✨TTCS: Test-Time Curriculum Synthesis for Self-Evolving

📝 Summary:
TTCS is a co-evolving test-time training framework for LLMs. It improves reasoning by using a question synthesizer to create challenging variants and a reasoning solver updated via self-consistency, with policies mutually guiding each other. This strengthens reasoning on various tasks.

🔹 Publication Date: Published on Jan 30

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.22628
• PDF: https://arxiv.org/pdf/2601.22628
• Github: https://github.com/XMUDeepLIT/TTCS

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

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

#LLM #AI #MachineLearning #CurriculumLearning #SelfEvolving

✨Real-Time Aligned Reward Model beyond Semantics

📝 Summary:
RLHF faces reward overoptimization from reward model misalignment. R2M introduces a new framework that uses real-time policy feedback to dynamically adapt the reward model. This improves alignment by responding to continuous policy distribution shifts beyond just semantics.

🔹 Publication Date: Published on Jan 30

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.22664
• PDF: https://arxiv.org/pdf/2601.22664

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

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

#ReinforcementLearning #AI #MachineLearning #RewardModels #AIAlignment

✨SSL: Sweet Spot Learning for Differentiated Guidance in Agentic Optimization

📝 Summary:
Sweet Spot Learning SSL is a novel RL framework employing tiered rewards for differentiated guidance. This directs agents to optimal solution regions, significantly boosting sample efficiency and cross-task transferability.

🔹 Publication Date: Published on Jan 30

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.22491
• PDF: https://arxiv.org/pdf/2601.22491

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

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

#ReinforcementLearning #AgenticAI #MachineLearning #Optimization #SampleEfficiency

✨Routing the Lottery: Adaptive Subnetworks for Heterogeneous Data

📝 Summary:
Routing the Lottery framework discovers multiple specialized subnetworks tailored to different data conditions, outperforming traditional pruning methods while using fewer parameters and identifying s...

🔹 Publication Date: Published on Jan 29

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.22141
• PDF: https://arxiv.org/pdf/2601.22141

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Latent Chain-of-Thought as Planning: Decoupling Reasoning from Verbalization

📝 Summary:
PLaT introduces a latent reasoning framework that decouples reasoning from verbalization, enabling dynamic termination and improved scalability over traditional approaches. AI-generated summary Chain-...

🔹 Publication Date: Published on Jan 29

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.21358
• PDF: https://arxiv.org/pdf/2601.21358

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨THINKSAFE: Self-Generated Safety Alignment for Reasoning Models

📝 Summary:
ThinkSafe is a self-aligned framework that enhances safety in large reasoning models. It uses lightweight refusal steering and fine-tuning on self-generated responses to preserve reasoning performance and reduce computational costs. ThinkSafe significantly improves safety without degrading native...

🔹 Publication Date: Published on Jan 30

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.23143
• PDF: https://arxiv.org/pdf/2601.23143

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

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

#AISafety #LLMs #AIAlignment #MachineLearning #DeepLearning

✨MemOCR: Layout-Aware Visual Memory for Efficient Long-Horizon Reasoning

📝 Summary:
MemOCR is a multimodal memory agent for long-horizon reasoning that compresses interaction histories into visual layouts. It adaptively allocates memory space, visually prioritizing crucial evidence while compressing details, outperforming text-based baselines under tight budgets.

🔹 Publication Date: Published on Jan 29

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.21468
• PDF: https://arxiv.org/pdf/2601.21468

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

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

#AI #MultimodalAI #LongHorizonReasoning #MemoryNetworks #ComputerVision

✨Pushing the Boundaries of Natural Reasoning: Interleaved Bonus from Formal-Logic Verification

📝 Summary:
This paper presents a framework that interleaves formal logic verification with natural language generation to improve LLM reasoning. It actively detects and corrects errors during the reasoning process. This method significantly outperforms state-of-the-art models on various reasoning benchmarks.

🔹 Publication Date: Published on Jan 30

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.22642
• PDF: https://arxiv.org/pdf/2601.22642

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨PaperBanana: Automating Academic Illustration for AI Scientists

📝 Summary:
_paperbanana is an agentic framework that automates the creation of publication-ready academic illustrations using advanced vision-language models and image generation techniques. AI-generated summary...

🔹 Publication Date: Published on Jan 30

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.23265
• PDF: https://arxiv.org/pdf/2601.23265
• Project Page: https://dwzhu-pku.github.io/PaperBanana/

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨ReGuLaR: Variational Latent Reasoning Guided by Rendered Chain-of-Thought

📝 Summary:
ReGuLaR introduces a variational auto-encoding framework that compresses reasoning processes into latent space while maintaining performance through image-rendered explicit reasoning chains for guidan...

🔹 Publication Date: Published on Jan 30

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.23184
• PDF: https://arxiv.org/pdf/2601.23184
• Github: https://github.com/FanmengWang/ReGuLaR

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨LMK > CLS: Landmark Pooling for Dense Embeddings

📝 Summary:
Landmark pooling improves long-context representation learning by partitioning sequences into chunks and using landmark tokens to preserve both global and local information more effectively than tradi...

