Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks54 from public data
- DREAM: Efficient Dataset Distillation by Representative Matching130
A novel matching strategy named as Dataset distillation by REpresentAtive Matching (DREAM), where only representative original images are selected for matching, which is able to be easily plugged into popular dataset distillation frameworks and reduce the distilling iterations by more than 8 times without performance drop.
- InfoBatch: Lossless Training Speed Up by Unbiased Dynamic Data Pruning117
InfoBatch is a novel framework aiming to achieve lossless training acceleration by unbiased dynamic data pruning that randomly prunes a portion of less informative samples based on the loss distribution and rescales the gradients of the remaining samples to approximate the original gradient.
- MSINet: Twins Contrastive Search of Multi-Scale Interaction for Object ReID99
This work proposes a novel Twins Contrastive Mechanism (TCM) to provide more appropriate supervision for ReID architecture search, reduces the category overlaps between the training and validation data, and assists NAS in simulating real-world ReID training schemes.
- Efficient Dataset Distillation via Minimax Diffusion98
This work designs additional minimax criteria in the generative training to enhance these facets for the generated images of diffusion models and presents a theoretical model of the process as hierarchical diffusion control demonstrating the flexibility of the diffusion process to target these criteria without jeopardizing the faithfulness of the sample to the desired distribution.
- Dataset Quantization69
DQ is the first method that can successfully distill large-scale datasets such as ImageNet-1k with a state-of-the-art compression ratio and can be trained with negligible or no performance drop for both vision tasks as well as language tasks.
- SoccerNet 2022 Challenges Results42
The SoccerNet 2022 challenges were the second annual video understanding challenges organized by the SoccerNet team, and tasks (3-6) were novel, including their underlying data and annotations.
- 41
- An Empirical Study of Vehicle Re-Identification on the AI City Challenge40
The solution for the Track2 in AI City Challenge 2021 (AICITY21) is introduced, a vehicle re-identification (ReID) task with both the real-world data and synthetic data, and unsupervised domain-adaptive (UDA) training, post-processing, model ensembling.
- 37
- Summarizing Stream Data for Memory-Constrained Online Continual Learning35
This work proposes to Summarize the knowledge from the Stream Data (SSD) into more informative samples by distilling the training characteristics of real images through maintaining the consistency of training gradients and relationship to the past tasks.
- Finer-CAM: Spotting the Difference Reveals Finer Details for Visual Explanation30
Finer-CAM is proposed, a method that retains CAM’s efficiency while achieving precise localization of discriminative regions and allows adjustable comparison strength, enabling users to selectively highlight coarse object contours or fine discriminative details.
- Taming Diffusion for Dataset Distillation with High Representativeness24
This paper systematically investigate issues present in current diffusion-based dataset distillation methods, including inaccurate distribution matching, distribution deviation with random noise, and separate sampling, and proposes D^3HR, a novel diffusion-based framework to generate distilled datasets with high representativeness.
- 18
- CoreInfer: Accelerating Large Language Model Inference with Semantics-Inspired Adaptive Sparse Activation18
This paper proposes the concept of sentence-wise core neurons, which refers to the subset of neurons most critical for a given sentence, and empirically demonstrates its effectiveness, and introduces CoreInfer, an MLP-free adaptive sparse activation inference method based on sentence-level prediction.
- 18
- Can pre-trained models assist in dataset distillation?16
This work systematically study different options in PTMs, revealing that increasing model diversity enhances the performance of synthetic datasets and significantly improves the cross-architecture generalization over baseline DD methods.
- Prompt-CAM: Making Vision Transformers Interpretable for Fine-Grained Analysis15
Extensive empirical studies on a dozen datasets from various domains validate the superior interpretation capability of Prompt-CAM, requiring only a modification to the prediction head of Visual Prompt Tuning (VPT), in stark contrast to other interpretable methods that require designing specific models and training processes.
- An efficient global representation constrained by Angular Triplet loss for vehicle re-identification14
A simple Angular Triplet loss is introduced on the basis of analysis of different feature representations constrained by softmax loss and triplet loss, which can cooperate with softmax consistently and be seen as an effective baseline for vehicle re-identification task.
- Group Distributionally Robust Dataset Distillation with Risk Minimization13
Drawing inspiration from distributionally robust optimization, this work introduces an algorithm that combines clustering with the minimization of a risk measure on the loss to conduct DD and demonstrates its effective generalization and robustness across subgroups through numerical experiments.
- Stochastic style perturbation modelling for visible-Infrared person re-Identification with severely modality imbalance12
The Co-Modality Balance Learning (CMBL) framework is introduced, designed to recalibrate the balance in cross-modality learning and enhance the extraction of discriminative features, and the novel Class-Aware Contrast Similarity Learning (CACS) strategy, which capitalizes on latent feature consistency to boost intra-class compactness and inter-class separation.
- 12
- 10
- HIERAMP: Coarse-to-Fine Autoregressive Amplification for Generative Dataset Distillation9
Across popular dataset distillation benchmarks, HIERAMP consistently improves validation performance without explicitly optimizing global proximity, demonstrating the importance of semantic amplification for effective dataset distillation.
