✨FlowAnchor: Stabilizing the Editing Signal for Inversion-Free Video Editing

📝 Summary:
FlowAnchor stabilizes inversion-free video editing by addressing signal instability in high-dimensional latent spaces. It uses spatial-aware attention refinement and adaptive magnitude modulation to ensure precise localization and sufficient editing strength, leading to faithful and coherent vide...

🔹 Publication Date: Published on Apr 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.22586
• PDF: https://arxiv.org/pdf/2604.22586

==================================

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#VideoEditing #DeepLearning #ComputerVision #GenerativeAI #AIResearch

✨Sessa: Selective State Space Attention

📝 Summary:
Sessa is a new decoder architecture that puts attention inside a recurrent feedback path. This allows it to model long contexts better than Transformers and state-space models, achieving power-law memory decay and flexible selective retrieval. It outperforms on long-context tasks.

🔹 Publication Date: Published on Apr 21

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.18580
• PDF: https://arxiv.org/pdf/2604.18580
• Github: https://github.com/LibratioAI/sessa

==================================

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#Sessa #DeepLearning #AttentionMechanisms #StateSpaceModels #LongContextAI

✨Sapiens2

📝 Summary:
Sapiens2 is a high-resolution transformer model for human-centric vision. It achieves state-of-the-art performance by combining unified pretraining objectives, a large 1-billion image dataset, and architectural improvements, excelling in tasks like pose and segmentation.

🔹 Publication Date: Published on Apr 23

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.21681
• PDF: https://arxiv.org/pdf/2604.21681
• Github: https://github.com/facebookresearch/sapiens2

🔹 Models citing this paper:
• https://huggingface.co/facebook/sapiens2
• https://huggingface.co/facebook/sapiens2-seg-5b
• https://huggingface.co/facebook/sapiens2-seg-1b

✨ Spaces citing this paper:
• https://huggingface.co/spaces/facebook/sapiens2-seg
• https://huggingface.co/spaces/facebook/sapiens2-pointmap
• https://huggingface.co/spaces/facebook/sapiens2-normal

==================================

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#Sapiens2 #ComputerVision #TransformerModels #HumanCentricAI #DeepLearning

✨Large Language Models Explore by Latent Distilling

📝 Summary:
Exploratory Sampling ESamp boosts LLM diversity beyond lexical variation. It uses a lightweight Distiller to predict hidden representations, biasing decoding towards novel semantic patterns via prediction error. ESamp boosts reasoning efficiency and creative writing, with low overhead.

🔹 Publication Date: Published on Apr 27

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.24927
• PDF: https://arxiv.org/pdf/2604.24927
• Github: https://github.com/LinesHogan/tllm

==================================

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#LLM #AI #NLP #DeepLearning #GenerativeAI

✨ViPO: Visual Preference Optimization at Scale

📝 Summary:
ViPO scales visual preference optimization using Poly-DPO for noisy data and constructing ViPO, a large high-quality dataset. This dual approach yields superior performance, emphasizing that algorithmic adaptability and data quality are crucial.

🔹 Publication Date: Published on Apr 29

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.24953
• PDF: https://arxiv.org/pdf/2604.24953
• Project Page: https://liming-ai.github.io/ViPO
• Github: https://liming-ai.github.io/ViPO

==================================

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#VisualAI #MachineLearning #DeepLearning #Optimization #DataScience

✨SignRoundV2: Closing the Performance Gap in Extremely Low-Bit Post-Training Quantization for LLMs

📝 Summary:
SignRoundV2 is a post-training quantization method for LLMs. It achieves competitive, near full-precision accuracy even at extremely low-bits like 2-bits. This is done via layer-wise bit allocation and pre-tuning scale search.

🔹 Publication Date: Published on Dec 4, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2512.04746
• PDF: https://arxiv.org/pdf/2512.04746
• Project Page: https://github.com/intel/auto-round
• Github: https://github.com/intel/auto-round

🔹 Models citing this paper:
• https://huggingface.co/Intel/MiroThinker-v1.5-30B-gguf-q2ks-mixed-AutoRound
• https://huggingface.co/Intel/DeepSeek-R1-0528-Qwen3-8B-int4-AutoRound

==================================

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#LLMs #Quantization #DeepLearning #AI #MachineLearning

✨Nemotron 3 Nano Omni: Efficient and Open Multimodal Intelligence

📝 Summary:
Nemotron 3 Nano Omni is a new efficient, open multimodal AI model. It natively supports audio, text, images, and video inputs, improving accuracy and efficiency over previous versions. It excels in document understanding and long audio-video comprehension.

