✨Revisiting Parameter Server in LLM Post-Training

📝 Summary:
On-Demand Communication (ODC) adapts parameter server principles to Fully Sharded Data Parallel training by replacing collective communication with point-to-point communication, improving device utili...

🔹 Publication Date: Published on Jan 27

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.19362
• PDF: https://arxiv.org/pdf/2601.19362
• Github: https://github.com/sail-sg/odc

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨GPCR-Filter: a deep learning framework for efficient and precise GPCR modulator discovery

📝 Summary:
GPCR-Filter is a deep learning framework that combines protein language models and graph neural networks to identify GPCR modulators with high accuracy and generalization across unseen receptors and l...

🔹 Publication Date: Published on Jan 27

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.19149
• PDF: https://arxiv.org/pdf/2601.19149

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨AdaReasoner: Dynamic Tool Orchestration for Iterative Visual Reasoning

📝 Summary:
AdaReasoner teaches multimodal models general tool use for visual reasoning using scalable data, reinforcement learning for tool selection, and adaptive learning. It dynamically orchestrates tools, generalizes to new ones, and achieves state-of-the-art performance on complex visual tasks.

🔹 Publication Date: Published on Jan 26

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.18631
• PDF: https://arxiv.org/pdf/2601.18631
• Project Page: https://adareasoner.github.io/
• Github: https://adareasoner.github.io

🔹 Models citing this paper:
• https://huggingface.co/AdaReasoner/AdaReasoner-7B-Randomized
• https://huggingface.co/AdaReasoner/AdaReasoner-TC-7B-Non-Randomized
• https://huggingface.co/AdaReasoner/AdaReasoner-7B-Non-Randomized

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps

📝 Summary:
A new solver, DPM-Solver, accelerates sampling from diffusion probabilistic models by analytically solving the diffusion ordinary differential equations, achieving high-quality results with fewer func...

🔹 Publication Date: Published on Jun 2, 2022

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2206.00927
• PDF: https://arxiv.org/pdf/2206.00927
• Project Page: https://huggingface.co/spaces/huggingface-projects/stable-diffusion-latent-upscaler
• Github: https://github.com/MaximeVandegar/Papers-in-100-Lines-of-Code/tree/main/DPM_Solver_A_Fast_ODE_Solver_for_Diffusion_Probabilistic_Model_Sampling_in_Around_10_Steps

🔹 Models citing this paper:
• https://huggingface.co/raisahil/scunge-model

✨ Spaces citing this paper:
• https://huggingface.co/spaces/huggingface-projects/stable-diffusion-latent-upscaler
• https://huggingface.co/spaces/Rooni/finetuned_diffusion
• https://huggingface.co/spaces/anzorq/finetuned_diffusion

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow

📝 Summary:
Rectified flow is a simple ODE-based method for efficient distribution transport and tasks like generative modeling and domain transfer, achieving high-quality results with minimal computational cost....

🔹 Publication Date: Published on Sep 7, 2022

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2209.03003
• PDF: https://arxiv.org/pdf/2209.03003
• Github: https://github.com/MaximeVandegar/Papers-in-100-Lines-of-Code/tree/main/Flow_Straight_and_Fast_Learning_to_Generate_and_Transfer_Data_with_Rectified_Flow

🔹 Models citing this paper:
• https://huggingface.co/nvidia/GR00T-N1.5-3B
• https://huggingface.co/XCLiu/2_rectified_flow_from_sd_1_5
• https://huggingface.co/XCLiu/instaflow_0_9B_from_sd_1_5

✨ Spaces citing this paper:
• https://huggingface.co/spaces/APGASU/FlowChef-InstaFlow-InverseProblem-Inpainting
• https://huggingface.co/spaces/APGASU/FlowChef-InstaFlow-Edit
• https://huggingface.co/spaces/XCLiu/InstaFlow

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨A Pragmatic VLA Foundation Model

📝 Summary:
A Vision-Language-Action model trained on extensive real-world robotic data demonstrates superior performance and generalization across multiple platforms while offering enhanced efficiency through op...

🔹 Publication Date: Published on Jan 26

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.18692
• PDF: https://arxiv.org/pdf/2601.18692
• Project Page: https://technology.robbyant.com/lingbot-vla
• Github: https://github.com/robbyant/lingbot-vla

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨FastNeRF: High-Fidelity Neural Rendering at 200FPS

📝 Summary:
FastNeRF enables high-speed rendering of photorealistic 3D environments by factorizing radiance maps for efficient pixel value estimation. AI-generated summary Recent work on Neural Radiance Fields ( ...

🔹 Publication Date: Published on Mar 18, 2021

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2103.10380
• PDF: https://arxiv.org/pdf/2103.10380
• Github: https://github.com/MaximeVandegar/Papers-in-100-Lines-of-Code/tree/main/FastNeRF_High_Fidelity_Neural_Rendering_at_200FPS

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨World Craft: Agentic Framework to Create Visualizable Worlds via Text

📝 Summary:
World Craft enables non-expert users to create executable and visualizable AI environments through textual denoscriptions by combining structured scaffolding and multi-agent intent analysis. AI-generate...

