Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks78 from public data
- CoCa: Contrastive Captioners are Image-Text Foundation Models1,864
Contrastive Captioner (CoCa), a minimalist design to pretrain an image-text encoder-decoder foundation model jointly with contrastive loss and captioning loss, thereby subsuming model capabilities from contrastive approaches like CLIP and generative methods like SimVLM.
- Scaling Autoregressive Models for Content-Rich Text-to-Image Generation1,583
The Pathways Autoregressive Text-to-Image (Parti) model is presented, which generates high-fidelity photorealistic images and supports content-rich synthesis involving complex compositions and world knowledge and explores and highlights limitations of the models.
- Ferret: Refer and Ground Anything Anywhere at Any Granularity625
The introduction of Ferret, a new Multimodal Large Language Model capable of understanding spatial referring of any shape or granularity within an image and accurately grounding open-vocabulary descriptions, and a novel and powerful hybrid region representation that integrates discrete coordinates and continuous features jointly to represent a region in the image.
- NoPe-NeRF: Optimising Neural Radiance Field with No Pose Prior378
Experiments on real-world indoor and outdoor scenes show that this method can handle challenging camera trajectories and outperforms existing methods in terms of novel view rendering quality and pose estimation accuracy.
- 327
- 245
- Reveal: Retrieval-Augmented Visual-Language Pre-Training with Multi-Source Multimodal Knowledge Memory194
An end-to-end Retrieval-Augmented Visual Language Model that learns to encode world knowledge into a large-scale memory, and to retrieve from it to answer knowledge-intensive queries, and achieves state-of-the-art results on visual question answering and image captioning.
- FlowNet3D++: Geometric Losses For Deep Scene Flow Estimation109
This work presents FlowNet3D++, a deep scene flow estimation network that incorporates geometric constraints in the form of point-toplane distance and angular alignment between individual vectors in the flow field, into Flow net3D, and proposes a benchmark for flow estimation on the task of dynamic 3D reconstruction.
- 103
- Direct-PoseNet: Absolute Pose Regression with Photometric Consistency93
This work designs a direct matching module that supplies a photometric supervision signal to refine the pose regression network via differentiable rendering and shows that the method can easily cope with additional unlabeled data without the need for external supervision.
- Efficient Meta Lifelong-Learning with Limited Memory84
This paper identifies three common principles of lifelong learning methods and proposes an efficient meta-lifelong framework that combines them in a synergistic fashion and alleviates both catastrophic forgetting and negative transfer at the same time.
- On Negative Interference in Multilingual Models: Findings and A Meta-Learning Treatment77
The results show that negative interference is more common than previously known, suggesting new directions for improving multilingual representations and presenting a meta-learning algorithm that obtains better cross-lingual transferability and alleviates negative interference.
- 66
- 63
- Generative models for protein sequence modeling: recent advances and future directions56
This study intends to provide structured guidance and robust framework for delivering a prospective outlook in the ML-driven protein engineering field by providing a comprehensive survey of model details, novel architecture developments, comparisons of model applications, and current challenges.
- NTIRE 2021 Challenge on Perceptual Image Quality Assessment55
The NTIRE 2021 challenge on perceptual image quality assessment (IQA), held in conjunction with the New Trends in Image Restoration and Enhancement workshop (NTIRE) workshop at CVPR 2021, achieved much better results than existing IQA methods, while the winning method can demonstrate state-of-the-art performance.
- TokenCompose: Text-to-Image Diffusion with Token-Level Supervision52
TokenCompose aims to improve multi-category instance composition by introducing the token-wise consistency terms between the image content and object segmentation maps in the finetuning stage of the Latent Diffusion Model.
- 48
- VideoCoCa: Video-Text Modeling with Zero-Shot Transfer from Contrastive Captioners48
VideoCoCa is presented, that reuses a pretrained image-text contrastive captioner (CoCa) model and adapt it to video-text tasks with minimal extra training, and achieves strong results on video question-answering and video captioning, and establishes a simple and effective video- text baseline for future research.
