✨UDM-GRPO: Stable and Efficient Group Relative Policy Optimization for Uniform Discrete Diffusion Models

📝 Summary:
UDM-GRPO integrates Uniform Discrete Diffusion Models with reinforcement learning, solving training instability issues. It optimizes using final samples as actions and reconstructed trajectories. This achieves state-of-the-art performance in text-to-image generation and OCR tasks.

🔹 Publication Date: Published on Apr 20

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.18518
• PDF: https://arxiv.org/pdf/2604.18518
• Project Page: https://yovecent.github.io/UDM-GRPO.github.io/
• Github: https://github.com/Yovecent/UDM-GRPO

🔹 Models citing this paper:
• https://huggingface.co/Yovecents/URSA-1.7B-IBQ512-UDMGRPO-GenEval
• https://huggingface.co/Yovecents/URSA-1.7B-IBQ512-UDMGRPO-PickScore

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

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

#DiffusionModels #ReinforcementLearning #GenerativeAI #TextToImage #DeepLearning

✨Mitigating Multimodal Hallucination via Phase-wise Self-reward

📝 Summary:
PSRD is a new self-rewarding framework that mitigates vision hallucination in LVLMs dynamically during inference. It uses phase-wise self-reward signals and a lightweight reward model for efficient online correction, significantly reducing hallucination rates.

🔹 Publication Date: Published on Apr 20

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.17982
• PDF: https://arxiv.org/pdf/2604.17982

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨The Cognitive Penalty: Ablating System 1 and System 2 Reasoning in Edge-Native SLMs for Decentralized Consensus

📝 Summary:
Research on SLMs in decentralized organizations finds that System 1 reasoning is superior for robust adversarial governance. System 2 inference-time compute introduces catastrophic instability, high latency, and vulnerabilities, making intuitive reasoning more effective.

🔹 Publication Date: Published on Apr 18

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.16913
• PDF: https://arxiv.org/pdf/2604.16913
• Github: https://github.com/smarizvi110/sentinel-bench

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

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

#SLMs #DecentralizedAI #CognitiveAI #AIGovernance #Blockchain

✨Chain-of-Thought Degrades Visual Spatial Reasoning Capabilities of Multimodal LLMs

📝 Summary:
Chain-of-Thought prompting in multimodal reasoning models degrades performance in visual spatial reasoning due to shortcut learning and hallucination of visual details from text alone. AI-generated su...

🔹 Publication Date: Published on Apr 17

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.16060
• PDF: https://arxiv.org/pdf/2604.16060

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Mind's Eye: A Benchmark of Visual Abstraction, Transformation and Composition for Multimodal LLMs

📝 Summary:
Multimodal large language models demonstrate significant limitations in visuospatial reasoning tasks compared to human performance, revealing deficiencies in visual attention, perceptual manipulation,...

🔹 Publication Date: Published on Apr 17

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.16054
• PDF: https://arxiv.org/pdf/2604.16054

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨ShadowPEFT: Shadow Network for Parameter-Efficient Fine-Tuning

📝 Summary:
ShadowPEFT is a new parameter-efficient fine-tuning framework that uses a depth-shared shadow module for layer-level refinement. This shifts adaptation from distributed weight perturbations to a shared layer-space process, matching or outperforming LoRA with reduced overhead and increased flexibi...

🔹 Publication Date: Published on Apr 21

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.19254
• PDF: https://arxiv.org/pdf/2604.19254
• Project Page: https://github.com/ShadowLLM/shadow-peft
• Github: https://github.com/ShadowLLM/shadow-peft

🔹 Models citing this paper:
• https://huggingface.co/shadow-llm/Qwen3-4B-GSM8k-Shadow
• https://huggingface.co/shadow-llm/Qwen3-4B-SquadV2-Shadow
• https://huggingface.co/shadow-llm/Qwen3-4B-MMLU-Shadow

✨ Datasets citing this paper:
• https://huggingface.co/datasets/shadow-llm/robot-dog-skills

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

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

#PEFT #FineTuning #MachineLearning #AI #LLMs

✨HP-Edit: A Human-Preference Post-Training Framework for Image Editing

📝 Summary:
A post-training framework called HP-Edit is introduced to align image editing models with human preferences using a novel automatic evaluator and a real-world dataset, improving editing quality throug...

🔹 Publication Date: Published on Apr 21

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.19406
• PDF: https://arxiv.org/pdf/2604.19406

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Understanding and Enforcing Weight Disentanglement in Task Arithmetic

📝 Summary:
Task arithmetic lacks theoretical explanation for its success, but the proposed OrthoReg method addresses this by promoting weight disentanglement through enforced orthogonality in weight updates duri...

🔹 Publication Date: Published on Apr 18

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.17078
• PDF: https://arxiv.org/pdf/2604.17078
• Github: https://github.com/RL-MIND/OrthoReg

🔹 Models citing this paper:
• https://huggingface.co/RL-MIND/OrthoReg_checkpoints
• https://huggingface.co/RL-MIND/OrthoReg

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨AJ-Bench: Benchmarking Agent-as-a-Judge for Environment-Aware Evaluation

📝 Summary:
Agent-as-a-Judge benchmark evaluates automated verification capabilities across multiple domains with comprehensive task assessment. AI-generated summary As reinforcement learning continues to scale t...

