✨RelayGen: Intra-Generation Model Switching for Efficient Reasoning

📝 Summary:
RelayGen is a training-free framework that dynamically switches between large and small models during reasoning by identifying difficulty transitions at the segment level, achieving faster inference w...

🔹 Publication Date: Published on Feb 6

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.06454
• PDF: https://arxiv.org/pdf/2602.06454
• Github: https://github.com/jiwonsong-dev/RelayGen

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Modality Gap-Driven Subspace Alignment Training Paradigm For Multimodal Large Language Models

📝 Summary:
Researchers address the modality gap in multimodal learning by proposing a fixed-frame theory and a training-free alignment method that enables efficient scaling of multimodal models using unpaired da...

🔹 Publication Date: Published on Feb 2

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.07026
• PDF: https://arxiv.org/pdf/2602.07026

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨ECO: Energy-Constrained Optimization with Reinforcement Learning for Humanoid Walking

📝 Summary:
Energy-constrained optimization framework separates energy metrics from rewards using Lagrangian method to achieve stable, energy-efficient humanoid robot locomotion with reduced hyperparameter tuning...

🔹 Publication Date: Published on Feb 6

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.06445
• PDF: https://arxiv.org/pdf/2602.06445

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Cybersecurity AI: Humanoid Robots as Attack Vectors

📝 Summary:
The Unitree G1 humanoid robot is vulnerable to BLE provisioning protocol exploits, exfiltrates sensor data, and can be repurposed for active cyber operations, highlighting the need for improved securi...

🔹 Publication Date: Published on Sep 17, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2509.14139
• PDF: https://arxiv.org/pdf/2509.14139
• Project Page: https://aliasrobotics.com
• Github: https://github.com/aliasrobotics/cai

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Towards Bridging the Gap between Large-Scale Pretraining and Efficient Finetuning for Humanoid Control

📝 Summary:
Off-policy Soft Actor-Critic with large-batch updates enables efficient humanoid locomotion policy pretraining, while model-based methods facilitate safe adaptation through deterministic data collecti...

🔹 Publication Date: Published on Jan 29

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.21363
• PDF: https://arxiv.org/pdf/2601.21363
• Github: https://github.com/bigai-ai/LIFT-humanoid

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨LOCA-bench: Benchmarking Language Agents Under Controllable and Extreme Context Growth

📝 Summary:
LOCA-bench is a new benchmark for evaluating language agents in long context, agentic scenarios with controlled environment state growth. It assesses how models and context management strategies perform as context extends, finding that advanced techniques significantly improve success rates.

🔹 Publication Date: Published on Feb 8

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.07962
• PDF: https://arxiv.org/pdf/2602.07962
• Github: https://github.com/hkust-nlp/LOCA-bench

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨LatentChem: From Textual CoT to Latent Thinking in Chemical Reasoning

📝 Summary:
LatentChem enables chemical reasoning through continuous latent space computations instead of discrete textual tokens, achieving superior performance and efficiency compared to traditional chain-of-th...

🔹 Publication Date: Published on Feb 6

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.07075
• PDF: https://arxiv.org/pdf/2602.07075

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨AgentCPM-Report: Interleaving Drafting and Deepening for Open-Ended Deep Research

📝 Summary:
AgentCPM-Report presents a lightweight local solution for deep research report generation using a Writing As Reasoning Policy framework and multi-stage agentic training to enhance small models' reason...

🔹 Publication Date: Published on Feb 6

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.06540
• PDF: https://arxiv.org/pdf/2602.06540
• Github: https://github.com/OpenBMB/AgentCPM

🔹 Models citing this paper:
• https://huggingface.co/openbmb/AgentCPM-Report
• https://huggingface.co/openbmb/AgentCPM-Report-GGUF

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨When and How Much to Imagine: Adaptive Test-Time Scaling with World Models for Visual Spatial Reasoning

📝 Summary:
Adaptive test-time framework with world models enables selective visual imagination for spatial reasoning, improving efficiency and reliability by determining when imagination is necessary. AI-generat...

🔹 Publication Date: Published on Feb 9

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.08236
• PDF: https://arxiv.org/pdf/2602.08236
• Project Page: https://adaptive-visual-tts.github.io/
• Github: https://adaptive-visual-tts.github.io/

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Recurrent-Depth VLA: Implicit Test-Time Compute Scaling of Vision-Language-Action Models via Latent Iterative Reasoning

📝 Summary:
RD-VLA introduces a recurrent architecture for VLA models, using latent iterative refinement for adaptive compute. It maintains constant memory, boosts success on complex tasks, and offers significant speedups.

🔹 Publication Date: Published on Feb 8

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.07845
• PDF: https://arxiv.org/pdf/2602.07845
• Project Page: https://rd-vla.github.io/
• Github: https://github.com/rd-vla/rd-vla

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Demo-ICL: In-Context Learning for Procedural Video Knowledge Acquisition

📝 Summary:
Researchers introduce a new video understanding task and benchmark that evaluates models' ability to learn from few-shot demonstrations, along with a specialized MLLM architecture trained using a two-...

