et al.

Currently end-to-end deep learning based open-domain dialogue systems remain black box models, making it easy to generate irrelevant contents with data-driven models. Specifically, latent variables are highly entangled with different semantics in the latent space due to the lack of priori knowledge to guide the training. To address this problem, this paper proposes to harness the generative model with a priori knowledge through a cognitive approach involving mesoscopic scale feature disentanglement. Particularly, the model integrates the macro-level guided-category knowledge and micro-level open-domain dialogue data for the training, leveraging the priori knowledge into the latent space, which enables the model to disentangle the latent variables within the mesoscopic scale. Besides, we propose a new metric for open-domain dialogues, which can objectively evaluate the interpretability of the latent space distribution. Finally, we validate our model on different datasets and experimentally demonstrate that our model is able to generate higher quality and more interpretable dialogues than other models.

Xiaofei Sun, Xiaoya Li, Shengyu Zhang et al.

A standard paradigm for sentiment analysis is to rely on a singular LLM and makes the decision in a single round under the framework of in-context learning. This framework suffers the key disadvantage that the single-turn output generated by a single LLM might not deliver the perfect decision, just as humans sometimes need multiple attempts to get things right. This is especially true for the task of sentiment analysis where deep reasoning is required to address the complex linguistic phenomenon (e.g., clause composition, irony, etc) in the input. To address this issue, this paper introduces a multi-LLM negotiation framework for sentiment analysis. The framework consists of a reasoning-infused generator to provide decision along with rationale, a explanation-deriving discriminator to evaluate the credibility of the generator. The generator and the discriminator iterate until a consensus is reached. The proposed framework naturally addressed the aforementioned challenge, as we are able to take the complementary abilities of two LLMs, have them use rationale to persuade each other for correction. Experiments on a wide range of sentiment analysis benchmarks (SST-2, Movie Review, Twitter, yelp, amazon, IMDB) demonstrate the effectiveness of proposed approach: it consistently yields better performances than the ICL baseline across all benchmarks, and even superior performances to supervised baselines on the Twitter and movie review datasets.

Shuyin Xia, Guoyin Wang, Xinbo Gao et al.

Human cognition operates on a "Global-first" cognitive mechanism, prioritizing information processing based on coarse-grained details. This mechanism inherently possesses an adaptive multi-granularity description capacity, resulting in computational traits such as efficiency, robustness, and interpretability. The analysis pattern reliance on the finest granularity and single-granularity makes most existing computational methods less efficient, robust, and interpretable, which is an important reason for the current lack of interpretability in neural networks. Multi-granularity granular-ball computing employs granular-balls of varying sizes to daptively represent and envelop the sample space, facilitating learning based on these granular-balls. Given that the number of coarse-grained "granular-balls" is fewer than sample points, granular-ball computing proves more efficient. Moreover, the inherent coarse-grained nature of granular-balls reduces susceptibility to fine-grained sample disturbances, enhancing robustness. The multi-granularity construct of granular-balls generates topological structures and coarse-grained descriptions, naturally augmenting interpretability. Granular-ball computing has successfully ventured into diverse AI domains, fostering the development of innovative theoretical methods, including granular-ball classifiers, clustering techniques, neural networks, rough sets, and evolutionary computing. This has notably ameliorated the efficiency, noise robustness, and interpretability of traditional methods. Overall, granular-ball computing is a rare and innovative theoretical approach in AI that can adaptively and simultaneously enhance efficiency, robustness, and interpretability. This article delves into the main application landscapes for granular-ball computing, aiming to equip future researchers with references and insights to refine and expand this promising theory.

Dawei Dai, Chunjie Wang, Shuyin Xia et al.

Edge detection, a basic task in the field of computer vision, is an important preprocessing operation for the recognition and understanding of a visual scene. In conventional models, the edge image generated is ambiguous, and the edge lines are also very thick, which typically necessitates the use of non-maximum suppression (NMS) and morphological thinning operations to generate clear and thin edge images. In this paper, we aim to propose a one-stage neural network model that can generate high-quality edge images without postprocessing. The proposed model adopts a classic encoder-decoder framework in which a pre-trained neural model is used as the encoder and a multi-feature-fusion mechanism that merges the features of each level with each other functions as a learnable decoder. Further, we propose a new loss function that addresses the pixel-level imbalance in the edge image by suppressing the false positive (FP) edge information near the true positive (TP) edge and the false negative (FN) non-edge. The results of experiments conducted on several benchmark datasets indicate that the proposed method achieves state-of-the-art results without using NMS and morphological thinning operations.