🔹 Publication Date: Published on Apr 27

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.24954
• PDF: https://arxiv.org/pdf/2604.24954

🔹 Models citing this paper:
• https://huggingface.co/nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16
• https://huggingface.co/nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4
• https://huggingface.co/nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8

✨ Spaces citing this paper:
• https://huggingface.co/spaces/akhaliq/Nemotron-3-Nano-Omni
• https://huggingface.co/spaces/developerjeremylive/Nemotron-3-Nano-Omni-etheroi

==================================

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#AI #MultimodalAI #DeepLearning #OpenSourceAI #AIResearch

✨DeepSeek-OCR: Contexts Optical Compression

📝 Summary:
DeepSeek-OCR compresses long contexts via optical 2D mapping to achieve high OCR precision with significantly reduced vision tokens. It shows 97% accuracy at 10x compression, outperforming other OCR models efficiently. This innovation holds practical value for document processing and LLM training...

🔹 Publication Date: Published on Oct 21, 2025

🔹 Paper Links:
• arXiv Page: https://arxivexplained.com/papers/deepseek-ocr-contexts-optical-compression
• PDF: https://arxiv.org/pdf/2510.18234
• Github: https://github.com/deepseek-ai/DeepSeek-OCR

🔹 Models citing this paper:
• https://huggingface.co/deepseek-ai/DeepSeek-OCR
• https://huggingface.co/deepseek-ai/DeepSeek-OCR-2
• https://huggingface.co/unsloth/DeepSeek-OCR

✨ Spaces citing this paper:
• https://huggingface.co/spaces/merterbak/DeepSeek-OCR-Demo
• https://huggingface.co/spaces/davidpcm/openclaw-stock-analyst
• https://huggingface.co/spaces/khang119966/DeepSeek-OCR-DEMO

==================================

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#OCR #AI #DeepLearning #ContextCompression #LLM

🔥 GoLongRL: Capability-Oriented Long Context Reinforcement Learning with Multitask Alignment

💡 The paper introduces GoLongRL, a new approach to long context reinforcement learning that focuses on capability oriented data construction and multitask alignment. The existing methods for long context reinforcement learning often result in homogeneous task coverage and reward formulations that do not accurately reflect real world requirements. To address this issue, the authors propose two main contributions.

First, they introduce a capability oriented data construction method that involves creating a dataset of 23,000 reinforcement learning samples with verifiable rewards, spanning 9 task types, each with its own evaluation metric. The dataset is openly released along with the construction pipeline and training code. The results show that this dataset outperforms a closed source dataset called QwenLong-L1.5 under the same training setup.

Second, the authors propose a new method called TMN-Reweight for heterogeneous multitask optimization. This method combines task level mean normalization for cross task reward scale alignment with difficulty adaptive weighting for more reliable advantage estimation. The results show that TMN-Reweight improves average performance over the vanilla GRPO method, while preserving or improving general capabilities across evaluations.

The authors also train a model called Qwen3-30B-A3B on the new dataset and achieve long context performance comparable to other state of the art models, such as DeepSeek-R1-0528 and Qwen3-235B-A22B-Thinking-2507. This suggests that the new dataset and TMN-Reweight method can substantially improve long context capability. Overall, the paper presents a new approach to long context reinforcement learning that focuses on capability oriented data construction and multitask alignment, and achieves state of the art results.

📅 Published on May 19

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2605.19577
• PDF: https://arxiv.org/pdf/2605.19577
• Project Page: https://huggingface.co/collections/Kwai-Klear/golongrl

🤖 Models citing this paper:
• https://huggingface.co/Kwai-Klear/GoLongRL-4B
• https://huggingface.co/Kwai-Klear/GoLongRL-30B-A3B

📊 Datasets citing this paper:
• https://huggingface.co/datasets/Kwai-Klear/GoLongRL

━━━━━━━━━━━━━━━━━━━━━━━━
📢 By: https://xn--r1a.website/PaperNexus

#ReinforcementLearning #LongContextLearning #MultitaskAlignment #CapabilityOrientedLearning #DeepLearning

🔥 GenEvolve: Self-Evolving Image Generation Agents via Tool-Orchestrated Visual Experience Distillation

💡 The paper proposes a self-evolving image generation framework called GenEvolve that improves generative capabilities through iterative learning and reference-based prompting. The problem addressed is that high-quality image generation often requires combining a model's internal generative ability with external resources, and existing methods have limitations in handling diverse and demanding requests.

