Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Qi SuSuggested from co-authorship
- Suggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works108 from public data
- Qwen2.5 Technical Report5,408
The open-weight flagship Qwen2.5-72B-Instruct outperforms a number of open and proprietary models and demonstrates competitive performance to the state-of-the-art open-weight model, Llama-3-405B-Instruct, which is around 5 times larger.
- Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution5,204
The Qwen2-VL Series introduces the Naive Dynamic Resolution mechanism, which enables the model to dynamically process images of varying resolutions into different numbers of visual tokens, allowing the model to generate more efficient and accurate visual representations, closely aligning with human perceptual processes.
- Qwen Technical Report4,443
Qwen is a comprehensive language model series that encompasses distinct models with varying parameter counts, and includes Qwen, the base pretrained language models, and Qwen-Chat, the chat models finetuned with human alignment techniques.
- Qwen2 Technical Report2,548
A comprehensive suite of foundational and instruction-tuned language models, encompassing a parameter range from 0.5 to 72 billion, featuring dense models and a Mixture-of-Experts model, which surpasses most prior open-weight models and demonstrates robust multilingual capabilities.
- Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond1,965
The Qwen-VL series is introduced, a set of large-scale vision-language models designed to perceive and understand both text and images that outperforms existing Large Vision Language Models (LVLMs).
- Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models1,618
The Qwen3 Embedding series is introduced, a significant advancement over its predecessor, the GTE-Qwen series, in text embedding and reranking capabilities, built upon the Qwen3 foundation models, and achieves state-of-the-art results across diverse benchmarks.
- Qwen2.5-Math Technical Report: Toward Mathematical Expert Model via Self-Improvement1,174
A series of math-specific large language models that possess advanced mathematical reasoning capabilities, including Chain-of-Thought (CoT) and Tool-Integrated Reasoning (TIR), and are evaluated on 10 mathematics datasets in both English and Chinese.
- Qwen3-Omni Technical Report582
Qwen3-Omni is presented, a single multimodal model that, for the first time, maintains state-of-the-art performance across text, image, audio, and video without any degradation relative to single-modal counterparts and introduces a Thinking model that explicitly reasons over inputs from any modality.
- Towards Knowledge-Based Recommender Dialog System318
A novel end-to-end framework called KBRD, which stands for Knowledge-Based Recommender Dialog System, integrates the recommender system and the dialog generation system and shows that the two systems can bring mutual benefits to each other.
- 250
- ExpertPrompting: Instructing Large Language Models to be Distinguished Experts235
This paper proposes ExpertPrompting to elicit the potential of LLMs to answer as distinguished experts, and produces a competitive open-source chat assistant called ExpertLLaMA, which outperforms existing open-source opponents and achieves 96\% of the original ChatGPT's capability.
- Chinese CLIP: Contrastive Vision-Language Pretraining in Chinese229
This work constructs a large-scale dataset of image-text pairs in Chinese, where most data is retrieved from publicly available datasets, and proposes a two-stage pretraining method, where the model is first trained with the image encoder frozen and then trained with all parameters being optimized to achieve enhanced model performance.
- Modality Competition: What Makes Joint Training of Multi-modal Network Fail in Deep Learning? (Provably)220
It is proved that for the multi-modal late-fusion network with (smoothed) ReLU activation trained jointly by gradient descent, different modalities will compete with each other, and the encoder networks will learn only a subset of modalities.
- Qwen2.5-1M Technical Report201
This framework provides an efficient and powerful solution for developing applications that require long-context processing using open-source models and detail the optimizations in the inference engine, including kernel optimization, pipeline parallelism, and scheduling optimization, which significantly enhance overall inference performance.
- ONE-PEACE: Exploring One General Representation Model Toward Unlimited Modalities176
This work releases ONE-PEACE, a highly extensible model with 4B parameters that can seamlessly align and integrate representations across vision, audio, and language modalities, and develops two modality-agnostic pretraining tasks, which align the semantic space of different modalities and capture fine-grained details within modalities concurrently.
- Qwen3-ASR Technical Report175
Timestamp accuracy experiments show that the proposed model outperforms the three strongest force alignment models and takes more advantages in efficiency and versatility, and is released under the Apache 2.0 license.
- 163
- Diversity-Promoting GAN: A Cross-Entropy Based Generative Adversarial Network for Diversified Text Generation154
A novel language-model based discriminator is proposed, which can better distinguish novel text from repeated text without the saturation problem compared with existing classifier-based discriminators.
- Global Encoding for Abstractive Summarization153
A global encoding framework, which controls the information flow from the encoder to the decoder based on the global information of the source context, and consists of a convolutional gated unit to perform global encoding to improve the representations of thesource-side information.
- 151
- 147
- #InsTag: Instruction Tagging for Analyzing Supervised Fine-tuning of Large Language Models135
This work proposes InsTag, an open-set fine-grained tagger, to tag samples within SFT datasets based on semantics and intentions and defines instruction diversity and complexity regarding tags, and obtains 6.6K tags to describe comprehensive user queries.