🔹 Publication Date: Published on Apr 20

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.18240
• PDF: https://arxiv.org/pdf/2604.18240
• Project Page: https://aj-bench.github.io/
• Github: https://github.com/aj-bench/AJ-Bench

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Accurate and scalable exchange-correlation with deep learning

📝 Summary:
A deep learning approach to density functional theory achieves higher accuracy than traditional methods while maintaining computational efficiency by learning electronic structure representations dire...

🔹 Publication Date: Published on Apr 21

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2506.14665
• PDF: https://arxiv.org/pdf/2506.14665
• Project Page: https://aka.ms/dft
• Github: https://github.com/microsoft/skala

🔹 Models citing this paper:
• https://huggingface.co/microsoft/skala-1.0
• https://huggingface.co/microsoft/skala-1.1

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨What Makes an LLM a Good Optimizer? A Trajectory Analysis of LLM-Guided Evolutionary Search

📝 Summary:
LLM-guided evolutionary search shows that optimization success depends on search trajectory characteristics rather than initial problem-solving ability alone, with strong optimizers refining locally w...

🔹 Publication Date: Published on Apr 21

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.19440
• PDF: https://arxiv.org/pdf/2604.19440
• Project Page: https://xinhao-zhang.github.io/traj_evo_search/
• Github: https://github.com/XINHAO-ZHANG/LLMEvo_Eval

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

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

#LLM #Optimization #EvolutionaryAlgorithms #AI #MachineLearning

✨MoVE: Translating Laughter and Tears via Mixture of Vocalization Experts in Speech-to-Speech Translation

📝 Summary:
MoVE, a Mixture-of-LoRA-Experts architecture with expressive-specialized adapters and a soft-weighting router, enables effective speech-to-speech translation with preserved non-verbal vocalizations wh...

🔹 Publication Date: Published on Apr 19

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.17435
• PDF: https://arxiv.org/pdf/2604.17435
• Github: https://github.com/47zzz/MoVE

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Micro Language Models Enable Instant Responses

📝 Summary:
Micro language models enable instant on-device response initiation with cloud-based continuation, achieving low-latency interactive AI through asymmetric collaboration between edge and cloud computing...

🔹 Publication Date: Published on Apr 21

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.19642
• PDF: https://arxiv.org/pdf/2604.19642
• Github: https://github.com/Sensente/micro_language_model_swen_project

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨CityRAG: Stepping Into a City via Spatially-Grounded Video Generation

📝 Summary:
CityRAG generates long-term, physically grounded video sequences that maintain environmental consistency and support complex navigation through real-world geography using geo-registered data as contex...

🔹 Publication Date: Published on Apr 21

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.19741
• PDF: https://arxiv.org/pdf/2604.19741
• Project Page: https://cityrag.github.io/

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

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

#VideoGeneration #GenerativeAI #SpatialAI #ComputerVision #UrbanSimulation

✨RDP LoRA: Geometry-Driven Identification for Parameter-Efficient Adaptation in Large Language Models

📝 Summary:
Using geometric trajectory analysis with the Ramer-Douglas-Peucker algorithm to select optimal layers for parameter-efficient fine-tuning of large language models, achieving better performance than fu...

🔹 Publication Date: Published on Apr 21

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.19321
• PDF: https://arxiv.org/pdf/2604.19321

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Cortex 2.0: Grounding World Models in Real-World Industrial Deployment

📝 Summary:
Cortex 2.0 introduces a plan-and-act control system for reliable long-horizon robotic manipulation. It generates and evaluates future trajectories in visual latent space, outperforming reactive Vision-Language-Action models. This demonstrates world-model-based planning's reliability in complex in...

🔹 Publication Date: Published on Apr 22

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.20246
• PDF: https://arxiv.org/pdf/2604.20246

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨SWE-chat: Coding Agent Interactions From Real Users in the Wild

📝 Summary:
SWE-chat presents a large-scale dataset of real coding agent interactions that reveals significant inefficiencies and challenges in current AI-assisted development practices. AI-generated summary A I ...

🔹 Publication Date: Published on Apr 22

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.20779
• PDF: https://arxiv.org/pdf/2604.20779

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨LLaDA2.0-Uni: Unifying Multimodal Understanding and Generation with Diffusion Large Language Model

📝 Summary:
LLaDA2.0-Uni is a unified discrete diffusion language model that integrates multimodal understanding and generation through a semantic discrete tokenizer, MoE-based backbone, and diffusion decoder, ac...

🔹 Publication Date: Published on Apr 22

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.20796
• PDF: https://arxiv.org/pdf/2604.20796
• Github: https://github.com/inclusionAI/LLaDA2.0-Uni

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨DR-Venus: Towards Frontier Edge-Scale Deep Research Agents with Only 10K Open Data

📝 Summary:
DR-Venus-4B is a 4-billion-parameter deep research agent trained entirely on open data using agentic supervised fine-tuning and reinforcement learning with turn-level rewards to achieve superior perfo...

🔹 Publication Date: Published on Apr 21

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.19859
• PDF: https://arxiv.org/pdf/2604.19859
• Project Page: https://huggingface.co/collections/inclusionAI/dr-venus
• Github: https://github.com/inclusionAI/DR-Venus/tree/master/Inference

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨WavAlign: Enhancing Intelligence and Expressiveness in Spoken Dialogue Models via Adaptive Hybrid Post-Training

📝 Summary:
Spoken dialogue models face challenges in expressiveness despite end-to-end approaches, but a modality-aware adaptive post-training method using constrained preference updates and explicit anchoring i...

🔹 Publication Date: Published on Apr 16

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.14932
• PDF: https://arxiv.org/pdf/2604.14932

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research