✨HyperAlign: Hypernetwork for Efficient Test-Time Alignment of Diffusion Models

📝 Summary:
HyperAlign uses a hypernetwork to efficiently align diffusion models at test-time. It dynamically adjusts denoising trajectories based on input conditions, improving semantic consistency and visual appeal. This outperforms existing methods.

🔹 Publication Date: Published on Jan 22

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.15968
• PDF: https://arxiv.org/pdf/2601.15968

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#DiffusionModels #Hypernetworks #GenerativeAI #AIResearch #DeepLearning

✨Towards Pixel-Level VLM Perception via Simple Points Prediction

📝 Summary:
SimpleSeg enables MLLMs to perform pixel-level segmentation by predicting point sequences in language space. A two-stage training with reinforcement learning refines these points. This simple method achieves competitive results, showing MLLMs have inherent low-level perception without specialized...

🔹 Publication Date: Published on Jan 27

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.19228
• PDF: https://arxiv.org/pdf/2601.19228
• Project Page: https://simpleseg.github.io/
• Github: https://github.com/songtianhui/SimpleSeg

🔹 Models citing this paper:
• https://huggingface.co/sthui/SimpleSeg-Kimi-VL
• https://huggingface.co/sthui/SimpleSeg-Qwen2.5-VL

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#VLM #MLLM #ImageSegmentation #DeepLearning #AIResearch

✨Youtu-VL: Unleashing Visual Potential via Unified Vision-Language Supervision

📝 Summary:
Youtu-VL introduces a Vision-Language Unified Autoregressive Supervision paradigm. It shifts from vision-as-input to vision-as-target, integrating visual tokens into the prediction stream. This improves multimodal comprehension and vision-centric task performance, fostering generalist visual agents.

🔹 Publication Date: Published on Jan 27

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.19798
• PDF: https://arxiv.org/pdf/2601.19798
• Project Page: https://youtu-tip.com/#llm
• Github: https://github.com/TencentCloudADP/youtu-vl

🔹 Models citing this paper:
• https://huggingface.co/tencent/Youtu-VL-4B-Instruct
• https://huggingface.co/tencent/Youtu-VL-4B-Instruct-GGUF
• https://huggingface.co/tencent/Youtu-Parsing

✨ Spaces citing this paper:
• https://huggingface.co/spaces/tencent/Youtu-Parsing

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#YoutuVL #VisionLanguage #MultimodalAI #ComputerVision #DeepLearning

✨CooperBench: Why Coding Agents Cannot be Your Teammates Yet

📝 Summary:
AI agents lack social intelligence for teamwork. CooperBench, a new collaborative coding benchmark, shows agents perform 30% worse together than individually. This 'curse of coordination' is due to poor communication, broken commitments, and incorrect expectations, calling for AI to develop socia...

🔹 Publication Date: Published on Jan 19

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.13295
• PDF: https://arxiv.org/pdf/2601.13295
• Project Page: https://cooperbench.com
• Github: https://github.com/cooperbench/CooperBench

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Self-Distillation Enables Continual Learning

📝 Summary:
Self-Distillation Fine-Tuning enables on-policy continual learning from demonstrations. It uses the model as its own teacher to acquire new skills while preserving prior knowledge. This method significantly reduces catastrophic forgetting and allows models to accumulate multiple skills over time.

🔹 Publication Date: Published on Jan 27

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.19897
• PDF: https://arxiv.org/pdf/2601.19897

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨GDCNet: Generative Discrepancy Comparison Network for Multimodal Sarcasm Detection

📝 Summary:
A multimodal sarcasm detection approach uses generative models to create stable semantic anchors and measures cross-modal discrepancies for improved accuracy and robustness. AI-generated summary Multi...

🔹 Publication Date: Published on Jan 28

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.20618
• PDF: https://arxiv.org/pdf/2601.20618

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Harder Is Better: Boosting Mathematical Reasoning via Difficulty-Aware GRPO and Multi-Aspect Question Reformulation

📝 Summary:
MathForge enhances mathematical reasoning in large models through a dual framework combining difficulty-aware policy optimization and multi-aspect question reformulation to address limitations in exis...

🔹 Publication Date: Published on Jan 28

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.20614
• PDF: https://arxiv.org/pdf/2601.20614
• Github: https://github.com/AMAP-ML/MathForge

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨RIR-Mega-Speech: A Reverberant Speech Corpus with Comprehensive Acoustic Metadata and Reproducible Evaluation

📝 Summary:
A large-scale reverberant speech corpus with detailed acoustic annotations is introduced to facilitate standardized comparison and reproduction of speech processing research. AI-generated summary Desp...

🔹 Publication Date: Published on Jan 25

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.19949
• PDF: https://arxiv.org/pdf/2601.19949
• Project Page: https://huggingface.co/datasets/mandipgoswami/rir-mega-speech

✨ Datasets citing this paper:
• https://huggingface.co/datasets/mandipgoswami/rirmega
• https://huggingface.co/datasets/mandipgoswami/rir-mega-speech

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Advancing Open-source World Models

📝 Summary:
LingBot-World is an open-source world simulator offering high-fidelity dynamics in diverse environments. It features long-term memory and real-time interactivity. This release empowers the community for applications like content creation, gaming, and robot learning.

🔹 Publication Date: Published on Jan 28

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.20540
• PDF: https://arxiv.org/pdf/2601.20540
• Project Page: https://technology.robbyant.com/lingbot-world
• Github: https://github.com/Robbyant/lingbot-world/

🔹 Models citing this paper:
• https://huggingface.co/robbyant/lingbot-world-base-cam

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨SketchDynamics: Exploring Free-Form Sketches for Dynamic Intent Expression in Animation Generation

📝 Summary:
Free-form sketching enables intuitive dynamic intent communication for automated content creation, bridging human intention and digital output in animation workflows. AI-generated summary Sketching pr...