Dawei Dai, Yutang Li, Liang Wang et al.

In some specific scenarios, face sketch was used to identify a person. However, drawing a complete face sketch often needs skills and takes time, which hinder its widespread applicability in the practice. In this study, we proposed a new task named sketch less face image retrieval (SLFIR), in which the retrieval was carried out at each stroke and aim to retrieve the target face photo using a partial sketch with as few strokes as possible (see Fig.1). Firstly, we developed a method to generate the data of sketch with drawing process, and opened such dataset; Secondly, we proposed a two-stage method as the baseline for SLFIR that (1) A triplet network, was first adopt to learn the joint embedding space shared between the complete sketch and its target face photo; (2) Regarding the sketch drawing episode as a sequence, we designed a LSTM module to optimize the representation of the incomplete face sketch. Experiments indicate that the new framework can finish the retrieval using a partial or pool drawing sketch.

Shuyin Xia, Jiang Xie, Guoyin Wang

Existing clustering methods are based on a single granularity of information, such as the distance and density of each data. This most fine-grained based approach is usually inefficient and susceptible to noise. Inspired by adaptive process of granular-ball division and differentiation, we present a novel clustering approach that retains the speed and efficiency of K-means clustering while out-performing time-tested density clustering approaches widely used in industry today. Our simple, robust, adaptive granular-ball clustering method can efficiently recognize clusters with unknown and complex shapes without the use of extra parameters. Moreover, the proposed method provides an efficient, adaptive way to depict the world, and will promote the research and development of adaptive and efficient AI technologies, especially density computing models, and improve the efficiency of many existing clustering methods.

Jie Yang, Lingyun Xiaodiao, Guoyin Wang et al.

The granular-ball (GB)-based classifier introduced by Xia, exhibits adaptability in creating coarse-grained information granules for input, thereby enhancing its generality and flexibility. Nevertheless, the current GB-based classifiers rigidly assign a specific class label to each data instance and lacks of the necessary strategies to address uncertain instances. These far-fetched certain classification approachs toward uncertain instances may suffer considerable risks. To solve this problem, we construct a robust three-way classifier with shadowed GBs for uncertain data. Firstly, combine with information entropy, we propose an enhanced GB generation method with the principle of justifiable granularity. Subsequently, based on minimum uncertainty, a shadowed mapping is utilized to partition a GB into Core region, Important region and Unessential region. Based on the constructed shadowed GBs, we establish a three-way classifier to categorize data instances into certain classes and uncertain case. Finally, extensive comparative experiments are conducted with 2 three-way classifiers, 3 state-of-the-art GB-based classifiers, and 3 classical machine learning classifiers on 12 public benchmark datasets. The results show that our model demonstrates robustness in managing uncertain data and effectively mitigates classification risks. Furthermore, our model almost outperforms the other comparison methods in both effectiveness and efficiency.

Ye Wang, Zixuan Wu, Lifeng Shen et al.

Imbalanced data distribution remains a critical challenge in sequential learning, leading models to easily recognize frequent categories while failing to detect minority classes adequately. The Mixture-of-Experts model offers a scalable solution, yet its application is often hindered by parameter inefficiency, poor expert specialization, and difficulty in resolving prediction conflicts. To Master the Minority classes effectively, we propose the Uncertainty-based Multi-Expert fusion network (UME) framework. UME is designed with three core innovations: First, we employ Ensemble LoRA for parameter-efficient modeling, significantly reducing the trainable parameter count. Second, we introduce Sequential Specialization guided by Dempster-Shafer Theory (DST), which ensures effective specialization on the challenging-tailed classes. Finally, an Uncertainty-Guided Fusion mechanism uses DST's certainty measures to dynamically weigh expert opinions, resolving conflicts by prioritizing the most confident expert for reliable final predictions. Extensive experiments across four public hierarchical text classification datasets demonstrate that UME achieves state-of-the-art performance. We achieve a performance gain of up to 17.97\% over the best baseline on individual categories, while reducing trainable parameters by up to 10.32\%. The findings highlight that uncertainty-guided expert coordination is a principled strategy for addressing challenging-tailed sequence learning. Our code is available at https://github.com/CQUPTWZX/Multi-experts.