- CodeElo: Benchmarking Competition-level Code Generation of LLMs with Human-comparable Elo Ratings120
This work introduces CodeElo, a standardized competition-level code generation benchmark that effectively addresses all these challenges for the first time and provides the Elo ratings of 30 existing popular open-source and 3 proprietary LLMs for the first time.
- CC-OCR: A Comprehensive and Challenging OCR Benchmark for Evaluating Large Multimodal Models in Literacy103
CC-OCR aims to comprehensively evaluate the capabilities of LMMs on OCR-centered tasks, facilitating continued progress in literacy with rich structure and fine-grained visual challenges.
- Imitation Learning for Non-Autoregressive Neural Machine Translation99
This paper proposes an imitation learning framework for non-autoregressive machine translation, which still enjoys the fast translation speed but gives comparable translation performance compared to its auto-regressive counterpart.
- Bag-of-Words as Target for Neural Machine Translation78
This paper proposes an approach that uses both the sentences and the bag-of-words as targets in the training stage, in order to encourage the model to generate the potentially correct sentences that are not appeared in theTraining set.
- 72
- Semantic-Unit-Based Dilated Convolution for Multi-Label Text Classification70
A novel model for multi-label text classification that generates higher-level semantic unit representations with multi-level dilated convolution as well as a corresponding hybrid attention mechanism that extracts both the information at the word-level and the level of the semantic unit.
- A Deep Reinforced Sequence-to-Set Model for Multi-Label Classification68
This work proposes a simple but effective sequence-to-set model, trained via reinforcement learning, that can substantially outperform competitive baselines, as well as effectively reduce the sensitivity to the label order.
- BaGuaLu65
BaGuaLu is proposed, the first work targeting training brain scale models on an entire exascale supercomputer, the New Generation Sunway Supercomputer, by combining hardware-specific intra-node optimization and hybrid parallel strategies, which enables decent performance and scalability on unprecedentedly large models.
- TouchStone: Evaluating Vision-Language Models by Language Models64
This paper proposes an evaluation method that uses strong LLMs as judges to comprehensively evaluate the various abilities of LVLMs, and constructs a comprehensive visual dialogue dataset TouchStone, consisting of open-world images and questions, covering five major categories of abilities and 27 subtasks.
- A Hierarchical End-to-End Model for Jointly Improving Text Summarization and Sentiment Classification61
This work proposes a hierarchical end-to-end model for joint learning of text summarization and sentiment classification, where the sentiment classification label is treated as the further ``summarization'' of theText summarization output.
- A Gesture Air-Writing Tracking Method that Uses 24 GHz SIMO Radar SoC52
A prototype of a gesture air-writing tracking system that is based on the 24-GHz frequency-modulated continuous-wave (FMCW) radar system-on-chip (SoC) is proposed, which provides sufficient range resolution to track hand gestures and the efficiency and accuracy of the proposed method is verified.
- 52
- 50
- M6: Multi-Modality-to-Multi-Modality Multitask Mega-transformer for Unified Pretraining47
This work proposes the largest dataset for pretraining in Chinese, which consists of over 1.9TB images and 292GB texts, and proposes a method called M6, referring to Multi-Modality-to-Multi- Modality Multitask Mega-transformer, for unified pretraining on the data of single modality and multiple modalities.
- Autoencoder as Assistant Supervisor: Improving Text Representation for Chinese Social Media Text Summarization47
This work supervise the learning of the representation of the source content with that of the summary of the sequence-to-sequence model (Seq2Seq), and evaluates the model on a popular Chinese social media dataset.
- 46
- Learning When to Concentrate or Divert Attention: Self-Adaptive Attention Temperature for Neural Machine Translation43
This work proposes a new model with a mechanism called Self-Adaptive Control of Temperature (SACT) to control the softness of attention by means of an attention temperature and demonstrates that the model outperforms the baseline models.
- Prompt Tuning for Generative Multimodal Pretrained Models39
This work implements prompt tuning on the unified sequence-to-sequence pretrained model adaptive to both understanding and generation tasks, with a focus on generative multimodal pretrained models, instead of contrastive ones.
- Learning Relation Alignment for Calibrated Cross-modal Retrieval39
A novel metric is proposed, Intra-modal Self-attention Distance (ISD), to quantify the relation consistency by measuring the semantic distance between linguistic and visual relations, and a regularized training method is presented, IAIS, to optimize the ISD and calibrate intra- modal self-attentions from the two modalities mutually via inter-modAL alignment.
- An Auto-Encoder Matching Model for Learning Utterance-Level Semantic Dependency in Dialogue Generation38
An Auto-Encoder Matching (AEM) model is proposed to learn utterance-level semantic dependency and is capable of generating responses of high coherence and fluency compared to baseline models.
- Connecting Language and Vision for Natural Language-Based Vehicle Retrieval37
This work proposes to jointly train the state-of-the-art vision models with the transformer-based language model in an end-to-end manner to connect language and vision, and achieves the 1st place on the 5th AI City Challenge.
