ML Research Hub
32.8K subscribers
4.32K photos
263 videos
23 files
4.66K links
Advancing research in Machine Learning – practical insights, tools, and techniques for researchers.

Admin: @HusseinSheikho || @Hussein_Sheikho
Download Telegram
πŸ”Ή Title: Glyph: Scaling Context Windows via Visual-Text Compression

πŸ”Ή Publication Date: Published on Oct 20

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.17800
β€’ PDF: https://arxiv.org/pdf/2510.17800

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
πŸ”Ή Title: Uniworld-V2: Reinforce Image Editing with Diffusion Negative-aware Finetuning and MLLM Implicit Feedback

πŸ”Ή Publication Date: Published on Oct 19

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.16888
β€’ PDF: https://arxiv.org/pdf/2510.16888
β€’ Github: https://github.com/PKU-YuanGroup/UniWorld-V2

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
πŸ”Ή Title: Annotation-Efficient Universal Honesty Alignment

πŸ”Ή Publication Date: Published on Oct 20

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.17509
β€’ PDF: https://arxiv.org/pdf/2510.17509

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
πŸ”Ή Title: RL makes MLLMs see better than SFT

πŸ”Ή Publication Date: Published on Oct 18

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.16333
β€’ PDF: https://arxiv.org/pdf/2510.16333
β€’ Project Page: https://june-page.github.io/pivot/

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
πŸ”Ή Title: UltraCUA: A Foundation Model for Computer Use Agents with Hybrid Action

πŸ”Ή Publication Date: Published on Oct 20

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.17790
β€’ PDF: https://arxiv.org/pdf/2510.17790

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
πŸ”Ή Title: Embody 3D: A Large-scale Multimodal Motion and Behavior Dataset

πŸ”Ή Publication Date: Published on Oct 17

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.16258
β€’ PDF: https://arxiv.org/pdf/2510.16258
β€’ Project Page: https://www.meta.com/emerging-tech/codec-avatars/embody-3d/

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
❀1
πŸ”Ή Title: When to Ensemble: Identifying Token-Level Points for Stable and Fast LLM Ensembling

πŸ”Ή Publication Date: Published on Oct 17

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.15346
β€’ PDF: https://arxiv.org/pdf/2510.15346

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
πŸ”Ή Title: Constantly Improving Image Models Need Constantly Improving Benchmarks

πŸ”Ή Publication Date: Published on Oct 16

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.15021
β€’ PDF: https://arxiv.org/pdf/2510.15021
β€’ Project Page: https://echo-bench.github.io/
β€’ Github: https://github.com/para-lost/ECHO

πŸ”Ή Datasets citing this paper:
β€’ https://huggingface.co/datasets/echo-bench/echo2025

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
πŸ”Ή Title: Foundational Automatic Evaluators: Scaling Multi-Task Generative Evaluator Training for Reasoning-Centric Domains

πŸ”Ή Publication Date: Published on Oct 20

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.17793
β€’ PDF: https://arxiv.org/pdf/2510.17793

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
❀1
πŸ”Ή Title: Chronos-2: From Univariate to Universal Forecasting

πŸ”Ή Publication Date: Published on Oct 17

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.15821
β€’ PDF: https://arxiv.org/pdf/2510.15821
β€’ Github: https://github.com/amazon-science/chronos-forecasting

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
πŸ”Ή Title: GuideFlow3D: Optimization-Guided Rectified Flow For Appearance Transfer

πŸ”Ή Publication Date: Published on Oct 17

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.16136
β€’ PDF: https://arxiv.org/pdf/2510.16136
β€’ Github: https://github.com/GradientSpaces/GuideFlow3D

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
πŸ”Ή Title: MultiVerse: A Multi-Turn Conversation Benchmark for Evaluating Large Vision and Language Models

πŸ”Ή Publication Date: Published on Oct 18

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.16641
β€’ PDF: https://arxiv.org/pdf/2510.16641
β€’ Github: https://github.com/passing2961/MultiVerse

πŸ”Ή Datasets citing this paper:
β€’ https://huggingface.co/datasets/passing2961/MultiVerse

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
πŸ”Ή Title: On Non-interactive Evaluation of Animal Communication Translators

πŸ”Ή Publication Date: Published on Oct 17

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.15768
β€’ PDF: https://arxiv.org/pdf/2510.15768

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
πŸ”Ή Title: Agentic Reinforcement Learning for Search is Unsafe

πŸ”Ή Publication Date: Published on Oct 20

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.17431
β€’ PDF: https://arxiv.org/pdf/2510.17431

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
πŸ”Ή Title: QueST: Incentivizing LLMs to Generate Difficult Problems

πŸ”Ή Publication Date: Published on Oct 20

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.17715
β€’ PDF: https://arxiv.org/pdf/2510.17715

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
❀1
πŸ€–πŸ§  Wan 2.1: Alibaba’s Open-Source Revolution in Video Generation

πŸ—“οΈ 21 Oct 2025
πŸ“š AI News & Trends

The landscape of artificial intelligence has been evolving rapidly, especially in the domain of video generation. Since OpenAI unveiled Sora in 2024, the world has witnessed an explosive surge in research and innovation within generative AI. However, most of these cutting-edge tools remained closed-source limiting transparency and accessibility. Recognizing this gap, Alibaba Group introduced Wan, ...

#Alibaba #Wan2.1 #VideoGeneration #GenerativeAI #OpenSource #ArtificialIntelligence
❀1
πŸ€–πŸ§  DeepSeek-OCR: Redefining Document Understanding Through Optical Context Compression

πŸ—“οΈ 21 Oct 2025
πŸ“š AI News & Trends

In the age of large language models (LLMs) and vision-language models (VLMs), handling long and complex textual data efficiently remains a massive challenge. Traditional models struggle with processing extended contexts because the computational cost increases quadratically with sequence length. To overcome this, researchers from DeepSeek-AI have introduced a groundbreaking approach – DeepSeek-OCR, a model that ...
πŸ”Ή Title: Test-Time Scaling of Reasoning Models for Machine Translation

πŸ”Ή Publication Date: Published on Oct 7

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.06471
β€’ PDF: https://arxiv.org/pdf/2510.06471

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
πŸ”Ή Title: Beacon: Single-Turn Diagnosis and Mitigation of Latent Sycophancy in Large Language Models

πŸ”Ή Publication Date: Published on Oct 19

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.16727
β€’ PDF: https://arxiv.org/pdf/2510.16727

πŸ”Ή Datasets citing this paper:
β€’ https://huggingface.co/datasets/sanskxr02/Beacon

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
❀1
πŸ”Ή Title: Automated Composition of Agents: A Knapsack Approach for Agentic Component Selection

πŸ”Ή Publication Date: Published on Oct 18

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.16499
β€’ PDF: https://arxiv.org/pdf/2510.16499

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
πŸ”Ή Title: What Limits Agentic Systems Efficiency?

πŸ”Ή Publication Date: Published on Oct 18

πŸ”Ή Paper Links:
β€’ arXiv Page: https://arxiv.org/abs/2510.16276
β€’ PDF: https://arxiv.org/pdf/2510.16276

πŸ”Ή Datasets citing this paper:
No datasets found

πŸ”Ή Spaces citing this paper:
No spaces found
==================================

For more data science resources:
βœ“ https://t.me/DataScienceT
❀2