✨MindWatcher: Toward Smarter Multimodal Tool-Integrated Reasoning

📝 Summary:
MindWatcher is a tool-integrated reasoning agent using interleaved thinking and multimodal chain-of-thought. It autonomously coordinates diverse tools for complex tasks without human prompts. It outperforms larger models and provides agent training insights.

🔹 Publication Date: Published on Dec 29, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2512.23412
• PDF: https://arxiv.org/pdf/2512.23412
• Github: https://github.com/TIMMY-CHAN/MindWatcher

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨MDAgent2: Large Language Model for Code Generation and Knowledge Q&A in Molecular Dynamics

📝 Summary:
MDAgent2 enables automated molecular dynamics code generation and question answering through domain-adapted language models and a multi-agent runtime system. AI-generated summary Molecular dynamics (M...

🔹 Publication Date: Published on Jan 5

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.02075
• PDF: https://arxiv.org/pdf/2601.02075
• Github: https://github.com/FredericVAN/PKU_MDAgent2

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Choreographing a World of Dynamic Objects

📝 Summary:
CHORD is a universal generative framework that extracts Lagrangian motion information from Eulerian video representations to synthesize diverse 4D dynamic scenes without requiring category-specific ru...

🔹 Publication Date: Published on Jan 7

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.04194
• PDF: https://arxiv.org/pdf/2601.04194
• Project Page: https://yanzhelyu.github.io/chord/

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨EpiQAL: Benchmarking Large Language Models in Epidemiological Question Answering for Enhanced Alignment and Reasoning

📝 Summary:
EpiQAL presents a novel benchmark for evaluating epidemiological reasoning in language models through three distinct subsets measuring factual recall, multi-step inference, and conclusion reconstructi...

🔹 Publication Date: Published on Jan 6

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.03471
• PDF: https://arxiv.org/pdf/2601.03471

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨E-GRPO: High Entropy Steps Drive Effective Reinforcement Learning for Flow Models

📝 Summary:
Entropy-aware policy optimization method for reinforcement learning in flow matching models that improves exploration through SDE and ODE sampling strategies. AI-generated summary Recent reinforcement...

🔹 Publication Date: Published on Jan 1

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.00423
• PDF: https://arxiv.org/pdf/2601.00423
• Github: https://github.com/shengjun-zhang/VisualGRPO

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Agentic Rubrics as Contextual Verifiers for SWE Agents

📝 Summary:
Agentic Rubrics enable efficient and scalable verification for software engineering agents by creating context-aware checklists that outperform traditional methods while maintaining interpretability. ...

🔹 Publication Date: Published on Jan 7

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.04171
• PDF: https://arxiv.org/pdf/2601.04171

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Klear: Unified Multi-Task Audio-Video Joint Generation

📝 Summary:
Klear addresses audio-video joint generation challenges through a unified model architecture, progressive multitask training, and large-scale dense-caption data construction, achieving superior alignm...

🔹 Publication Date: Published on Jan 7

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.04151
• PDF: https://arxiv.org/pdf/2601.04151

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨RedBench: A Universal Dataset for Comprehensive Red Teaming of Large Language Models

📝 Summary:
RedBench presents a unified dataset with standardized risk categorization for evaluating LLM vulnerabilities across multiple domains and attack types. AI-generated summary As large language models (LL...

🔹 Publication Date: Published on Jan 7

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.03699
• PDF: https://arxiv.org/pdf/2601.03699
• Github: https://github.com/knoveleng/redeval

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Entropy-Adaptive Fine-Tuning: Resolving Confident Conflicts to Mitigate Forgetting

📝 Summary:
Supervised Fine-Tuning causes catastrophic forgetting due to 'Confident Conflicts.' Entropy-Adaptive Fine-Tuning EAFT uses token-level entropy to distinguish uncertainty from knowledge conflict. EAFT suppresses conflicting gradients, mitigating forgetting while matching performance.

🔹 Publication Date: Published on Jan 5

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.02151
• PDF: https://arxiv.org/pdf/2601.02151
• Project Page: https://ymxyll.github.io/EAFT/
• Github: https://ymxyll.github.io/EAFT/

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Benchmark^2: Systematic Evaluation of LLM Benchmarks

📝 Summary:
Researchers developed Benchmark^2, a framework with three metrics to evaluate benchmark quality for large language models, revealing significant variations in existing benchmarks and enabling more eff...

🔹 Publication Date: Published on Jan 7

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.03986
• PDF: https://arxiv.org/pdf/2601.03986

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨ThinkRL-Edit: Thinking in Reinforcement Learning for Reasoning-Centric Image Editing

📝 Summary:
ThinkRL-Edit enhances reasoning-centric image editing through reinforcement learning by expanding visual reasoning exploration beyond denoising stochasticity and using unbiased reward strategies. AI-g...

🔹 Publication Date: Published on Jan 6

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.03467
• PDF: https://arxiv.org/pdf/2601.03467

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Atlas: Orchestrating Heterogeneous Models and Tools for Multi-Domain Complex Reasoning

📝 Summary:
ATLAS is a dual-path framework that dynamically selects optimal model-tool combinations for complex reasoning. It uses cluster-based routing for domain-specific tasks and RL-based multi-step routing for generalization. ATLAS outperforms GPT-4o and other methods on diverse benchmarks.

