✨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

✨On Data Engineering for Scaling LLM Terminal Capabilities

📝 Summary:
Researchers developed a synthetic task generation pipeline and analyzed data strategies to improve terminal agent performance, creating a large-scale dataset and models that outperform larger counterp...

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.21193
• PDF: https://arxiv.org/pdf/2602.21193
• Project Page: https://huggingface.co/collections/nvidia/nemotron-terminal

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Learning from Trials and Errors: Reflective Test-Time Planning for Embodied LLMs

📝 Summary:
Reflective Test-Time Planning enhances robot decision-making by integrating multiple reflection mechanisms that enable learning from experience and improving long-horizon task performance. AI-generate...

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.21198
• PDF: https://arxiv.org/pdf/2602.21198
• Project Page: https://reflective-test-time-planning.github.io/

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Aletheia tackles FirstProof autonomously

📝 Summary:
We report the performance of Aletheia (Feng et al., 2026b), a mathematics research agent powered by Gemini 3 Deep Think, on the inaugural FirstProof challenge. Within the allowed timeframe of the chal...

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.21201
• PDF: https://arxiv.org/pdf/2602.21201
• Project Page: https://github.com/google-deepmind/superhuman/tree/main/aletheia

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨The Diffusion Duality, Chapter II: Ψ-Samplers and Efficient Curriculum

📝 Summary:
Discrete diffusion models with predictor-corrector samplers surpass traditional methods in generation quality and efficiency, challenging assumptions about masked diffusion's necessity in language mod...

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.21185
• PDF: https://arxiv.org/pdf/2602.21185
• Project Page: https://s-sahoo.com/duo-ch2/

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Test-Time Training with KV Binding Is Secretly Linear Attention

📝 Summary:
This paper reinterprets Test-Time Training TTT with KV binding. Instead of memorization, it shows TTT is a form of learned linear attention with enhanced representational capacity. This new perspective explains puzzling behaviors, simplifies architectures, and boosts efficiency.

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.21204
• PDF: https://arxiv.org/pdf/2602.21204
• Project Page: https://research.nvidia.com/labs/sil/projects/tttla/

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Generative AI and Machine Learning Collaboration for Container Dwell Time Prediction via Data Standardization

📝 Summary:
A collaborative framework integrating generative artificial intelligence with machine learning improves container dwell time prediction by standardizing unstructured text data, leading to reduced reha...

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.20540
• PDF: https://arxiv.org/pdf/2602.20540

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨DREAM: Deep Research Evaluation with Agentic Metrics

📝 Summary:
Deep Research Agents generate analyst-grade reports, yet evaluating them remains challenging due to the absence of a single ground truth and the multidimensional nature of research quality. Recent ben...

🔹 Publication Date: Published on Feb 21

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.18940
• PDF: https://arxiv.org/pdf/2602.18940

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Untied Ulysses: Memory-Efficient Context Parallelism via Headwise Chunking

📝 Summary:
UPipe enables efficient processing of long sequences in Transformer models through fine-grained chunking at the attention head level, significantly reducing activation memory usage while maintaining t...

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.21196
• PDF: https://arxiv.org/pdf/2602.21196
• Project Page: https://rghadia.github.io/untied_ulysses_proj/
• Github: https://github.com/togethercomputer/Untied-Ulysses

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨OCR-Agent: Agentic OCR with Capability and Memory Reflection

📝 Summary:
A novel iterative self-correction framework enhances vision-language models' reasoning robustness through capability reflection and memory reflection mechanisms, achieving superior performance on visu...

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.21053
• PDF: https://arxiv.org/pdf/2602.21053
• Github: https://github.com/AIGeeksGroup/OCR-Agent

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨OmniOCR: Generalist OCR for Ethnic Minority Languages

📝 Summary:
OmniOCR presents a universal framework for ethnic minority scripts using Dynamic LoRA and sparsity regularization to achieve state-of-the-art accuracy with improved parameter efficiency in low-resourc...

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.21042
• PDF: https://arxiv.org/pdf/2602.21042
• Github: https://github.com/AIGeeksGroup/OmniOCR

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨LaS-Comp: Zero-shot 3D Completion with Latent-Spatial Consistency

📝 Summary:
LaS-Comp is a zero-shot 3D shape completion method that leverages 3D foundation models. It uses a two-stage approach for faithful reconstruction and seamless boundary refinement. This training-free framework outperforms prior state-of-the-art methods.

🔹 Publication Date: Published on Feb 21

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.18735
• PDF: https://arxiv.org/pdf/2602.18735
• Github: https://github.com/DavidYan2001/LaS-Comp

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

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

#3DCompletion #ZeroShotLearning #FoundationModels #ComputerVision #AI

✨One-step Language Modeling via Continuous Denoising

📝 Summary:
This paper introduces flow-based language models that use continuous denoising over one-hot token encodings. They surpass discrete diffusion models in quality and speed, particularly for few-step generation, challenging discrete diffusion's necessity for discrete data.