The GenEvolve framework models each generation attempt as a tool-orchestrated trajectory, where the agent gathers evidence, selects references, invokes generation skills, and composes them into a prompt-reference program. Unlike existing methods that rely on image-level scalar rewards, GenEvolve compares multiple trajectories for the same request and abstracts best-worst differences into structured visual experience.

This visual experience is provided to a privileged teacher branch, which uses visual experience distillation to provide dense token-level supervision to a student branch. This helps the student internalize better search, knowledge activation, reference selection, and prompt construction. The authors also construct GenEvolve-Data and GenEvolve-Bench to evaluate the framework.

The results show that GenEvolve achieves substantial gains over strong baselines, achieving state-of-the-art performance among current image-generation frameworks. The experiments on public benchmarks and GenEvolve-Bench demonstrate the effectiveness of the proposed framework. Overall, the paper contributes a novel self-evolving image generation framework that can effectively handle diverse and demanding generation challenges.

📅 Published on May 20

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2605.21605
• PDF: https://arxiv.org/pdf/2605.21605
• Project Page: https://ephemeral182.github.io/GenEvolve/

🤖 Models citing this paper:
• https://huggingface.co/MeiGen-AI/GenEvolve

📊 Datasets citing this paper:
• https://huggingface.co/datasets/MeiGen-AI/GenEvolve-Data-Bench

━━━━━━━━━━━━━━━━━━━━━━━━
📢 By: https://xn--r1a.website/PaperNexus

#ComputerVision #ImageGeneration #GenerativeModels #SelfEvolvingSystems #DeepLearning

🔥 Representation Distribution Matching for One-Step Visual Generation

💡 The paper introduces Representation Distribution Matching, a method for one-step visual generation that matches feature distributions under pretrained encoders. The goal is to generate high-quality images by comparing the distributions of generated and reference features. The authors identify two key design axes: how the distributions are compared and the representations they are compared in. They conduct controlled studies and find three main results.

First, they show that the Maximum Mean Discrepancy, a classical method that was previously ineffective, becomes a strong and scalable objective when estimated correctly. Second, they find that the batch size of the generated images has a significant impact on performance, with an optimum batch size above 2048, which is much larger than typical batch sizes. Third, they demonstrate that using a single representation can be gamed, resulting in low scores despite visibly fake images, and instead propose using a balanced set of encoders and evaluating with a Sliced-Wasserstein distance over 14 encoders.

The authors combine these findings to develop an improved Representation Distribution Matching method, which they call iRDM. They evaluate iRDM on the ImageNet dataset and achieve state-of-the-art results, with a Sliced-Wasserstein distance of 1.30. Additionally, they use a human-preference proxy, called PickScore, which shows that iRDM is preferred over the previous best one-step generator on 71.2% of matched samples. They also apply the same method to post-train a four-step generator, called FLUX.2, and achieve better results than the original four-step version, with improved performance on GenEval and PickScore, and requiring only 90 GPU-hours. Overall, the paper presents a new method for one-step visual generation that achieves state-of-the-art results and can be used to improve existing generators.

📅 Published on Jul 2

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2607.02375
• PDF: https://arxiv.org/pdf/2607.02375
• Project Page: https://alan-lanfeng.github.io/rdm/

🤖 Models citing this paper:
• https://huggingface.co/epfl-vita/flux2-klein-1step-rdm

🚀 Spaces citing this paper:
• https://huggingface.co/spaces/epfl-vita/flux2-klein-1step-demo

━━━━━━━━━━━━━━━━━━━━━━━━
📢 By: https://xn--r1a.website/PaperNexus

#VisualGeneration #RepresentationLearning #DistributionMatching #ImageSynthesis #DeepLearning