🔹 Publication Date: Published on Jan 29

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.21525
• PDF: https://arxiv.org/pdf/2601.21525

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨DINO-SAE: DINO Spherical Autoencoder for High-Fidelity Image Reconstruction and Generation

📝 Summary:
A novel vision autoencoder framework combines semantic representation with pixel-level reconstruction using spherical latent space and Riemannian flow matching for improved fidelity and efficiency. AI...

🔹 Publication Date: Published on Jan 30

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.22904
• PDF: https://arxiv.org/pdf/2601.22904

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨NativeTok: Native Visual Tokenization for Improved Image Generation

📝 Summary:
NativeTok introduces a novel visual tokenization approach that enforces causal dependencies during image encoding, using a Meta Image Transformer and Mixture of Causal Expert Transformer for efficient...

🔹 Publication Date: Published on Jan 30

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.22837
• PDF: https://arxiv.org/pdf/2601.22837
• Github: https://github.com/wangbei1/Nativetok

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨DIFFA-2: A Practical Diffusion Large Language Model for General Audio Understanding

📝 Summary:
DIFFA-2, a diffusion-based large audio language model, achieves competitive audio understanding performance with improved efficiency over autoregressive counterparts through enhanced encoding, dual ad...

🔹 Publication Date: Published on Jan 30

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.23161
• PDF: https://arxiv.org/pdf/2601.23161
• Github: https://github.com/NKU-HLT/DIFFA

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Deep Search with Hierarchical Meta-Cognitive Monitoring Inspired by Cognitive Neuroscience

📝 Summary:
Deep search agents with hierarchical metacognitive monitoring enhance reasoning and retrieval performance through fast consistency checks and experience-driven corrective interventions. AI-generated s...

🔹 Publication Date: Published on Jan 30

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.23188
• PDF: https://arxiv.org/pdf/2601.23188

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Robust Tool Use via Fission-GRPO: Learning to Recover from Execution Errors

📝 Summary:
A framework called Fission-GRPO is introduced to improve multi-turn tool execution in large language models by converting execution errors into corrective supervision during reinforcement learning tra...

🔹 Publication Date: Published on Jan 22

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.15625
• PDF: https://arxiv.org/pdf/2601.15625

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨FourierSampler: Unlocking Non-Autoregressive Potential in Diffusion Language Models via Frequency-Guided Generation

📝 Summary:
Diffusion language models face positional bias. FourierSampler uses frequency analysis to guide generation by separating global structure from local details. This sliding window approach significantly outperforms previous methods and autoregressive models.

🔹 Publication Date: Published on Jan 30

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.23182
• PDF: https://arxiv.org/pdf/2601.23182
• Github: https://github.com/ShirleYoung/FourierSampler

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Statistical Estimation of Adversarial Risk in Large Language Models under Best-of-N Sampling

📝 Summary:
A scaling-aware risk estimation method called SABER is introduced for predicting large-scale adversarial vulnerability in language models through Best-of-N sampling, enabling accurate assessment with ...

🔹 Publication Date: Published on Jan 30

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.22636
• PDF: https://arxiv.org/pdf/2601.22636

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨PaddleOCR-VL-1.5: Towards a Multi-Task 0.9B VLM for Robust In-the-Wild Document Parsing

📝 Summary:
A compact vision-language model achieves state-of-the-art accuracy on document understanding tasks while maintaining efficiency through specialized benchmarking and extended functionality. AI-generate...

🔹 Publication Date: Published on Jan 29

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.21957
• PDF: https://arxiv.org/pdf/2601.21957

🔹 Models citing this paper:
• https://huggingface.co/PaddlePaddle/PaddleOCR-VL-1.5
• https://huggingface.co/PaddlePaddle/PP-DocLayoutV3

✨ Spaces citing this paper:
• https://huggingface.co/spaces/PaddlePaddle/PaddleOCR-VL-1.5_Online_Demo
• https://huggingface.co/spaces/AAAASSSASDASD3000/PaddleOCR-VL-1.5_Online_Demo

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Revisiting Diffusion Model Predictions Through Dimensionality

📝 Summary:
Diffusion models using direct data prediction outperform traditional noise or velocity prediction in high-dimensional settings, with a proposed framework automatically learning optimal prediction para...

🔹 Publication Date: Published on Jan 29

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.21419
• PDF: https://arxiv.org/pdf/2601.21419

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research

✨Machine Learning for Energy-Performance-aware Scheduling

📝 Summary:
We propose a Bayesian Optimization framework using Gaussian Processes to automate scheduling configuration on multi-core systems. It approximates the energy-time Pareto Frontier and reveals dominant hardware parameters through sensitivity analysis.

🔹 Publication Date: Published on Jan 30

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.23134
• PDF: https://arxiv.org/pdf/2601.23134

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

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

#MachineLearning #Optimization #EnergyEfficiency #ComputerArchitecture #DataScience