✨VLANeXt: Recipes for Building Strong VLA Models

📝 Summary:
This paper systematically analyzes Vision-Language-Action VLA models through a unified framework, distilling 12 key design principles. The resulting VLANeXt model achieves superior performance on benchmarks and strong real-world generalization.

🔹 Publication Date: Published on Feb 20

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.18532
• PDF: https://arxiv.org/pdf/2602.18532
• Project Page: https://dravenalg.github.io/VLANeXt/
• Github: https://github.com/DravenALG/awesome-vla

🔹 Models citing this paper:
• https://huggingface.co/DravenALG/VLANeXt

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

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

#VLANeXt #VLAModels #ComputerVision #Robotics #AIResearch

✨RoboCurate: Harnessing Diversity with Action-Verified Neural Trajectory for Robot Learning

📝 Summary:
RoboCurate enhances synthetic robot learning data by evaluating action quality through simulator replay consistency. It also augments observation diversity via image editing and video transfer techniques. This leads to substantial improvements in robot task success rates compared to using real da...

🔹 Publication Date: Published on Feb 21

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.18742
• PDF: https://arxiv.org/pdf/2602.18742
• Project Page: https://seungkukim.github.io/robocurate/

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

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

#RobotLearning #Robotics #SyntheticData #DataAugmentation #AI

✨Nacrith: Neural Lossless Compression via Ensemble Context Modeling and High-Precision CDF Coding

📝 Summary:
Nacrith is a lossless compression system combining an LLM, lightweight predictors, and arithmetic coding. It introduces innovations like enhanced CDF precision and a hybrid binary format, achieving state-of-the-art compression efficiency, significantly outperforming existing methods on various da...

🔹 Publication Date: Published on Feb 23

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.19626
• PDF: https://arxiv.org/pdf/2602.19626
• Project Page: https://github.com/robtacconelli/Nacrith-GPU
• Github: https://github.com/robtacconelli/Nacrith-GPU

✨ Datasets citing this paper:
• https://huggingface.co/datasets/robtacconelli/smollm2-135M-GGUF

✨ Spaces citing this paper:
• https://huggingface.co/spaces/robtacconelli/Nacrith-GPU

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

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

#LosslessCompression #NeuralNetworks #LLM #DataCompression #MachineLearning

✨Large Causal Models for Temporal Causal Discovery

📝 Summary:
Large Causal Models LCMs introduce a foundation-model approach for temporal causal discovery. They combine diverse synthetic and realistic time-series data, enabling scalable learning, improved generalization, and fast, single-pass inference beyond traditional methods.

🔹 Publication Date: Published on Feb 20

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.18662
• PDF: https://arxiv.org/pdf/2602.18662

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

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

#CausalDiscovery #FoundationModels #MachineLearning #TimeSeries #DataScience

✨ImplicitRDP: An End-to-End Visual-Force Diffusion Policy with Structural Slow-Fast Learning

📝 Summary:
ImplicitRDP is an end-to-end visual-force diffusion policy that integrates asynchronous vision and force sensing using structural slow-fast learning and virtual-target regularization. It improves reactivity and success in contact-rich manipulation tasks.

🔹 Publication Date: Published on Dec 11, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2512.10946
• PDF: https://arxiv.org/pdf/2512.10946
• Project Page: https://implicit-rdp.github.io
• Github: https://github.com/Chen-Wendi/ImplicitRDP

🔹 Models citing this paper:
• https://huggingface.co/WendiChen/ImplicitRDP_model

✨ Datasets citing this paper:
• https://huggingface.co/datasets/WendiChen/ImplicitRDP_dataset

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

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

#Robotics #DiffusionPolicy #MachineLearning #Manipulation #ComputerVision

✨Contact-Anchored Proprioceptive Odometry for Quadruped Robots

📝 Summary:
This purely proprioceptive state estimator uses IMU and motor data, treating contacting legs as kinematic anchors. It leverages footfall positions and height clustering to reduce pose and elevation drift without external sensors, showing low errors on quadruped robots.

🔹 Publication Date: Published on Feb 19

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.17393
• PDF: https://arxiv.org/pdf/2602.17393
• Project Page: https://github.com/ShineMinxing/Ros2Go2Estimator
• Github: https://github.com/ShineMinxing/Ros2Go2Estimator

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

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

#Robotics #QuadrupedRobots #StateEstimation #Odometry #ProprioceptiveSensing

✨DODO: Discrete OCR Diffusion Models

📝 Summary:
DODO introduces block discrete diffusion for optical character recognition to overcome the slowness of autoregressive models. It achieves near state-of-the-art accuracy and up to 3x faster inference by processing text in blocks, which resolves instability issues of global diffusion for precise OC...

🔹 Publication Date: Published on Feb 18

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.16872
• PDF: https://arxiv.org/pdf/2602.16872

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Anatomy of Agentic Memory: Taxonomy and Empirical Analysis of Evaluation and System Limitations

📝 Summary:
Agentic memory systems for LLM agents face empirical challenges including inadequate benchmarks, misaligned metrics, and performance variability that limit their practical effectiveness. AI-generated ...

🔹 Publication Date: Published on Feb 22

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.19320
• PDF: https://arxiv.org/pdf/2602.19320
• Github: https://github.com/FredJiang0324/Anatomy-of-Agentic-Memory

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Decoding ML Decision: An Agentic Reasoning Framework for Large-Scale Ranking System

📝 Summary:
GEARS presents a framework that reframes ranking optimization as an autonomous discovery process using specialized agent skills and validation hooks to balance algorithmic signals with ranking context...