Jinda Lu, Junkang Wu, Jinghan Li et al.

Extending Reinforcement Learning with Verifiable Rewards (RLVR) to multimodal large language models (MLLMs) faces a fundamental challenge: their responses inherently interleave perception-related tokens, which ground visual content, with reasoning-related tokens, which construct reasoning chains. These token types instantiate distinct yet interdependent capacities -- visual grounding and symbolic reasoning -- making isolated optimization insufficient. Through token-level empirical analysis, we demonstrate that optimizing either perception- or reasoning-only tokens consistently underperforms full optimization, underscoring their inherent coupling. To address this, we propose a plug-and-play Token-Reweighting (ToR) strategy that explicitly models this interdependence by identifying critical tokens of both types and dynamically reweighting them during RLVR training. Applied on top of existing methods (e.g., GRPO and DAPO), ToR delivers consistent performance gains across multiple multi-modal reasoning benchmarks, achieving state-of-the-art performance with both accurate visual grounding and coherent reasoning.

Shuyin Xia, Xinyu Lin, Guan Wang et al.

Optimization problems aim to find the optimal solution, which is becoming increasingly complex and difficult to solve. Traditional evolutionary optimization methods always overlook the granular characteristics of solution space. In the real scenario of numerous optimizations, the solution space is typically partitioned into sub-regions characterized by varying degree distributions. These sub-regions present different granularity characteristics at search potential and difficulty. Considering the granular characteristics of the solution space, the number of coarse-grained regions is smaller than the number of points, so the calculation is more efficient. On the other hand, coarse-grained characteristics are not easily affected by fine-grained sample points, so the calculation is more robust. To this end, this paper proposes a new multi-granularity evolutionary optimization method, namely the Granular-ball Optimization (GBO) algorithm, which characterizes and searches the solution space from coarse to fine. Specifically, using granular-balls instead of traditional points for optimization increases the diversity and robustness of the random search process. At the same time, the search range in different iteration processes is limited by the radius of granular-balls, covering the solution space from large to small. The mechanism of granular-ball splitting is applied to continuously split and evolve the large granular-balls into smaller ones for refining the solution space. Extensive experiments on commonly used benchmarks have shown that GBO outperforms popular and advanced evolutionary algorithms. The code can be found in the supporting materials.

Shuyin Xia, Meng Yang, Dawei Dai et al.

Superpixels provide a compact region-based representation that preserves object boundaries and local structures, and have therefore been widely used in a variety of vision tasks to reduce computational cost. However, most existing superpixel algorithms produce irregularly shaped regions, which are not well aligned with regular operators such as convolutions. Consequently, superpixels are often treated as an offline preprocessing step, limiting parallel implementation and hindering end-to-end optimization within deep learning pipelines. Motivated by the adaptive representation and coverage property of granular-ball computing, we develop a square superpixel generation approach. Specifically, we approximate superpixels using multi-scale square blocks to avoid the computational and implementation difficulties induced by irregular shapes, enabling efficient parallel processing and learnable feature extraction. For each block, a purity score is computed based on pixel-intensity similarity, and high-quality blocks are selected accordingly. The resulting square superpixels can be readily integrated as graph nodes in graph neural networks (GNNs) or as tokens in Vision Transformers (ViTs), facilitating multi-scale information aggregation and structured visual representation. Experimental results on downstream tasks demonstrate consistent performance improvements, validating the effectiveness of the proposed method.

Jiang Xie, Pengfei Zhao, Shuyin Xia et al.

It is crucial to evaluate the quality and determine the optimal number of clusters in cluster analysis. In this paper, the multi-granularity characterization of the data set is carried out to obtain the hyper-balls. The cluster internal evaluation index based on hyper-balls(HCVI) is defined. Moreover, a general method for determining the optimal number of clusters based on HCVI is proposed. The proposed methods can evaluate the clustering results produced by the several classic methods and determine the optimal cluster number for data sets containing noises and clusters with arbitrary shapes. The experimental results on synthetic and real data sets indicate that the new index outperforms existing ones.

Hongyu Shen, Jinoh Oh, Shuai Zhao et al.