- 43
- DifIISR: A Diffusion Model with Gradient Guidance for Infrared Image Super-Resolution39
The proposed DifIISR is an infrared image super-resolution diffusion model optimized for visual quality and perceptual performance that achieves task-based guidance for diffusion by injecting gradients derived from visual and perceptual priors into the noise during the reverse process.
- 36
- OmniControlNet: Dual-stage Integration for Conditional Image Generation30
This work provides a two-way integration for the widely adopted ControlNet by integrating external condition generation algorithms into a single dense prediction method and incorporating its individually trained image generation processes into a single model.
- Language Models as Science Tutors25
This work introduces TutorEval and TutorChat, and uses TutorChat to fine-tune Llemma models with 7B and 34B parameters, and creates TutorChat, a dataset of 80,000 long synthetic dialogues about textbooks.
- Understanding Alignment in Multimodal LLMs: A Comprehensive Study21
A novel way of creating multimodal preference data called Bias-Driven Hallucination Sampling (BDHS) is introduced that needs neither additional annotation nor external models, and can achieve competitive performance to previously published alignment work for multimodal models across a range of benchmarks.
- 20
- On the Feasibility of Cross-Task Transfer with Model-Based Reinforcement Learning19
This work proposes Model-Based Cross-Task Transfer (XTRA), a framework for sample-efficient online RL with scalable pretraining and finetuning of learned world models, and achieves substantial improvements over a baseline trained from scratch.
- 15
- A distributed cross-chain mechanism based on notary schemes and group signatures15
This paper improves the group signature scheme by setting up a group of tracing authorities to achieve distributed tracing and incorporates this signature scheme into the notary scheme, where a group of notaries signs cross-chain transactions using group signatures.
- Revisiting MoE and Dense Speed-Accuracy Comparisons for LLM Training13
Experimental results show that MoE consistently outperform dense LLMs on the speed-accuracy trade-off curve with meaningful gaps, and the total compute budget under the Chinchilla compute-optimal settings is determined.
- Towards More Reliable Transfer Learning13
This paper addresses challenges related to sources with diverse labeling volume and diverse reliability by proposing a new transfer learning method that utilizes both source-target similarities and inter-source relationships and involves pool-based active learning.
- 12
- Ray-ONet: Efficient 3D Reconstruction From A Single RGB Image12
By predicting a series of occupancy probabilities along a ray that is back-projected from a pixel in the camera coordinate, the method Ray-ONet improves the reconstruction accuracy in comparison with Occupancy Networks (ONet), while reducing the network inference complexity to O($N^2$).
- 11
- 10
- 10
- 9
- 9
- 8
- GS-CPR: Efficient Camera Pose Refinement via 3D Gaussian Splatting8
This work leverages 3D Gaussian Splatting as a scene representation and proposes a novel test-time camera pose refinement framework, GSLoc, which enhances the localization accuracy of state-of-the-art absolute pose regression and scene coordinate regression methods.
- 8
- Evaluating a Tactile Approach to Programming Scratch8
An innovative peripheral device that can help students with visual impairments access Scratch by simply remembering six buttons and guided by auditory cues generated by the computer is described.
- Lightweight Single Image Super-Resolution With Similar Feature Fusion Block7
This work proposes a generic, lightweight and efficient feature fusion block to replace the commonly used 1*1 convolution, and proposes the enhanced shallow residual blocks to improve the super-resolution performance.
- Phytochemical drug discovery for COVID-19 using high-resolution computational docking and machine learning assisted binder prediction7
This study screens a wide range of natural drug products against a comprehensive set of SARS-CoV-2 proteins using a high-resolution computational workflow and demonstrates the advantage of incorporating machine learning elements into a virtual screening workflow.
- DyPoA: enhanced PoA protocol with a dynamic set of validators for IoT6
DyPoA is proposed, a proof of authority based on a dynamic set of validators that enables unauthenticated nodes to perform the process of packing transactions into blocks, while authenticated nodes validate and commit blocks to the blockchain.
- Exploiting Category Names for Few-Shot Classification with Vision-Language Models6
This paper shows that it can significantly improve the performance of few-shot classification by using the category names to initialize the classification head, and obtains the state-of-the-art performance on a number ofFew-shot image classification benchmarks.
