✨TRUST-SQL: Tool-Integrated Multi-Turn Reinforcement Learning for Text-to-SQL over Unknown Schemas

📝 Summary:
TRUST-SQL addresses unknown schema Text-to-SQL by employing a four-phase protocol and a Dual-Track GRPO strategy. This resolves credit assignment, achieving significant performance gains and matching baselines without pre-loaded metadata.

🔹 Publication Date: Published on Mar 17

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.16448
• PDF: https://arxiv.org/pdf/2603.16448

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Thinking in Uncertainty: Mitigating Hallucinations in MLRMs with Latent Entropy-Aware Decoding

📝 Summary:
Recent advancements in multimodal large reasoning models (MLRMs) have significantly improved performance in visual question answering. However, we observe that transition words (e.g., because, however...

🔹 Publication Date: Published on Mar 9

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.13366
• PDF: https://arxiv.org/pdf/2603.13366
• Project Page: https://mlrm-lead.github.io/
• Github: https://github.com/mlrm-LEAD/mlrm-LEAD

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨FinToolBench: Evaluating LLM Agents for Real-World Financial Tool Use

📝 Summary:
FinToolBench presents the first real-world benchmark for evaluating financial tool learning agents, featuring 760 executable tools and comprehensive evaluation criteria beyond simple execution success...

🔹 Publication Date: Published on Mar 9

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.08262
• PDF: https://arxiv.org/pdf/2603.08262
• Github: https://github.com/Double-wk/FinToolBench

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨One-Eval: An Agentic System for Automated and Traceable LLM Evaluation

📝 Summary:
One-Eval is an agentic evaluation system that automates large language model assessment by converting natural-language requests into executable workflows with integrated benchmark planning, dataset ha...

🔹 Publication Date: Published on Mar 10

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.09821
• PDF: https://arxiv.org/pdf/2603.09821
• Project Page: https://github.com/OpenDCAI/One-Eval
• Github: https://github.com/OpenDCAI/One-Eval

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨SK-Adapter: Skeleton-Based Structural Control for Native 3D Generation

📝 Summary:
SK-Adapter enables precise 3D structural control by treating skeletons as direct inputs through a lightweight adapter network that injects learnable tokens into frozen 3D generation models via cross-a...

🔹 Publication Date: Published on Mar 14

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.14152
• PDF: https://arxiv.org/pdf/2603.14152
• Project Page: https://sk-adapter.github.io/
• Github: https://github.com/sk-adapter/SK-Adapter

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨InCoder-32B: Code Foundation Model for Industrial Scenarios

📝 Summary:
InCoder-32B is a 32-billion-parameter code model for industrial programming tasks like chip design and GPU optimization. It was trained with extended context and execution verification, achieving strong performance on industrial benchmarks and competitive results on general tasks.

🔹 Publication Date: Published on Mar 17

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.16790
• PDF: https://arxiv.org/pdf/2603.16790
• Project Page: https://huggingface.co/Multilingual-Multimodal-NLP/IndustrialCoder
• Github: https://github.com/CSJianYang/Industrial-Coder

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Semi-Autonomous Formalization of the Vlasov-Maxwell-Landau Equilibrium

📝 Summary:
We present a complete Lean 4 formalization of the equilibrium characterization in the Vlasov-Maxwell-Landau (VML) system, which describes the motion of charged plasma. The project demonstrates the ful...

🔹 Publication Date: Published on Mar 16

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.15929
• PDF: https://arxiv.org/pdf/2603.15929
• Project Page: https://github.com/Vilin97/Clawristotle
• Github: https://github.com/Vilin97/Clawristotle

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨From Passive Observer to Active Critic: Reinforcement Learning Elicits Process Reasoning for Robotic Manipulation

📝 Summary:
Accurate process supervision remains a critical challenge for long-horizon robotic manipulation. A primary bottleneck is that current video MLLMs, trained primarily under a Supervised Fine-Tuning (SFT...