Generating representations that precisely reflect customers' behavior is an important task for providing personalized skill routing experience in Alexa. Currently, Dynamic Routing (DR) team, which is responsible for routing Alexa traffic to providers or skills, relies on two features to be served as personal signals: absolute traffic count and normalized traffic count of every skill usage per customer. Neither of them considers the network based structure for interactions between customers and skills, which contain richer information for customer preferences. In this work, we first build a heterogeneous edge attributed graph based customers' past interactions with the invoked skills, in which the user requests (utterances) are modeled as edges. Then we propose a graph convolutional network(GCN) based model, namely Personalized Dynamic Routing Feature Encoder(PDRFE), that generates personalized customer representations learned from the built graph. Compared with existing models, PDRFE is able to further capture contextual information in the graph convolutional function. The performance of our proposed model is evaluated by a downstream task, defect prediction, that predicts the defect label from the learned embeddings of customers and their triggered skills. We observe up to 41% improvements on the cross entropy metric for our proposed models compared to the baselines.

Kexin Huang, Haoming Meng, Junkang Wu et al.

Reinforcement learning with verifiable rewards (RLVR) has substantially improved the reasoning capabilities of large language models. While existing analyses identify that RLVR-induced changes are sparse, they primarily focus on the \textbf{magnitude} of these updates, largely overlooking their \textbf{direction}. In this work, we argue that the direction of updates is a more critical lens for understanding RLVR's effects, which can be captured by the signed, token-level log probability difference $Δ\log p$ between the base and final RLVR models. Through statistical analysis and token-replacement interventions, we demonstrate that $Δ\log p$ more effectively identifies sparse, yet reasoning-critical updates than magnitude-based metrics (\eg divergence or entropy). Building on this insight, we propose two practical applications: (1) a \textit{test-time extrapolation} method that amplifies the policy along the learned $Δ\log p$ direction to improve reasoning accuracy without further training; (2) a \textit{training-time reweighting} method that focuses learning on low-probability (corresponding to higher $Δ\log p$) tokens, which improves reasoning performance across models and benchmarks. Our work establishes the direction of change as a key principle for analyzing and improving RLVR.

Xueyu Hu, Ziyu Zhao, Shuang Wei et al.

In this paper, we introduce InfiAgent-DABench, the first benchmark specifically designed to evaluate LLM-based agents on data analysis tasks. These tasks require agents to end-to-end solving complex tasks by interacting with an execution environment. This benchmark contains DAEval, a dataset consisting of 257 data analysis questions derived from 52 CSV files, and an agent framework which incorporates LLMs to serve as data analysis agents for both serving and evaluation. Since data analysis questions are often open-ended and hard to evaluate without human supervision, we adopt a format-prompting technique to convert each question into a closed-form format so that they can be automatically evaluated. Our extensive benchmarking of 34 LLMs uncovers the current challenges encountered in data analysis tasks. In addition, building on top of our agent framework, we develop a specialized agent, DAAgent, which surpasses GPT-3.5 by 3.9% on DABench. Evaluation datasets and toolkits for InfiAgent-DABench are released at https://github.com/InfiAgent/InfiAgent .

Kaiyuan Chen, Qimin Wu, Taiyu Hou et al.

The capacity of AI agents to effectively handle tasks of increasing duration and complexity continues to grow, demonstrating exceptional performance in coding, deep research, and complex problem-solving evaluations. However, in daily scenarios, the perception of these advanced AI capabilities among general users remains limited. We argue that current evaluations prioritize increasing task difficulty without sufficiently addressing the diversity of agentic tasks necessary to cover the daily work, life, and learning activities of a broad demographic. To address this, we propose AgentIF-OneDay, aimed at determining whether general users can utilize natural language instructions and AI agents to complete a diverse array of daily tasks. These tasks require not only solving problems through dialogue but also understanding various attachment types and delivering tangible file-based results. The benchmark is structured around three user-centric categories: Open Workflow Execution, which assesses adherence to explicit and complex workflows; Latent Instruction, which requires agents to infer implicit instructions from attachments; and Iterative Refinement, which involves modifying or expanding upon ongoing work. We employ instance-level rubrics and a refined evaluation pipeline that aligns LLM-based verification with human judgment, achieving an 80.1% agreement rate using Gemini-3-Pro. AgentIF-OneDay comprises 104 tasks covering 767 scoring points. We benchmarked four leading general AI agents and found that agent products built based on APIs and ChatGPT agents based on agent RL remain in the first tier simultaneously. Leading LLM APIs and open-source models have internalized agentic capabilities, enabling AI application teams to develop cutting-edge Agent products.