- 28
- 27
- Language Models can Self-Lengthen to Generate Long Texts26
Experiments on benchmarks and human evaluations show that Self-Lengthen outperforms existing methods in long-text generation, when applied to top open-source LLMs such as Qwen2 and LLaMA3.
- Sketch and Refine: Towards Faithful and Informative Table-to-Text Generation26
A novel Skeleton-based two-stage method that combines both Autoregressive and Non-Autoregressive generations (SANA) that outperforms the previous state-of-the-art methods in both automatic and human evaluation, especially on coverage and faithfulness.
- AutoLogi: Automated Generation of Logic Puzzles for Evaluating Reasoning Abilities of Large Language Models25
This approach features program-based verification and controllable difficulty levels, enabling more reliable evaluation that better distinguishes models' reasoning abilities, and develops a bilingual benchmark, AutoLogi, to obtain more accurate assessments of models' reasoning capabilities.
- VideoAgentTrek: Computer Use Pretraining from Unlabeled Videos20
VideoAgentTrek is presented, a scalable pipeline that automatically mines training data from publicly available screen-recorded videos at web scale, eliminating the need for manual annotation, providing a scalable alternative to expensive manual annotation.
- CUA-Gym: Scaling Verifiable Training Environments and Tasks for Computer-Use Agents18
This work presents CUA-Gym, a scalable pipeline that co-generates task instructions, environment states, and reward functions, and constructs CUA-Gym, a dataset of 32,112 verified RLVR training tuples grounded in 110 environments and synthesizes CUA-Gym-Hub, a broad suite of high-fidelity mock web applications grounded in real-world software-use distributions.
- DataMan: Data Manager for Pre-training Large Language Models18
This paper trains a Data Manager to learn quality ratings and domain recognition from pointwise rating, and uses it to annotate a 447B token pre-training corpus with 14 quality ratings and domain type, and derives 14 quality criteria from the causes of text perplexity anomalies and introduces 15 common application domains to support domain mixing.
- 17
- 17
- 15
- 14
- RealCritic: Towards Effectiveness-Driven Evaluation of Language Model Critiques13
A new benchmark designed to assess the critique capabilities of Large Language Models, which employs a closed-loop methodology that evaluates the quality of corrections generated from critiques.
- 13
- 13
- Multi-Agent Collaboration for Multilingual Code Instruction Tuning12
This work introduces a novel multi-agent collaboration framework to enhance multilingual instruction tuning for code LLMs, where multiple language-specific intelligent agent components with generation memory work together to transfer knowledge from one language to another efficiently and effectively.
- 12
- 12
- 12
- CoRT: Code-integrated Reasoning within Thinking11
CoRT is introduced, a post-training framework for teaching LRMs to leverage CI effectively and efficiently, by synthesizing code-integrated reasoning data through Hint-Engineering, which strategically inserts different hints at appropriate positions to optimize LRM-CI interaction.
- 10
- 9
- A Spark of Vision-Language Intelligence: 2-Dimensional Autoregressive Transformer for Efficient Finegrained Image Generation9
This work tackles the information loss bottleneck of vector-quantization (VQ) autoregressive image generation by introducing a novel model architecture called the 2-Dimensional Autoregression (DnD) Transformer, an end-to-end model that can generate higher quality images with the same backbone model size and sequence length.
- InterBERT: An Effective Multi-Modal Pretraining Approach via Vision-and-Language Interaction9
The analysis shows that the proposed MGM is effective for pretraining, and the method for multi-modal pretraining can adapt to single- modal tasks without significant performance decrease in comparison with the BERT-base model.
- 8
- WorldPM: Scaling Human Preference Modeling7
Through evaluations on 7 benchmarks with 20 subtasks, it is found that WorldPM broadly improves the generalization performance across human preference datasets of varying sizes, with performance gains exceeding 5% on many key subtasks.
- Towards Better Correctness and Efficiency in Code Generation7
A deeper dive into the code efficiency problem is taken, identifying then proposing methods to overcome key bottlenecks and proposing a two-stage tuning method, which achieves high and balanced performance across correctness and efficiency.
- Instance-wise Prompt Tuning for Pretrained Language Models7
Instance-wise Prompt Tuning (IPT) is introduced, the first prompt learning paradigm that injects knowledge from the input data instances to the prompts, thereby providing PLMs with richer and more concrete context information.
- 6
- 6
- 6
- 6
- 4
- 4
- LongWeave: A Long-Form Generation Benchmark Bridging Real-World Relevance and Verifiability3
LongWeave is introduced, which balances real-world and verifiable assessment with Constraint-Verifier Evaluation (CoV-Eval), and ensures that tasks are both realistic and objectively assessable, enabling rigorous assessment of model capabilities in meeting complex real-world constraints.
- 3
- 3
- 2
- 2
- 2
- 2
- 1
- 1
- 1
- 1
- 1
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-16. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.