🔹 Publication Date: Published on Mar 16

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.15600
• PDF: https://arxiv.org/pdf/2603.15600
• Project Page: https://huggingface.co/collections/LeonOverload/primo-r1

🔹 Models citing this paper:
• https://huggingface.co/LeonOverload/PRIMO-R1-7B
• https://huggingface.co/LeonOverload/PRIMO-COT-SFT-7B

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Kinema4D: Kinematic 4D World Modeling for Spatiotemporal Embodied Simulation

📝 Summary:
Kinema4D is a 4D generative robotic simulator for precise robot-world interactions. It combines kinematic robot control with spatiotemporal environmental reaction synthesis. This enables physically plausible, embodiment-agnostic simulations with zero-shot transfer capability.

🔹 Publication Date: Published on Mar 17

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.16669
• PDF: https://arxiv.org/pdf/2603.16669
• Project Page: https://mutianxu.github.io/Kinema4D-project-page/
• Github: https://github.com/mutianxu/Kinema4D

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

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

#Robotics #Simulation #GenerativeAI #Kinematics #EmbodiedAI

✨Rethinking UMM Visual Generation: Masked Modeling for Efficient Image-Only Pre-training

📝 Summary:
IOMM is a data-efficient framework for UMM visual generation. It pre-trains with image-only data then fine-tunes with mixed data, achieving SOTA performance while significantly reducing computational costs.

🔹 Publication Date: Published on Mar 17

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.16139
• PDF: https://arxiv.org/pdf/2603.16139
• Github: https://github.com/LINs-lab/IOMM

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

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

#UMMVisualGeneration #MaskedModeling #EfficientAI #ComputerVision #GenerativeAI

✨Anticipatory Planning for Multimodal AI Agents

📝 Summary:
TraceR1 is a two-stage reinforcement learning framework for multimodal AI agents. It enhances planning by training anticipatory trajectory reasoning to forecast future actions and refine them with execution feedback. This significantly improves planning stability, execution robustness, and genera...

🔹 Publication Date: Published on Mar 17

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.16777
• PDF: https://arxiv.org/pdf/2603.16777

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

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

#AIAgents #MultimodalAI #ReinforcementLearning #AIPlanning #MachineLearning

✨ViT-AdaLA: Adapting Vision Transformers with Linear Attention

📝 Summary:
ViT-AdaLA adapts vision foundation models to linear attention Vision Transformers through attention alignment, feature alignment, and supervised fine-tuning to overcome quadratic complexity limitation...

🔹 Publication Date: Published on Mar 17

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.16063
• PDF: https://arxiv.org/pdf/2603.16063

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Test-Time Strategies for More Efficient and Accurate Agentic RAG

📝 Summary:
Test-time modifications to the Search-R1 pipeline improve retrieval efficiency and answer accuracy through contextualization and de-duplication modules. AI-generated summary Retrieval-Augmented Genera...

🔹 Publication Date: Published on Mar 12

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.12396
• PDF: https://arxiv.org/pdf/2603.12396

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨SWE-Skills-Bench: Do Agent Skills Actually Help in Real-World Software Engineering?

📝 Summary:
Research using SWE-Skills-Bench shows agent skills offer limited benefits in real-world software engineering. Most skills yield no improvement, with an average pass-rate gain of only 1.2 percent. Only specialized skills provide meaningful gains, while some can even degrade performance.

🔹 Publication Date: Published on Mar 16

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.15401
• PDF: https://arxiv.org/pdf/2603.15401
• Github: https://github.com/GeniusHTX/SWE-Skills-Bench

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

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

#SoftwareEngineering #AIagents #Benchmarking #AIresearch #LLM

✨VAREX: A Benchmark for Multi-Modal Structured Extraction from Documents

📝 Summary:
VAREX is a multimodal benchmark for structured data extraction from government forms. It provides four input modalities per document to systematically assess how input format affects extraction accuracy. Key findings show layout-preserving text significantly boosts accuracy and output compliance ...