Jiang Xie, Qiao Deng, Shuyin Xia et al.

In recent years, the problem of fuzzy clustering has been widely concerned. The membership iteration of existing methods is mostly considered globally, which has considerable problems in noisy environments, and iterative calculations for clusters with a large number of different sample sizes are not accurate and efficient. In this paper, starting from the strategy of large-scale priority, the data is fuzzy iterated using granular-balls, and the membership degree of data only considers the two granular-balls where it is located, thus improving the efficiency of iteration. The formed fuzzy granular-balls set can use more processing methods in the face of different data scenarios, which enhances the practicability of fuzzy clustering calculations.

Shuyin Xia, Xiaoyu Lian, Guoyin Wang et al.

Most existing fuzzy set methods use points as their input, which is the finest granularity from the perspective of granular computing. Consequently, these methods are neither efficient nor robust to label noise. Therefore, we propose a frame-work called granular-ball fuzzy set by introducing granular-ball computing into fuzzy set. The computational framework is based on the granular-balls input rather than points; therefore, it is more efficient and robust than traditional fuzzy methods, and can be used in various fields of fuzzy data processing according to its extensibility. Furthermore, the framework is extended to the classifier fuzzy support vector machine (FSVM), to derive the granular ball fuzzy SVM (GBFSVM). The experimental results demonstrate the effectiveness and efficiency of GBFSVM.

Shuhe Wang, Shengyu Zhang, Jie Zhang et al.

Reinforcement learning (RL) enhanced large language models (LLMs), particularly exemplified by DeepSeek-R1, have exhibited outstanding performance. Despite the effectiveness in improving LLM capabilities, its implementation remains highly complex, requiring complex algorithms, reward modeling strategies, and optimization techniques. This complexity poses challenges for researchers and practitioners in developing a systematic understanding of RL-enhanced LLMs. Moreover, the absence of a comprehensive survey summarizing existing research on RL-enhanced LLMs has limited progress in this domain, hindering further advancements. In this work, we are going to make a systematic review of the most up-to-date state of knowledge on RL-enhanced LLMs, attempting to consolidate and analyze the rapidly growing research in this field, helping researchers understand the current challenges and advancements. Specifically, we (1) detail the basics of RL; (2) introduce popular RL-enhanced LLMs; (3) review researches on two widely-used reward model-based RL techniques: Reinforcement Learning from Human Feedback (RLHF) and Reinforcement Learning from AI Feedback (RLAIF); and (4) explore Direct Preference Optimization (DPO), a set of methods that bypass the reward model to directly use human preference data for aligning LLM outputs with human expectations. We will also point out current challenges and deficiencies of existing methods and suggest some avenues for further improvements. Project page of this work can be found at https://github.com/ShuheWang1998/Reinforcement-Learning-Enhanced-LLMs-A-Survey.

Dawei Dai, Hao Zhu, Shuyin Xia et al.

In actual scenarios, whether manually or automatically annotated, label noise is inevitably generated in the training data, which can affect the effectiveness of deep CNN models. The popular solutions require data cleaning or designing additional optimizations to punish the data with mislabeled data, thereby enhancing the robustness of models. However, these methods come at the cost of weakening or even losing some data during the training process. As we know, content is the inherent attribute of an image that does not change with changes in annotations. In this study, we propose a general granular-ball computing (GBC) module that can be embedded into a CNN model, where the classifier finally predicts the label of granular-ball ($gb$) samples instead of each individual samples. Specifically, considering the classification task: (1) in forward process, we split the input samples as $gb$ samples at feature-level, each of which can correspond to multiple samples with varying numbers and share one single label; (2) during the backpropagation process, we modify the gradient allocation strategy of the GBC module to enable it to propagate normally; and (3) we develop an experience replay policy to ensure the stability of the training process. Experiments demonstrate that the proposed method can improve the robustness of CNN models with no additional data or optimization.

Xueyu Hu, Tao Xiong, Biao Yi et al.