🔹 Publication Date: Published on Feb 20

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.18640
• PDF: https://arxiv.org/pdf/2602.18640

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨SimToolReal: An Object-Centric Policy for Zero-Shot Dexterous Tool Manipulation

📝 Summary:
SimToolReal enables generalizable robot manipulation of diverse tools through procedural simulation and universal reinforcement learning policies without task-specific training. AI-generated summary T...

🔹 Publication Date: Published on Feb 18

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.16863
• PDF: https://arxiv.org/pdf/2602.16863
• Project Page: https://simtoolreal.github.io/
• Github: https://github.com/tylerlum/simtoolreal

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨On the "Induction Bias" in Sequence Models

📝 Summary:
Transformers require exponentially more data than RNNs for state tracking tasks. They also fail to share learned mechanisms across different sequence lengths, unlike RNNs which exhibit effective amortized learning by sharing weights. This reveals a fundamental in-distribution challenge for transf...

🔹 Publication Date: Published on Feb 20

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.18333
• PDF: https://arxiv.org/pdf/2602.18333

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Ani3DHuman: Photorealistic 3D Human Animation with Self-guided Stochastic Sampling

📝 Summary:
Ani3DHuman generates photorealistic 3D human animations by merging kinematics and video diffusion. It uses a layered motion representation and a novel self-guided stochastic sampling method to ensure photorealistic non-rigid motion and identity preservation.

🔹 Publication Date: Published on Feb 22

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.19089
• PDF: https://arxiv.org/pdf/2602.19089
• Github: https://github.com/qiisun/ani3dhuman

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨From Perception to Action: An Interactive Benchmark for Vision Reasoning

📝 Summary:
Current vision-language models struggle with physical structures and causal constraints for complex 3D tasks. The new CHAIN benchmark evaluates this capability, revealing that state-of-the-art models still fail to plan effective actions based on perceived physical structure.

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.21015
• PDF: https://arxiv.org/pdf/2602.21015
• Project Page: https://social-ai-studio.github.io/CHAIN/
• Github: https://social-ai-studio.github.io/CHAIN/

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨PyVision-RL: Forging Open Agentic Vision Models via RL

📝 Summary:
PyVision-RL framework addresses interaction collapse in multimodal models through enhanced reinforcement learning techniques and efficient video processing strategies. AI-generated summary Reinforceme...

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.20739
• PDF: https://arxiv.org/pdf/2602.20739
• Project Page: https://agent-x.space/pyvision-rl/
• Github: https://github.com/agents-x-project/PyVision-RL

🔹 Models citing this paper:
• https://huggingface.co/Agents-X/PyVision-Image-7B-SFT
• https://huggingface.co/Agents-X/PyVision-Image-7B-RL
• https://huggingface.co/Agents-X/PyVision-Video-7B-RL

✨ Datasets citing this paper:
• https://huggingface.co/datasets/Agents-X/PyVision-Image-SFT-Data
• https://huggingface.co/datasets/Agents-X/PyVision-Video-RL-Data
• https://huggingface.co/datasets/Agents-X/PyVision-Image-RL-Data

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨LongCLI-Bench: A Preliminary Benchmark and Study for Long-horizon Agentic Programming in Command-Line Interfaces

📝 Summary:
LongCLI-Bench evaluates AI agents' ability to complete complex, multi-step programming tasks through command-line interfaces with detailed failure analysis and human-agent collaboration insights. AI-g...

🔹 Publication Date: Published on Feb 15

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.14337
• PDF: https://arxiv.org/pdf/2602.14337
• Project Page: https://github.com/finyorko/longcli-bench
• Github: https://github.com/finyorko/longcli-bench

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Conv-FinRe: A Conversational and Longitudinal Benchmark for Utility-Grounded Financial Recommendation

📝 Summary:
A new conversational financial recommendation benchmark evaluates large language models' ability to balance rational decision-making with user behavior alignment using multi-view references derived fr...

🔹 Publication Date: Published on Feb 19

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.16990
• PDF: https://arxiv.org/pdf/2602.16990
• Github: https://github.com/The-FinAI/Conv-FinRe

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨FlowPrefill: Decoupling Preemption from Prefill Scheduling Granularity to Mitigate Head-of-Line Blocking in LLM Serving

📝 Summary:
FlowPrefill addresses head-of-line blocking in large language model serving by decoupling preemption granularity from scheduling frequency through operator-level preemption and event-driven scheduling...

🔹 Publication Date: Published on Feb 18

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.16603
• PDF: https://arxiv.org/pdf/2602.16603
• Github: https://github.com/HSIEHCHIACHI/FlowPrefill

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨The Art of Efficient Reasoning: Data, Reward, and Optimization

📝 Summary:
Large language models benefit from scaled chain-of-thought reasoning through efficient training methods that balance trajectory length and accuracy using reinforcement learning with reward shaping. AI...

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.20945
• PDF: https://arxiv.org/pdf/2602.20945

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Implicit Intelligence -- Evaluating Agents on What Users Don't Say

📝 Summary:
AI agents struggle to interpret implicitly specified real-world requests that require contextual reasoning beyond explicit instructions, as demonstrated by an evaluation framework using interactive YA...

🔹 Publication Date: Published on Feb 23

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.20424
• PDF: https://arxiv.org/pdf/2602.20424

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research