🔹 Publication Date: Published on Jan 14

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.09150
• PDF: https://arxiv.org/pdf/2601.09150
• Github: https://github.com/HerzogFL/World-Craft

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research

✨TriPlay-RL: Tri-Role Self-Play Reinforcement Learning for LLM Safety Alignment

📝 Summary:
TriPlay-RL is a closed-loop reinforcement learning framework for LLM safety alignment. It iteratively improves attacker, defender, and evaluator roles with near-zero manual annotation. This leads to better adversarial effectiveness, enhanced safety performance, and refined judgment.

🔹 Publication Date: Published on Jan 26

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.18292
• PDF: https://arxiv.org/pdf/2601.18292

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#LLM #ReinforcementLearning #AISafety #MachineLearning #SelfPlay

✨FABLE: Forest-Based Adaptive Bi-Path LLM-Enhanced Retrieval for Multi-Document Reasoning

📝 Summary:
FABLE is a new retrieval framework enhancing LLM-based multi-document reasoning through hierarchical forest indexes and a bi-path strategy. It outperforms traditional RAG with up to 94 percent token reduction, proving the ongoing need for structured retrieval.

🔹 Publication Date: Published on Jan 26

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.18116
• PDF: https://arxiv.org/pdf/2601.18116

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#LLM #InformationRetrieval #MultiDocumentReasoning #RAG #NLP

✨HalluCitation Matters: Revealing the Impact of Hallucinated References with 300 Hallucinated Papers in ACL Conferences

📝 Summary:
Hallucinated citations HalluCitation are a growing problem in NLP papers. This study found nearly 300 papers from 2024-2025 contain HalluCitations, with a rapid increase at EMNLP 2025, threatening scientific reliability and conference credibility.

🔹 Publication Date: Published on Jan 26

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.18724
• PDF: https://arxiv.org/pdf/2601.18724

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#HalluCitation #NLP #ResearchIntegrity #AI #AcademicPublishing

✨Benchmarks Saturate When The Model Gets Smarter Than The Judge

📝 Summary:
This paper introduces Omni-MATH-2, a manually audited mathematical benchmark dataset to reduce noise. It reveals that existing judges like Omni-Judge are highly inaccurate, masking real model performance differences. Accurate benchmarks require both high-quality datasets and more competent judges.

🔹 Publication Date: Published on Jan 27

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.19532
• PDF: https://arxiv.org/pdf/2601.19532

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#AI #MachineLearning #Benchmarking #ModelEvaluation #Datasets

✨Post-LayerNorm Is Back: Stable, ExpressivE, and Deep

📝 Summary:
Keel is a novel Post-LayerNorm Transformer using Highway-style connections instead of residual ones. This enables stable training of networks over 1000 layers deep, preventing gradient vanishing and improving expressivity for LLMs.

🔹 Publication Date: Published on Jan 27

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.19895
• PDF: https://arxiv.org/pdf/2601.19895

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#Transformers #DeepLearning #LLM #NeuralNetworks #AIResearch

✨EvolVE: Evolutionary Search for LLM-based Verilog Generation and Optimization

📝 Summary:
EvolVE improves LLM-based Verilog generation and optimization through evolutionary search. It uses MCTS for correctness and IGR for optimization, accelerated by STG. EvolVE achieves state-of-the-art performance and reduces PPA on industry-scale designs.

🔹 Publication Date: Published on Jan 26

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.18067
• PDF: https://arxiv.org/pdf/2601.18067
• Github: https://github.com/weiber2002/ICRTL

✨ Datasets citing this paper:
• https://huggingface.co/datasets/weiber2002/ICRTL

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#LLM #Verilog #EvolutionaryAlgorithms #HardwareDesign #AI

❗️LISA HELPS EVERYONE EARN MONEY!$29,000 HE'S GIVING AWAY TODAY!

Everyone can join his channel and make money! He gives away from $200 to $5.000 every day in his channel

https://news.1rj.ru/str/+HDFF3Mo_t68zNWQy

⚡️FREE ONLY FOR THE FIRST 500 SUBSCRIBERS! FURTHER ENTRY IS PAID! 👆👇

https://news.1rj.ru/str/+HDFF3Mo_t68zNWQy

✨DeFM: Learning Foundation Representations from Depth for Robotics

📝 Summary:
DeFM is a self-supervised foundation model for depth representation learning in robotics. It learns geometric and semantic features from 60M depth images, achieving state-of-the-art performance across diverse robotic tasks and strong sim-to-real generalization.

🔹 Publication Date: Published on Jan 26

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.18923
• PDF: https://arxiv.org/pdf/2601.18923
• Github: https://de-fm.github.io/

🔹 Models citing this paper:
• https://huggingface.co/leggedrobotics/defm

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

For more data science resources:
✓ https://news.1rj.ru/str/DataScienceT

#Robotics #FoundationModels #SelfSupervisedLearning #ComputerVision #MachineLearning

✨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://news.1rj.ru/str/DataScienceT

#DiffusionModels #Hypernetworks #GenerativeAI #AIResearch #DeepLearning