The dream to create AI assistants as capable and versatile as the fictional J.A.R.V.I.S from Iron Man has long captivated imaginations. With the evolution of (multi-modal) large language models ((M)LLMs), this dream is closer to reality, as (M)LLM-based Agents using computing devices (e.g., computers and mobile phones) by operating within the environments and interfaces (e.g., Graphical User Interface (GUI)) provided by operating systems (OS) to automate tasks have significantly advanced. This paper presents a comprehensive survey of these advanced agents, designated as OS Agents. We begin by elucidating the fundamentals of OS Agents, exploring their key components including the environment, observation space, and action space, and outlining essential capabilities such as understanding, planning, and grounding. We then examine methodologies for constructing OS Agents, focusing on domain-specific foundation models and agent frameworks. A detailed review of evaluation protocols and benchmarks highlights how OS Agents are assessed across diverse tasks. Finally, we discuss current challenges and identify promising directions for future research, including safety and privacy, personalization and self-evolution. This survey aims to consolidate the state of OS Agents research, providing insights to guide both academic inquiry and industrial development. An open-source GitHub repository is maintained as a dynamic resource to foster further innovation in this field. We present a 9-page version of our work, accepted by ACL 2025, to provide a concise overview to the domain.

Shuo Yang, Jinda Lu, Kexin Huang et al.

Reinforcement Learning with Verifiable Rewards (RLVR) has become a promising paradigm for scaling reasoning capabilities of Large Language Models (LLMs). However, the sparsity of binary verifier rewards often leads to low efficiency and optimization instability. To stabilize training, existing methods typically impose token-level constraints relative to a reference policy. We identify that such constraints penalize deviations indiscriminately; this can flip verifier-determined direction when the policy attempts to outperform the reference, thereby suppressing gains. To resolve this, we propose One-Way Policy Optimization (OWPO), a method based on the principle of decoupling optimization direction from update magnitude. In OWPO, the verifier dictates the update direction, while the reference policy serves only to adjust the magnitude. Specifically, OWPO applies asymmetric reweighting: it performs Accelerated Alignment for inferior deviations (where the policy lags behind the reference) and Gain Locking for superior deviations (where the policy surpasses the reference). Furthermore, by incorporating iterative reference updates, OWPO creates a ``Ratchet Effect'' that continuously consolidates gains. Experimental results demonstrate that OWPO outperforms strong baselines, including DAPO, OPD, and MOPD, breaking the bottleneck of fixed priors to enable continuous self-evolution without reliance on external reference models.

Shuo Yang, Jinda Lu, Chiyu Ma et al.

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a central paradigm for scaling LLM reasoning, yet its optimization often suffers from training instability and suboptimal convergence. Through a systematic dissection of clipping-based GRPO-style objectives, we identify the rigid clipping decision induced by hard clipping as a key practical bottleneck in the studied RLVR setups. Specifically, our analysis suggests that informative signals can lie in the near-boundary region just beyond the clipping threshold, and are therefore discarded by the standard hard-clipping rule. Notably, once this bottleneck is precisely identified, even simple stochastic perturbations at the boundary can recover meaningful performance gains. Building on this finding, we propose Near-boundary Stochastic Rescue (NSR), a minimal, plug-and-play modification that stochastically retains these slightly out-of-bound tokens to recover lost signals. While NSR, via stochastic sampling, can be interpreted as inducing an implicit gradient decay in expectation, our ablations reveal that its stochastic, boundary-local rescue mechanism is consistently more effective than deterministic gradient decay. Validated by extensive experiments across model sizes from 7B to 30B and both dense and MoE architectures, as a plug-and-play solution, NSR substantially improves training stability and delivers consistent gains over strong baselines such as DAPO and GSPO.

Wenhao Zhang, Yuexiang Xie, Yuchang Sun et al.

Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) are two prominent post-training paradigms for refining the capabilities and aligning the behavior of Large Language Models (LLMs). Existing approaches that integrate SFT and RL often face the risk of disrupting established response patterns and inducing overfitting to expert data. To address this, we present a novel investigation into the unified view of SFT and RL through an off-policy versus on-policy lens. We propose CHORD, a framework for Controllable Harmonization of On- and Off-Policy Reinforcement Learning via Dynamic Weighting, which reframes SFT not as a separate stage but as a dynamically weighted auxiliary objective within the on-policy RL process. Based on an analysis of off-policy expert data's influence at both holistic and granular levels, we incorporate a dual-control mechanism in CHORD. Specifically, the framework first employs a global coefficient to holistically guide the transition from off-policy imitation to on-policy exploration, and then applies a token-wise weighting function that enables granular learning from the expert, which promotes on-policy exploration and mitigates disruption from off-policy data. We conduct extensive experiments on mathematical reasoning problems and practical tool-use tasks, providing empirical evidence that CHORD achieves a stable and efficient learning process. By effectively harmonizing off-policy expert data with on-policy exploration, CHORD demonstrates significant improvements over baselines. We release the implementation at https://github.com/modelscope/Trinity-RFT/tree/main/examples/mix_chord to inspire further research.

Shuyin Xia, Fan Chen, Dawei Dai et al.

Deep learning models have achieved remarkable success in computer vision but still rely heavily on large-scale labeled data and tend to overfit when data is limited or distributions shift. Data augmentation -- particularly mask-based information dropping -- can enhance robustness by forcing models to explore complementary cues; however, existing approaches often lack structural awareness and risk discarding essential semantics. We propose Granular Ball Guided Masking (GBGM), a structure-aware augmentation strategy guided by Granular Ball Computing (GBC). GBGM adaptively preserves semantically rich, structurally important regions while suppressing redundant areas through a coarse-to-fine hierarchical masking process, producing augmentations that are both representative and discriminative. Extensive experiments on multiple benchmarks demonstrate consistent improvements not only in image classification and masked image reconstruction, but also in image tampering detection, validating the effectiveness and generalization of GBGM across both recognition and forensic scenarios. Simple and model-agnostic, GBGM integrates seamlessly into CNNs and Vision Transformers, offering a practical paradigm for structure-aware data augmentation.

Yuan Li, Li Liu, Penggang Chen et al.

Graph data widely exists in real life, with large amounts of data and complex structures. It is necessary to map graph data to low-dimensional embedding. Graph classification, a critical graph task, mainly relies on identifying the important substructures within the graph. At present, some graph classification methods do not combine the multi-granularity characteristics of graph data. This lack of granularity distinction in modeling leads to a conflation of key information and false correlations within the model. So, achieving the desired goal of a credible and interpretable model becomes challenging. This paper proposes a causal disentangled multi-granularity graph representation learning method (CDM-GNN) to solve this challenge. The CDM-GNN model disentangles the important substructures and bias parts within the graph from a multi-granularity perspective. The disentanglement of the CDM-GNN model reveals important and bias parts, forming the foundation for its classification task, specifically, model interpretations. The CDM-GNN model exhibits strong classification performance and generates explanatory outcomes aligning with human cognitive patterns. In order to verify the effectiveness of the model, this paper compares the three real-world datasets MUTAG, PTC, and IMDM-M. Six state-of-the-art models, namely GCN, GAT, Top-k, ASAPool, SUGAR, and SAT are employed for comparison purposes. Additionally, a qualitative analysis of the interpretation results is conducted.

01. AI, Alex Young, Bei Chen et al.

We introduce the Yi model family, a series of language and multimodal models that demonstrate strong multi-dimensional capabilities. The Yi model family is based on 6B and 34B pretrained language models, then we extend them to chat models, 200K long context models, depth-upscaled models, and vision-language models. Our base models achieve strong performance on a wide range of benchmarks like MMLU, and our finetuned chat models deliver strong human preference rate on major evaluation platforms like AlpacaEval and Chatbot Arena. Building upon our scalable super-computing infrastructure and the classical transformer architecture, we attribute the performance of Yi models primarily to its data quality resulting from our data-engineering efforts. For pretraining, we construct 3.1 trillion tokens of English and Chinese corpora using a cascaded data deduplication and quality filtering pipeline. For finetuning, we polish a small scale (less than 10K) instruction dataset over multiple iterations such that every single instance has been verified directly by our machine learning engineers. For vision-language, we combine the chat language model with a vision transformer encoder and train the model to align visual representations to the semantic space of the language model. We further extend the context length to 200K through lightweight continual pretraining and demonstrate strong needle-in-a-haystack retrieval performance. We show that extending the depth of the pretrained checkpoint through continual pretraining further improves performance. We believe that given our current results, continuing to scale up model parameters using thoroughly optimized data will lead to even stronger frontier models.