🔹 Publication Date: Published on Jan 7

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.03872
• PDF: https://arxiv.org/pdf/2601.03872

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨MAGMA: A Multi-Graph based Agentic Memory Architecture for AI Agents

📝 Summary:
MAGMA is a multi-graph memory architecture that improves AI agent long-context reasoning. It decouples memory representation from retrieval logic across semantic, temporal, causal, and entity graphs for query-adaptive selection, outperforming existing agentic memory systems.

🔹 Publication Date: Published on Jan 6

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.03236
• PDF: https://arxiv.org/pdf/2601.03236

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

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

#AIAgents #MemoryArchitecture #LongContextReasoning #GraphAI #ArtificialIntelligence

✨SimpleMem: Efficient Lifelong Memory for LLM Agents

📝 Summary:
SimpleMem is an efficient memory framework for LLM agents that uses semantic lossless compression. It employs a three-stage pipeline to distill, consolidate, and retrieve historical experiences efficiently. SimpleMem significantly improves accuracy and reduces token consumption by up to 30-fold c...

🔹 Publication Date: Published on Jan 5

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.02553
• PDF: https://arxiv.org/pdf/2601.02553
• Project Page: https://aiming-lab.github.io/SimpleMem-Page/
• Github: https://aiming-lab.github.io/SimpleMem-Page/

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

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

#LLM #AIAgents #LifelongLearning #AI #DeepLearning

✨RGS-SLAM: Robust Gaussian Splatting SLAM with One-Shot Dense Initialization

📝 Summary:
RGS-SLAM is a robust Gaussian-splatting SLAM framework that uses a one-shot, correspondence-to-Gaussian initialization with DINOv3 descriptors. This method improves stability, accelerates convergence, and yields higher rendering fidelity and accuracy compared to existing systems.

🔹 Publication Date: Published on Dec 28, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.00705
• PDF: https://arxiv.org/pdf/2601.00705

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

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

#SLAM #GaussianSplatting #ComputerVision #Robotics #DeepLearning

✨Enhancing Linguistic Competence of Language Models through Pre-training with Language Learning Tasks

📝 Summary:
L2T, a new pre-training framework, integrates language learning tasks with next-token prediction. This enhances language models' linguistic competence and acquisition without sacrificing general reasoning.

🔹 Publication Date: Published on Jan 6

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.03448
• PDF: https://arxiv.org/pdf/2601.03448
• Project Page: https://huggingface.co/l2t-project/l2t-500m-disjoint-mix_75
• Github: https://github.com/gucci-j/l2t

🔹 Models citing this paper:
• https://huggingface.co/l2t-project/l2t-500m-disjoint
• https://huggingface.co/l2t-project/l2t-500m-disjoint-mix_0
• https://huggingface.co/l2t-project/l2t-500m-disjoint-mix_25

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

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

#LLM #NLP #Pretraining #LanguageLearning #AI

✨Why LLMs Aren't Scientists Yet: Lessons from Four Autonomous Research Attempts

📝 Summary:
A case study of four LLM agent attempts to autonomously generate ML research papers reveals six recurring failure modes. Most attempts failed, though one was accepted to a special AI-first author venue, leading to proposed design principles for future AI-scientist systems.

🔹 Publication Date: Published on Jan 6

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.03315
• PDF: https://arxiv.org/pdf/2601.03315

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

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

#LLMs #AIResearch #MachineLearning #AIAgents #AutonomousSystems

✨Gen3R: 3D Scene Generation Meets Feed-Forward Reconstruction

📝 Summary:
Gen3R combines reconstruction and video diffusion models to generate 3D scenes. It produces RGB videos and 3D geometry by aligning geometric and appearance latents. This achieves state-of-the-art results and improves reconstruction robustness.

🔹 Publication Date: Published on Jan 7

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.04090
• PDF: https://arxiv.org/pdf/2601.04090
• Project Page: https://xdimlab.github.io/Gen3R/
• Github: https://xdimlab.github.io/Gen3R/

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

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

#3DGeneration #DiffusionModels #ComputerVision #3DReconstruction #DeepLearning

✨Evolving Programmatic Skill Networks

📝 Summary:
The Programmatic Skill Network PSN enables continual skill acquisition through executable symbolic programs that evolve via reflection, progressive optimization, and structural refactoring. This framework demonstrates robust skill reuse, rapid adaptation, and strong generalization in open-ended e...

🔹 Publication Date: Published on Jan 7

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.03509
• PDF: https://arxiv.org/pdf/2601.03509

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

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

#ProgrammaticAI #SkillAcquisition #EvolutionaryAI #MachineLearning #AIResearch

✨Pearmut: Human Evaluation of Translation Made Trivial

📝 Summary:
Pearmut is a lightweight platform that simplifies complex human evaluation for multilingual NLP, particularly machine translation. It removes setup barriers by supporting various protocols, document context, and learning strategies. This makes reliable human evaluation a routine and practical par...

🔹 Publication Date: Published on Jan 6

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.02933
• PDF: https://arxiv.org/pdf/2601.02933
• Github: https://github.com/zouharvi/pearmut

✨ Datasets citing this paper:
• https://huggingface.co/datasets/zouharvi/hearing2translate-humeval

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research