- DREAM+: Efficient Dataset Distillation by Bidirectional Representative Matching9
A novel matching strategy called Dataset Distillation by Bidirectional REpresentAtive Matching (DREAM+), which selects representative original images for bidirectional matching, is proposed, which significantly reduces the number of distillation iterations by more than 15 times without affecting performance.
- Unsupervised Domain Adaptive Person Re-Identification via Intermediate Domains9
This work is the first to apply intermediate domains to unsupervised domain adaptive person re-identification problems and achieves a state-of-the-art performance.
- 8
- 8
- Color Prompting for Data-Free Continual Unsupervised Domain Adaptive Person Re-Identification8
A Color Prompting (CoP) method for data-free continual unsupervised domain adaptive person Re-ID, which achieves accurate color style recovery for past tasks with adequate data diversity, leading to superior anti-forgetting effects compared with image rehearsal methods.
- 7
- 6
- 3D-Guided Frontal Face Generation for Pose-Invariant Recognition6
This work proposes to recognize faces via generation frontal face images with a 3D-Guided Deep Pose-Invariant Face Recognition Model (3D-PIM) consisted of a simulator and a refiner module and proves the superiority of the proposed “recognition via generation” framework over state-of-the-art.
- 6
- Exploring the Impact of Dataset Bias on Dataset Distillation5
Experiments demonstrate that biases present in the original dataset significantly impact the performance of the synthetic dataset in most cases, which highlights the necessity of identifying and mitigating biases in the original datasets during DD.
- Leveraging Latent Visual Reasoning in Silence4
This work proposes an attention-based reward that encourages generated latent tokens to interact with later text tokens during RL, and shows that this reward promotes latent utilization when the latent mode is activated while preserving the flexibility to use pure-text reasoning.
- BioCAP: Exploiting Synthetic Captions Beyond Labels in Biological Foundation Models4
This work generates synthetic captions with multimodal large language models (MLLMs) guided by Wikipedia-derived visual information and taxon-tailored format examples to train BioCAP, a biological foundation model that captures rich semantics and achieves strong performance in species classification and text-image retrieval.
- 4
- 4
- 4
- Concord: Concept-Informed Diffusion for Dataset Distillation3
This work proposes incorporating the concept understanding of large language models (LLMs) to perform Concept-Informed Diffusion (Concord) for dataset distillation, and demonstrates the efficacy of Concord by achieving state-of-the-art performance on ImageNet-1K and subsets.
- Dataset Distillation from First Principles: Integrating Core Information Extraction and Purposeful Learning3
This work presents a formal model of DD, arguing that a precise characterization of the underlying optimization problem must specify the inference task associated with the application of interest, and reveals novel applications of DD across different modeling environments.
- TaxaAdapter: Vision Taxonomy Models are Key to Fine-grained Image Generation over the Tree of Life2
This work proposes TaxaAdapter, a simple and lightweight approach that incorporates Vision Taxonomy Models (VTMs) such as BioCLIP to guide fine-grained species generation, and introduces a multimodal Large Language Model-based metric that summarizes trait-level descriptions from generated and real images, providing a more interpretable measure of morphological consistency.
- Finer-Personalization Rank: Fine-Grained Retrieval Examines Identity Preservation for Personalized Generation2
Across CUB, Stanford Cars, and animal Re-ID benchmarks, Finer-Personalization Rank more faithfully reflects identity retention than semantic-only metrics and reveals substantial identity drift in several popular personalization methods.
- Lost at the End: Primacy Bias in Multimodal Retrieval-Augmented Question Answering1
The first controlled probe of reader-side position dependence in multimodal KB-VQA is designed, a gold-position protocol in which only the gold passage's prompt slot varies within question, indicating that recall@k is the wrong metric for deployed KB-VQA and that the remaining headroom sits on the reader side.
- 1
- Static Segmentation by Tracking: A Frustratingly Label-Efficient Approach to Fine-Grained Segmentation1
Powered by Segment Anything Model 2 (SAM 2) initially developed for video segmentation, it is shown that SST can achieve high-quality trait and part segmentation with merely one labeled image per species—a breakthrough for analyzing.
- 1
- Better with Experience: Self-Evolving LLM Agents for Evidence-Grounded Health Community Notes–
EvoNote is introduced, an agentic framework that enables health Community Notes generation to self-evolve through an evolving experience memory of prior misinformation correction episodes, positioning self-evolving note generation as a promising paradigm for health misinformation governance.
- –
- –
- –
- Beyond Flat Labels: Level-Restricted Contrastive Learning for Hierarchical Fine-Grained Vision Classification–
This work proposes to restrict contrastive comparisons to categories within the same taxonomic level, and adopts a group-balanced design, ensuring each taxonomic level receives adequate optimization and improves both hierarchical consistency and classification accuracy from coarse to fine granularity.
- AVION: Aerial Vision-Language Instruction from Offline Teacher to Prompt-Tuned Network–
Experiments show that AVION improves few-shot classification and base-class accuracy without degrading generalization to novel categories, and enhances mean recall for cross-modal retrieval, with minimal additional trainable parameters.
- –
- Technical Report for ActivityNet Challenge 2022 -- Temporal Action Localization–
Faster-TAD simplifies the pipeline of TAD and gets remarkable performance, obtaining comparable results as those of multi-step approaches.
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 2026-10-12. 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.