- 6
- Privacy-preserving cross-chain asset transfers using notary schemes and zero-knowledge proofs5
Experimental results demonstrated that the proposed privacy-preserving notary scheme has significantly improved efficiency compared to existing schemes, particularly regarding computational cost and time consumption when handling large prime numbers.
- 5
- Improving Multi-agent Reinforcement Learning with Stable Prefix Policy4
This work uses the previously collected trajectories to construct a Monte-Carlo Trajectory Tree and achieves not only better performance but also faster convergence speed than baseline algorithms within early time steps.
- 4
- Research on Teaching Reform of Artificial Intelligence Course Based on CDIO4
The “12365” principle is proposed based on the CDIO concept, and corresponding teaching reform and practice are carried out to set Artificial Intelligence courses in the lower grades of the university.
- AS-IntroVAE: Adversarial Similarity Distance Makes Robust IntroVAE4
The vanishing gradient problem is theoretically analyzed and a new Adversarial Similarity Distance Introspective Variational Autoencoder (AS-IntroVAE) is proposed, which fosters more diverse distributions in the latent space, allowing the model to produce images of great diversity.
- 4
- Reflect3r: Single-View 3D Stereo Reconstruction Aided by Mirror Reflections2
This work treats the reflection as an auxiliary view and designs a transformation that constructs a physically valid virtual camera, allowing direct pixel-domain generation of the virtual view while adhering to the real-world imaging process, enabling a multi-view stereo setup from a single image.
- 2
- Scalable batch verification of ECDSA for blockchain using IVC2
BEATS (Batch ECDSA Transaction verification Scheme), where the underlying SNARK is instantiated by Spartan with Bulletproof commitment, is introduced, where the underlying SNARK is instantiated by Spartan with Bulletproof commitment.
- 2
- HiMemFormer: Hierarchical Memory-Aware Transformer for Multi-Agent Action Anticipation2
The Hierarchical Memory-Aware Transformer (HiMemFormer), a transformer-based model for online multi-agent action anticipation, is presented, which uniquely hierarchically applies the global context with agent-specific preferences to avoid noisy or redundant information in multi-agent action anticipation.
- 2
- Stochastic C-GNet Environment Modeling and Path Planning Optimization in a Narrow and Long Space2
This study considers the walking and space constraints of a roadheader to construct the network topology for the traffic cost value weights and the improved algorithm of variation self-adapting particle swarm optimization is proposed to optimize the regional path.
- 1
- 1
- AXLearn: Modular, Hardware-Agnostic Large Model Training1
A novel method of quantifying modularity via Lines-of-Code (LoC)-complexity is introduced, which demonstrates how the AXLearn system maintains constant complexity as the authors scale the components in the system, compared to linear or quadratic complexity in other systems.
- 1
- 1
- 1
- Using Touch and Sound to Program Scratch1
In this evaluation, CodeBox64 is a possible solution, offering a tactile approach to input, a peripheral device containing six buttons that generates auditory cues to indicate the name of the block when the mouse lands on it.
- A Study of Language Practices in Human-Ai Interaction: A Case Study of Posts from the Douban “Human-Ai Romance” Community–
The research uses data from user posts in Douban's “Human-AI Romance” community to explore the linguistic practices of Chinese users in Human-AI Interaction and provides empirical insights for the design of language style and interaction modes of AI companion software.
- –
- –
- –
- –
- –
- Subjective well-being of women undergoing in vitro fertilization and embryo transfer and influencing factors–
Overall, female subjects receiving IVF-ET treatment have a moderate level of SWB, which is affected by various factors, and clinical care should identify factors influencing SWB and provide interventions to enhance SWB of infertile women during IVF-ET treatment.
- Extracting function relation from less than ideal digital curve–
A least squares model with outlier elimination strategy is presented, which can be solved to obtain the function parameters corresponding to the curve by iterative method and shows that the difference between the obtained fitting function and the original function is positively related to the effective resolution of the digital image.
- Novel image fusion methods using intuitionistic index–
These new algorithms are better comparing to other algorithms, and they can get a good fusion result, especially the contourlet transformation algorithm using the intuitionistic index.
- –
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 2026-10-15. 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.