🔹 Publication Date: Published on Mar 16

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.15118
• PDF: https://arxiv.org/pdf/2603.15118
• Project Page: https://udibarzi.github.io/varex-bench/
• Github: https://github.com/udibarzi/varex-bench

✨ Datasets citing this paper:
• https://huggingface.co/datasets/ibm-research/VAREX

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

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

#DataExtraction #MultimodalAI #DocumentAI #AIbenchmark #NLP

✨Qianfan-OCR: A Unified End-to-End Model for Document Intelligence

📝 Summary:
Qianfan-OCR is a 4B vision-language model that unifies document parsing, layout analysis, and understanding. It features Layout-as-Thought to improve accuracy on complex layouts and achieves state-of-the-art performance across multiple OCR and document intelligence benchmarks.

🔹 Publication Date: Published on Mar 11

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.13398
• PDF: https://arxiv.org/pdf/2603.13398
• Project Page: https://github.com/baidubce/Qianfan-VL
• Github: https://github.com/baidubce/Qianfan-VL

🔹 Models citing this paper:
• https://huggingface.co/baidu/Qianfan-OCR

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

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

#OCR #DocumentIntelligence #VisionLanguageModel #AI #MachineLearning

✨WiT: Waypoint Diffusion Transformers via Trajectory Conflict Navigation

📝 Summary:
Waypoint Diffusion Transformers WiT address trajectory conflicts in pixel-space flow matching using semantic waypoints from pre-trained vision models. WiT disentangles generation paths into segments, accelerating training convergence. It outperforms pixel-space baselines and speeds up JiT trainin...

🔹 Publication Date: Published on Mar 16

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.15132
• PDF: https://arxiv.org/pdf/2603.15132
• Project Page: https://hainuo-wang.github.io/WiT/
• Github: https://github.com/hainuo-wang/WiT

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

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

#DiffusionModels #Transformers #ComputerVision #DeepLearning #AI

✨GradMem: Learning to Write Context into Memory with Test-Time Gradient Descent

📝 Summary:
GradMem writes LLM context into memory efficiently via test-time gradient descent on memory tokens. It optimizes a reconstruction loss, outperforming forward-only methods in capacity and efficiency on synthetic and natural language tasks.

🔹 Publication Date: Published on Mar 14

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.13875
• PDF: https://arxiv.org/pdf/2603.13875
• Github: https://github.com/yurakuratov/gradmem

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

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

#LLM #GradientDescent #MachineLearning #NLP #AIResearch

✨SuperLocalMemory V3: Information-Geometric Foundations for Zero-LLM Enterprise Agent Memory

📝 Summary:
This paper establishes information-geometric foundations for AI agent memory. It introduces a new retrieval metric, principled lifecycle management, and formal contradiction detection, improving performance on benchmarks with a zero-LLM architecture.

🔹 Publication Date: Published on Mar 15

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.14588
• PDF: https://arxiv.org/pdf/2603.14588
• Github: https://github.com/qualixar/superlocalmemory

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

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

#AIAgents #AgentMemory #InformationGeometry #ZeroLLM #EnterpriseAI

✨Theoretical Foundations of Latent Posterior Factors: Formal Guarantees for Multi-Evidence Reasoning

📝 Summary:
Latent Posterior Factors LPF is a theoretical framework for trustworthy AI that combines heterogeneous evidence in probabilistic prediction tasks. It offers formal guarantees for key desiderata like calibration, error decay, and graceful degradation under corruption, all empirically validated.

🔹 Publication Date: Published on Mar 13

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.15674
• PDF: https://arxiv.org/pdf/2603.15674

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

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

#TrustworthyAI #AIResearch #ProbabilisticAI #MachineLearning #FormalGuarantees

✨I Know What I Don't Know: Latent Posterior Factor Models for Multi-Evidence Probabilistic Reasoning

📝 Summary:
This paper introduces Latent Posterior Factors LPF, a framework combining VAE latent posteriors with Sum-Product Network inference. LPF enables tractable probabilistic reasoning over unstructured evidence while maintaining calibrated uncertainty. It achieves high accuracy and low calibration erro...

🔹 Publication Date: Published on Mar 13

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.15670
• PDF: https://arxiv.org/pdf/2603.15670
• Github: https://github.com/aaaEpalea/epalea

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research