Search Results for author: Debing Zhang

Found 23 papers, 9 papers with code

MLLM-Selector: Necessity and Diversity-driven High-Value Data Selection for Enhanced Visual Instruction Tuning

no code implementations26 Mar 2025 Yiwei Ma, Guohai Xu, Xiaoshuai Sun, Jiayi Ji, Jie Lou, Debing Zhang, Rongrong Ji

Visual instruction tuning (VIT) has emerged as a crucial technique for enabling multi-modal large language models (MLLMs) to follow user instructions adeptly.

Diversity

The Devil Is in the Details: Tackling Unimodal Spurious Correlations for Generalizable Multimodal Reward Models

no code implementations5 Mar 2025 Zichao Li, Xueru Wen, Jie Lou, Yuqiu Ji, Yaojie Lu, Xianpei Han, Debing Zhang, Le Sun

Multimodal Reward Models (MM-RMs) are crucial for aligning Large Language Models (LLMs) with human preferences, particularly as LLMs increasingly interact with multimodal data.

Scalable Oversight for Superhuman AI via Recursive Self-Critiquing

no code implementations7 Feb 2025 Xueru Wen, Jie Lou, Xinyu Lu, Junjie Yang, Yanjiang Liu, Yaojie Lu, Debing Zhang, Xingyu

As AI capabilities increasingly surpass human proficiency in complex tasks, current alignment techniques including SFT and RLHF face fundamental challenges in ensuring reliable oversight.

SedarEval: Automated Evaluation using Self-Adaptive Rubrics

1 code implementation26 Jan 2025 Zhiyuan Fan, Weinong Wang, Xing Wu, Debing Zhang

Using the same training data, our evaluator LM achieves a higher concordance rate with human grading results than other paradigms, including GPT-4, highlighting the superiority and efficiency of our approach.

Logical Reasoning

NExtLong: Toward Effective Long-Context Training without Long Documents

1 code implementation22 Jan 2025 Chaochen Gao, Xing Wu, Zijia Lin, Debing Zhang, Songlin Hu

Large language models (LLMs) with extended context windows have made significant strides yet remain a challenge due to the scarcity of long documents.

RedStar: Does Scaling Long-CoT Data Unlock Better Slow-Reasoning Systems?

no code implementations20 Jan 2025 Haotian Xu, Xing Wu, Weinong Wang, Zhongzhi Li, Da Zheng, Boyuan Chen, Yi Hu, Shijia Kang, Jiaming Ji, Yingying Zhang, Zhijiang Guo, Yaodong Yang, Muhan Zhang, Debing Zhang

In this work, we explore the untapped potential of scaling Long Chain-of-Thought (Long-CoT) data to 1000k samples, pioneering the development of a slow-thinking model, RedStar.

Math Reinforcement Learning (RL)

Diff-Instruct*: Towards Human-Preferred One-step Text-to-image Generative Models

1 code implementation28 Oct 2024 Weijian Luo, Colin Zhang, Debing Zhang, Zhengyang Geng

In this paper, we introduce the Diff-Instruct* (DI*), an image data-free approach for building one-step text-to-image generative models that align with human preference while maintaining the ability to generate highly realistic images.

CodePMP: Scalable Preference Model Pretraining for Large Language Model Reasoning

no code implementations3 Oct 2024 Huimu Yu, Xing Wu, Weidong Yin, Debing Zhang, Songlin Hu

Large language models (LLMs) have made significant progress in natural language understanding and generation, driven by scalable pretraining and advanced finetuning.

GSM8K Language Modeling +7

Untie the Knots: An Efficient Data Augmentation Strategy for Long-Context Pre-Training in Language Models

no code implementations7 Sep 2024 Junfeng Tian, Da Zheng, Yang Cheng, Rui Wang, Colin Zhang, Debing Zhang

Large language models (LLM) have prioritized expanding the context window from which models can incorporate more information.

Data Augmentation

MPI-Flow: Learning Realistic Optical Flow with Multiplane Images

1 code implementation ICCV 2023 Yingping Liang, Jiaming Liu, Debing Zhang, Ying Fu

The accuracy of learning-based optical flow estimation models heavily relies on the realism of the training datasets.

Optical Flow Estimation

Recurrent Self-Supervised Video Denoising with Denser Receptive Field

no code implementations7 Aug 2023 Zichun Wang, Yulun Zhang, Debing Zhang, Ying Fu

However, under their blind spot constraints, previous self-supervised video denoising methods suffer from significant information loss and texture destruction in either the whole reference frame or neighbor frames, due to their inadequate consideration of the receptive field.

Denoising Video Denoising

End-to-End Video Text Spotting with Transformer

1 code implementation20 Mar 2022 Weijia Wu, Yuanqiang Cai, Chunhua Shen, Debing Zhang, Ying Fu, Hong Zhou, Ping Luo

Recent video text spotting methods usually require the three-staged pipeline, i. e., detecting text in individual images, recognizing localized text, tracking text streams with post-processing to generate final results.

Text Detection Text Spotting

TransAug: Translate as Augmentation for Sentence Embeddings

no code implementations30 Oct 2021 Jue Wang, Haofan Wang, Xing Wu, Chaochen Gao, Debing Zhang

In this paper, we present TransAug (Translate as Augmentation), which provide the first exploration of utilizing translated sentence pairs as data augmentation for text, and introduce a two-stage paradigm to advances the state-of-the-art sentence embeddings.

Contrastive Learning Data Augmentation +4

Partial FC: Training 10 Million Identities on a Single Machine

7 code implementations11 Oct 2020 Xiang An, Xuhan Zhu, Yang Xiao, Lan Wu, Ming Zhang, Yuan Gao, Bin Qin, Debing Zhang, Ying Fu

The experiment demonstrates no loss of accuracy when training with only 10\% randomly sampled classes for the softmax-based loss functions, compared with training with full classes using state-of-the-art models on mainstream benchmarks.

Face Identification Face Recognition +2

EasyQuant: Post-training Quantization via Scale Optimization

1 code implementation30 Jun 2020 Di Wu, Qi Tang, Yongle Zhao, Ming Zhang, Ying Fu, Debing Zhang

The 8 bits quantization has been widely applied to accelerate network inference in various deep learning applications.

Quantization

Joint Camera Spectral Sensitivity Selection and Hyperspectral Image Recovery

no code implementations ECCV 2018 Ying Fu, Tao Zhang, Yinqiang Zheng, Debing Zhang, Hua Huang

Hyperspectral image (HSI) recovery from a single RGB image has attracted much attention, whose performance has recently been shown to be sensitive to the camera spectral sensitivity (CSS).

Cannot find the paper you are looking for? You can Submit a new open access paper.