🔹 Publication Date: Published on Feb 18

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.16813
• PDF: https://arxiv.org/pdf/2602.16813
• Project Page: https://one-step-lm.github.io/
• Github: https://github.com/david3684/flm

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

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

#LanguageModels #GenerativeAI #DeepLearning #NLP #AI

✨TextPecker: Rewarding Structural Anomaly Quantification for Enhancing Visual Text Rendering

📝 Summary:
TextPecker proposes a reinforcement learning strategy to improve visual text rendering by perceiving and mitigating structural anomalies in text-to-image generation. It uses a new annotated dataset and synthesis engine to significantly enhance structural fidelity and semantic alignment, setting a...

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.20903
• PDF: https://arxiv.org/pdf/2602.20903
• Project Page: https://github.com/CIawevy/TextPecker
• Github: https://github.com/CIawevy/TextPecker

🔹 Models citing this paper:
• https://huggingface.co/CIawevy/TextPecker-8B-InternVL3
• https://huggingface.co/CIawevy/TextPecker-8B-Qwen3VL
• https://huggingface.co/CIawevy/QwenImage-TextPecker-SQPA

✨ Datasets citing this paper:
• https://huggingface.co/datasets/CIawevy/TextPecker-1.5M

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Communication-Inspired Tokenization for Structured Image Representations

📝 Summary:
COMiT introduces a framework for learning structured, object-centric visual tokens through iterative encoding and flow-matching decoding. This single-transformer approach improves compositional generalization and relational reasoning by creating interpretable token structures.

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.20731
• PDF: https://arxiv.org/pdf/2602.20731
• Project Page: https://araachie.github.io/comit/
• Github: https://github.com/araachie/comit

🔹 Models citing this paper:
• https://huggingface.co/cvg-unibe/comit-xl
• https://huggingface.co/cvg-unibe/comit-l
• https://huggingface.co/cvg-unibe/comit-b

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

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

#ComputerVision #Transformers #ImageRecognition #RepresentationLearning #AIResearch

✨Adaptive Text Anonymization: Learning Privacy-Utility Trade-offs via Prompt Optimization

📝 Summary:
This paper introduces adaptive text anonymization, a framework that uses prompt optimization to automatically adjust anonymization strategies for language models. It adapts to varying privacy-utility requirements across diverse domains, achieving a better trade-off than baselines. It is efficient...

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.20743
• PDF: https://arxiv.org/pdf/2602.20743

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

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

#TextAnonymization #Privacy #PromptOptimization #LLM #NLP

✨Query-focused and Memory-aware Reranker for Long Context Processing

📝 Summary:
This reranking framework uses attention scores from selected LLM heads to estimate passage-query relevance. It's lightweight, achieves strong performance, and outperforms state-of-the-art rerankers across various domains, including long narrative datasets and the LoCoMo benchmark.

🔹 Publication Date: Published on Feb 12

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.12192
• PDF: https://arxiv.org/pdf/2602.12192
• Project Page: https://qdcassie-li.github.io/QRRanker/

🔹 Models citing this paper:
• https://huggingface.co/MindscapeRAG/QRRanker

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

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

#Reranking #LLM #NLP #InformationRetrieval #LongContext

✨QuantVLA: Scale-Calibrated Post-Training Quantization for Vision-Language-Action Models

📝 Summary:
QuantVLA is a training-free post-training quantization framework for vision-language-action models. Through scale-calibrated components, it significantly reduces memory and speeds up inference while maintaining performance, enabling efficient deployment for embodied AI.

🔹 Publication Date: Published on Feb 23

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.20309
• PDF: https://arxiv.org/pdf/2602.20309
• Project Page: https://quantvla.github.io/

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

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

#Quantization #VLAModels #EmbodiedAI #AIResearch #DeepLearning

✨Multi-Vector Index Compression in Any Modality

📝 Summary:
This paper introduces attention-guided clustering AGC for compressing multi-vector document representations across various modalities. AGC consistently outperforms other compression methods in text, visual-document, and video retrieval, often matching or improving upon uncompressed indexes.

🔹 Publication Date: Published on Feb 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.21202
• PDF: https://arxiv.org/pdf/2602.21202

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

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

#IndexCompression #MultiModal #InformationRetrieval #MachineLearning #VectorDatabases

✨PETS: A Principled Framework Towards Optimal Trajectory Allocation for Efficient Test-Time Self-Consistency

📝 Summary:
PETS is a principled framework for efficient test-time self-consistency that optimizes trajectory allocation. It defines a new self-consistency rate, reducing sampling requirements while maintaining accuracy. PETS significantly cuts sampling budgets by up to 75 percent offline and 55 percent onli...

🔹 Publication Date: Published on Feb 18

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2602.16745
• PDF: https://arxiv.org/pdf/2602.16745
• Github: https://github.com/ZDCSlab/PETS

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

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

#SelfConsistency #MachineLearning #Optimization #AI #Efficiency