ML Research Hub
32.8K subscribers
4.11K photos
241 videos
23 files
4.43K links
Advancing research in Machine Learning – practical insights, tools, and techniques for researchers.

Admin: @HusseinSheikho || @Hussein_Sheikho
Download Telegram
PhyGDPO: Physics-Aware Groupwise Direct Preference Optimization for Physically Consistent Text-to-Video Generation

📝 Summary:
Recent advances in text-to-video (T2V) generation have achieved good visual quality, yet synthesizing videos that faithfully follow physical laws remains an open challenge. Existing methods mainly bas...

🔹 Publication Date: Published on Dec 31, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2512.24551
• PDF: https://arxiv.org/pdf/2512.24551
• Project Page: https://caiyuanhao1998.github.io/project/PhyGDPO/
• Github: https://github.com/caiyuanhao1998/Open-PhyGDPO

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research
Scaling Open-Ended Reasoning to Predict the Future

📝 Summary:
This work trains language models for open-ended future prediction using a new dataset synthesized from news. Their OpenForecaster 8B model matches larger proprietary models in accuracy, calibration, and consistency. All resources are open-sourced.

🔹 Publication Date: Published on Dec 31, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2512.25070
• PDF: https://arxiv.org/pdf/2512.25070
• Project Page: https://www.openforecaster.github.io
• Github: https://github.com/OpenForecaster/scaling-forecasting-training

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

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

#LLMs #FuturePrediction #AI #OpenSourceAI #MachineLearning
Fantastic Reasoning Behaviors and Where to Find Them: Unsupervised Discovery of the Reasoning Process

📝 Summary:
This paper introduces RISE, an unsupervised framework using sparse auto-encoders to discover and control LLM reasoning behaviors. It identifies interpretable reasoning vectors like reflection and backtracking, enabling targeted interventions and discovery of novel behaviors without retraining.

🔹 Publication Date: Published on Dec 30, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2512.23988
• PDF: https://arxiv.org/pdf/2512.23988

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

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

#LLM #AI #MachineLearning #AIReasoning #Interpretability
Let It Flow: Agentic Crafting on Rock and Roll, Building the ROME Model within an Open Agentic Learning Ecosystem

📝 Summary:
The Agentic Learning Ecosystem ALE is a new infrastructure to streamline LLM agent development for real-world tasks. ALE comprises ROLL for optimization, ROCK for sandboxing, and iFlow CLI for context. Their agent ROME, built with ALE, shows strong benchmark performance.

🔹 Publication Date: Published on Dec 31, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2512.24873
• PDF: https://arxiv.org/pdf/2512.24873

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

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

#AIAgents #LLMDevelopment #AgenticLearning #AIArchitecture #MachineLearning
Figure It Out: Improving the Frontier of Reasoning with Active Visual Thinking

📝 Summary:
Complex reasoning problems often involve implicit spatial, geometric, and structural relationships that are not explicitly encoded in text. While recent reasoning models have achieved strong performan...

🔹 Publication Date: Published on Dec 30, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2512.24297
• PDF: https://arxiv.org/pdf/2512.24297
• Github: https://github.com/chenmeiqii/FIGR

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research
Pretraining Frame Preservation in Autoregressive Video Memory Compression

📝 Summary:
We present PFP, a neural network structure to compress long videos into short contexts, with an explicit pretraining objective to preserve the high-frequency details of single frames at arbitrary temp...

🔹 Publication Date: Published on Dec 29, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2512.23851
• PDF: https://arxiv.org/pdf/2512.23851
• Github: https://github.com/lllyasviel/PFP

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research
Factorized Learning for Temporally Grounded Video-Language Models

📝 Summary:
Video-language models struggle with temporal grounding from coupled tasks. Our D^2VLM framework decouples grounding and textual response using evidence tokens. Factorized preference optimization explicitly optimizes temporal grounding for both tasks.

🔹 Publication Date: Published on Dec 30, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2512.24097
• PDF: https://arxiv.org/pdf/2512.24097
• Project Page: https://github.com/nusnlp/d2vlm
• Github: https://github.com/nusnlp/d2vlm

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research
JavisGPT: A Unified Multi-modal LLM for Sounding-Video Comprehension and Generation

📝 Summary:
This paper presents JavisGPT, the first unified multimodal large language model (MLLM) for Joint Audio-Video (JAV) comprehension and generation. JavisGPT adopts a concise encoder-LLM-decoder architect...

🔹 Publication Date: Published on Dec 28, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2503.23377
• PDF: https://arxiv.org/pdf/2512.22905
• Project Page: https://javisverse.github.io/JavisGPT-page/
• Github: https://github.com/JavisVerse/JavisGPT

🔹 Models citing this paper:
https://huggingface.co/JavisVerse/JavisGPT-v0.1-7B-Instruct

Datasets citing this paper:
https://huggingface.co/datasets/JavisVerse/MM-PreTrain
https://huggingface.co/datasets/JavisVerse/JavisUnd-Eval
https://huggingface.co/datasets/JavisVerse/AV-FineTune

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research
Forging Spatial Intelligence: A Roadmap of Multi-Modal Data Pre-Training for Autonomous Systems

📝 Summary:
The rapid advancement of autonomous systems, including self-driving vehicles and drones, has intensified the need to forge true Spatial Intelligence from multi-modal onboard sensor data. While foundat...

🔹 Publication Date: Published on Dec 30, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2512.24385
• PDF: https://arxiv.org/pdf/2512.24385
• Github: https://github.com/worldbench/awesome-spatial-intelligence

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research
Valori: A Deterministic Memory Substrate for AI Systems

📝 Summary:
Valori introduces a deterministic AI memory substrate using fixed-point arithmetic, ensuring bit-identical results across platforms. This eliminates non-determinism from floating-point operations in vector embeddings and search, making AI systems trustworthy and verifiable.

🔹 Publication Date: Published on Dec 25, 2025

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2512.22280
• PDF: https://arxiv.org/pdf/2512.22280
• Project Page: https://valori.systems/
• Github: https://github.com/varshith-Git/Valori-Kernel

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

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

#AI #DataScience #MachineLearning #HuggingFace #Research