🔹 Publication Date: Published on Jan 28

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.20622
• PDF: https://arxiv.org/pdf/2601.20622

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨DeepSeek-OCR 2: Visual Causal Flow

📝 Summary:
DeepSeek-OCR 2 introduces DeepEncoder V2 that dynamically reorders visual tokens based on semantic content, enabling more human-like causal reasoning in 2D image understanding through cascaded 1D caus...

🔹 Publication Date: Published on Jan 28

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.20552
• PDF: https://arxiv.org/pdf/2601.20552
• Github: https://github.com/deepseek-ai/DeepSeek-OCR-2

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Spark: Strategic Policy-Aware Exploration via Dynamic Branching for Long-Horizon Agentic Learning

📝 Summary:
Spark is a reinforcement learning framework that strategically allocates computational resources by branching at critical decision states, improving sample efficiency and generalization for long-horiz...

🔹 Publication Date: Published on Jan 28

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.20209
• PDF: https://arxiv.org/pdf/2601.20209

🔹 Models citing this paper:
• https://huggingface.co/Jinyang23/Spark-1.5B-ALFWorld
• https://huggingface.co/Jinyang23/Spark-1.5B-ScienceWorld
• https://huggingface.co/Jinyang23/Spark-1.5B-WebShop

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Linear representations in language models can change dramatically over a conversation

📝 Summary:
Linear representation directions in language models dynamically shift during conversations, affecting how factual information is encoded while preserving generic content, with implications for interpr...

🔹 Publication Date: Published on Jan 28

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.20834
• PDF: https://arxiv.org/pdf/2601.20834

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨SERA: Soft-Verified Efficient Repository Agents

📝 Summary:
Soft-Verified Efficient Repository Agents (SERA) enables cost-effective training of coding agents through supervised fine-tuning, achieving state-of-the-art performance while enabling specialization t...

🔹 Publication Date: Published on Jan 28

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.20789
• PDF: https://arxiv.org/pdf/2601.20789
• Github: https://github.com/allenai/SERA

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Innovator-VL: A Multimodal Large Language Model for Scientific Discovery

📝 Summary:
Innovator-VL demonstrates that principled training design and transparent methodology can achieve strong scientific intelligence with reduced data requirements while maintaining general vision perform...

🔹 Publication Date: Published on Jan 27

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.19325
• PDF: https://arxiv.org/pdf/2601.19325
• Project Page: https://innovatorlm.github.io/Innovator-VL
• Github: https://github.com/InnovatorLM/Innovator-VL

🔹 Models citing this paper:
• https://huggingface.co/InnovatorLab/Innovator-VL-8B-Instruct
• https://huggingface.co/InnovatorLab/Innovator-VL-8B-Thinking

✨ Datasets citing this paper:
• https://huggingface.co/datasets/InnovatorLab/Innovator-VL-Instruct-46M
• https://huggingface.co/datasets/InnovatorLab/EMVista
• https://huggingface.co/datasets/InnovatorLab/MolParse

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨OmegaUse: Building a General-Purpose GUI Agent for Autonomous Task Execution

📝 Summary:
OmegaUse is a general-purpose GUI agent model that achieves state-of-the-art performance on mobile and desktop platforms through a combination of high-quality data construction, decoupled training met...

🔹 Publication Date: Published on Jan 28

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.20380
• PDF: https://arxiv.org/pdf/2601.20380

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨SE-DiCoW: Self-Enrolled Diarization-Conditioned Whisper

📝 Summary:
SE-DiCoW improves speaker-attributed ASR by using diarization output to identify an enrollment segment for each speaker. This segment provides fixed conditioning in cross-attention layers, resolving ambiguities and significantly reducing transcription error rates compared to DiCoW.

🔹 Publication Date: Published on Jan 27

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.19194
• PDF: https://arxiv.org/pdf/2601.19194

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨UPLiFT: Efficient Pixel-Dense Feature Upsampling with Local Attenders

📝 Summary:
UPLiFT is an efficient iterative upsampling architecture with a Local Attender operator that creates dense features from visual backbones. It achieves state-of-the-art performance with lower inference costs than cross-attention methods, overcoming prior limitations.

🔹 Publication Date: Published on Jan 25

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.17950
• PDF: https://arxiv.org/pdf/2601.17950
• Project Page: https://www.cs.umd.edu/~mwalmer/uplift/
• Github: https://github.com/mwalmer-umd/UPLiFT/

🔹 Models citing this paper:
• https://huggingface.co/UPLiFT-upsampler/uplift_dinov2-s14
• https://huggingface.co/UPLiFT-upsampler/uplift_dinov3-splus16
• https://huggingface.co/UPLiFT-upsampler/uplift_sd1.5vae

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#ComputerVision #DeepLearning #FeatureUpsampling #AttentionMechanisms #EfficientAI

✨Shallow-π: Knowledge Distillation for Flow-based VLAs

📝 Summary:
Shallow-pi is a knowledge distillation framework that reduces transformer depth in vision-language-action models. It achieves over two times faster inference with less than one percent performance drop, enabling efficient real-world robotic deployment.

🔹 Publication Date: Published on Jan 28

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.20262
• PDF: https://arxiv.org/pdf/2601.20262
• Project Page: https://icsl-jeon.github.io/shallow-pi/
• Github: https://icsl-jeon.github.io/shallow-pi/

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

For more data science resources:
✓ https://t.iss.one/DataScienceT

#KnowledgeDistillation #Robotics #VLAModels #EfficientAI #DeepLearning