🔹 Publication Date: Published on Feb 9

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.08439
• PDF: https://arxiv.org/pdf/2602.08439
• Github: https://github.com/dongyh20/Demo-ICL

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨QuantaAlpha: An Evolutionary Framework for LLM-Driven Alpha Mining

📝 Summary:
Financial markets are noisy and non-stationary, making alpha mining highly sensitive to noise in backtesting results and sudden market regime shifts. While recent agentic frameworks improve alpha mini...

🔹 Publication Date: Published on Feb 6

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.07085
• PDF: https://arxiv.org/pdf/2602.07085
• Github: https://github.com/QuantaAlpha/QuantaAlpha

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨LLaDA2.1: Speeding Up Text Diffusion via Token Editing

📝 Summary:
LLaDA2.1 introduces a novel token-to-token editing approach with speed and quality modes, enhanced through reinforcement learning for improved reasoning and instruction following in large language dif...

🔹 Publication Date: Published on Feb 9

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.08676
• PDF: https://arxiv.org/pdf/2602.08676
• Github: https://github.com/inclusionAI/LLaDA2.X

🔹 Models citing this paper:
• https://huggingface.co/inclusionAI/LLaDA2.1-mini
• https://huggingface.co/inclusionAI/LLaDA2.1-flash

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨WorldCompass: Reinforcement Learning for Long-Horizon World Models

📝 Summary:
WorldCompass enhances long-horizon video-based world models through reinforcement learning post-training with clip-level rollouts, complementary rewards, and efficient RL algorithms. AI-generated summ...

🔹 Publication Date: Published on Feb 9

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.09022
• PDF: https://arxiv.org/pdf/2602.09022
• Project Page: https://3d-models.hunyuan.tencent.com/world/

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨WildReward: Learning Reward Models from In-the-Wild Human Interactions

📝 Summary:
WildReward demonstrates that reward models can be effectively trained from in-the-wild user interactions using ordinal regression, achieving performance comparable to traditional methods while benefit...

🔹 Publication Date: Published on Feb 9

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.08829
• PDF: https://arxiv.org/pdf/2602.08829

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Reliable and Responsible Foundation Models: A Comprehensive Survey

📝 Summary:
Foundation models including LLMs, MLLMs, and generative models require reliable and responsible development addressing bias, security, explainability, and other critical issues for trustworthy deploym...

🔹 Publication Date: Published on Feb 4

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.08145
• PDF: https://arxiv.org/pdf/2602.08145

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨MOVA: Towards Scalable and Synchronized Video-Audio Generation

📝 Summary:
MOVA is an open-source model generating synchronized video-audio content, including lip-synced speech and sound effects. It employs a 32B-parameter Mixture-of-Experts architecture for image-text to video-audio generation, overcoming limitations of previous cascaded and closed-source systems.

🔹 Publication Date: Published on Feb 9

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.08794
• PDF: https://arxiv.org/pdf/2602.08794
• Project Page: https://mosi.cn/models/mova
• Github: https://github.com/OpenMOSS/MOVA

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨InternAgent-1.5: A Unified Agentic Framework for Long-Horizon Autonomous Scientific Discovery

📝 Summary:
InternAgent-1.5 is a unified system for autonomous scientific discovery that integrates computational modeling and experimental research through coordinated subsystems for generation, verification, an...

🔹 Publication Date: Published on Feb 9

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.08990
• PDF: https://arxiv.org/pdf/2602.08990

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨How2Everything: Mining the Web for How-To Procedures to Evaluate and Improve LLMs

📝 Summary:
A scalable framework for evaluating and improving goal-conditioned procedure generation using large-scale web mining, automated scoring, and reinforcement learning to enhance step-by-step instruction ...

🔹 Publication Date: Published on Feb 9

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.08808
• PDF: https://arxiv.org/pdf/2602.08808
• Github: https://github.com/lilakk/how2everything

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨GISA: A Benchmark for General Information-Seeking Assistant

📝 Summary:
A new benchmark called GISA is introduced for evaluating information-seeking assistants, featuring human-crafted queries with structured answer formats and live updates to prevent memorization. AI-gen...

🔹 Publication Date: Published on Feb 9

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.08543
• PDF: https://arxiv.org/pdf/2602.08543

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Rolling Sink: Bridging Limited-Horizon Training and Open-Ended Testing in Autoregressive Video Diffusion

📝 Summary:
Autoregressive video diffusion models suffer from train-test gaps when generating long videos, but a training-free approach called Rolling Sink addresses this by maintaining AR cache and enabling ultr...

🔹 Publication Date: Published on Feb 8

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.07775
• PDF: https://arxiv.org/pdf/2602.07775
• Project Page: https://rolling-sink.github.io/

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research