ML Research Hub – Telegram
ML Research Hub
32.7K subscribers
4.01K photos
229 videos
23 files
4.32K links
Advancing research in Machine Learning – practical insights, tools, and techniques for researchers.

Admin: @HusseinSheikho || @Hussein_Sheikho
Download Telegram
Black-Box On-Policy Distillation of Large Language Models

📝 Summary:
Generative Adversarial Distillation GAD is a new black-box on-policy method for distilling LLMs. GAD trains a student generator and a discriminator for adaptive feedback, surpassing traditional distillation. It enables student LLMs to perform comparably to proprietary teachers.

🔹 Publication Date: Published on Nov 13

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.10643
• PDF: https://arxiv.org/pdf/2511.10643

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#LLMs #AIDistillation #MachineLearning #GenerativeAI #DeepLearning
AlphaResearch: Accelerating New Algorithm Discovery with Language Models

📝 Summary:
AlphaResearch is an autonomous agent that discovers new algorithms using a dual research environment. It achieved a 2/8 win rate against human researchers and found a best-of-known solution for the packing circles problem, showing LLMs potential for algorithm discovery.

🔹 Publication Date: Published on Nov 11

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.08522
• PDF: https://arxiv.org/pdf/2511.08522
• Github: https://github.com/answers111/alpha-research

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#AlgorithmDiscovery #LLMs #AIResearch #AutonomousAgents #MachineLearning
1
Music Flamingo: Scaling Music Understanding in Audio Language Models

📝 Summary:
Music Flamingo, a large audio-language model, advances music understanding through fine-tuning on a rich dataset and post-training with novel methods, achieving state-of-the-art results across various...

🔹 Publication Date: Published on Nov 13

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.10289
• PDF: https://arxiv.org/pdf/2511.10289

🔹 Models citing this paper:
https://huggingface.co/nvidia/music-flamingo-hf

Spaces citing this paper:
https://huggingface.co/spaces/nvidia/music-flamingo

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research
Superpositional Gradient Descent: Harnessing Quantum Principles for Model Training

📝 Summary:
Superpositional Gradient Descent SGD is a new quantum-inspired optimizer. It uses quantum superposition to enhance gradient updates, leading to faster convergence and lower final loss in LLM training than AdamW.

🔹 Publication Date: Published on Nov 1

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.01918
• PDF: https://arxiv.org/pdf/2511.01918
• Github: https://github.com/The-Aqua-Labs/Superpositional-Gradient-Descent

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#MachineLearning #AI #LLM #QuantumInspired #Optimization
1
One Small Step in Latent, One Giant Leap for Pixels: Fast Latent Upscale Adapter for Your Diffusion Models

📝 Summary:
LUA performs efficient super-resolution directly in diffusion models' latent space. This lightweight module enables faster, high-quality image synthesis by upscaling before VAE decoding, cutting time versus pixel-space methods, and generalizing across VAEs.

🔹 Publication Date: Published on Nov 13

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.10629
• PDF: https://arxiv.org/pdf/2511.10629

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#DiffusionModels #SuperResolution #LatentSpace #ImageGeneration #AIResearch
Benchmarking Diversity in Image Generation via Attribute-Conditional Human Evaluation

📝 Summary:
This paper introduces a framework to robustly evaluate diversity in text-to-image models. It uses a novel human evaluation template, curated prompts with variation factors, and systematic analysis of image embeddings to rank models and identify diversity weaknesses.

🔹 Publication Date: Published on Nov 13

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.10547
• PDF: https://arxiv.org/pdf/2511.10547

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#ImageGeneration #TextToImage #AIDiversity #Benchmarking #HumanEvaluation
Rubric-Based Benchmarking and Reinforcement Learning for Advancing LLM Instruction Following

📝 Summary:
AdvancedIF benchmark and RIFL pipeline improve instruction-following capabilities in large language models by using expert-curated rubrics and reinforcement learning techniques. AI-generated summary R...

🔹 Publication Date: Published on Nov 13

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.10507
• PDF: https://arxiv.org/pdf/2511.10507

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research
AffordBot: 3D Fine-grained Embodied Reasoning via Multimodal Large Language Models

📝 Summary:
AffordBot uses MLLMs and chain-of-thought reasoning for fine-grained 3D embodied reasoning. It predicts affordance elements' location, motion type, and axis in 3D scenes per instructions. It achieves state-of-the-art by projecting 3D elements for 2D MLLMs.

🔹 Publication Date: Published on Nov 13

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.10017
• PDF: https://arxiv.org/pdf/2511.10017

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#AffordBot #MLLM #EmbodiedAI #3DReasoning #Robotics
SliderEdit: Continuous Image Editing with Fine-Grained Instruction Control

📝 Summary:
SliderEdit enables continuous, fine-grained control over image editing instructions by using low-rank adaptation matrices, improving edit controllability, visual consistency, and user steerability. AI...

🔹 Publication Date: Published on Nov 12

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.09715
• PDF: https://arxiv.org/pdf/2511.09715

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research
ResearchRubrics: A Benchmark of Prompts and Rubrics For Evaluating Deep Research Agents

📝 Summary:
ResearchRubrics is a benchmark for evaluating deep research agents, using expert rubrics to assess their factual grounding, reasoning, and clarity across diverse, complex tasks. AI-generated summary D...

🔹 Publication Date: Published on Nov 10

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.07685
• PDF: https://arxiv.org/pdf/2511.07685

Datasets citing this paper:
https://huggingface.co/datasets/ScaleAI/researchrubrics

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#AI #DataScience #MachineLearning #HuggingFace #Research
PAN: A World Model for General, Interactable, and Long-Horizon World Simulation

📝 Summary:
PAN is a general interactable world model that predicts future states through high-quality action-conditioned video simulation. It uses a GLP architecture combining LLM-based latent dynamics with a video diffusion decoder for detailed long-term coherent results enabling reasoning and acting.

🔹 Publication Date: Published on Nov 12

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.09057
• PDF: https://arxiv.org/pdf/2511.09057

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#WorldModels #AI #Simulation #GenerativeAI #Robotics
1
Hail to the Thief: Exploring Attacks and Defenses in Decentralised GRPO

📝 Summary:
This study identifies and demonstrates adversarial attacks in decentralized GRPO for LLMs, achieving 100% success rates by injecting malicious tokens. It also proposes effective defense mechanisms that can stop these attacks completely.

🔹 Publication Date: Published on Nov 12

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.09780
• PDF: https://arxiv.org/pdf/2511.09780

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#LLMs #AdversarialAttacks #AISecurity #DecentralizedAI #GRPO
1
Solving a Million-Step LLM Task with Zero Errors

📝 Summary:
MAKER solves million-step LLM tasks with zero errors. It uses extreme task decomposition for microagents and applies error correction at each step with multi-agent voting. This offers a new scalable approach for complex LLM processes.

🔹 Publication Date: Published on Nov 12

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.09030
• PDF: https://arxiv.org/pdf/2511.09030

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#LLM #AI #ErrorCorrection #MultiAgent #TaskDecomposition
CC30k: A Citation Contexts Dataset for Reproducibility-Oriented Sentiment Analysis

📝 Summary:
CC30k is a new dataset of 30,000 machine learning paper citation contexts, labeled with reproducibility-oriented sentiments. It enables large language models to better predict paper reproducibility, filling a crucial gap in computational reproducibility studies.

🔹 Publication Date: Published on Nov 11

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.07790
• PDF: https://arxiv.org/pdf/2511.07790

Datasets citing this paper:
https://huggingface.co/datasets/rochanaro/CC30k

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#MachineLearning #Reproducibility #LLM #SentimentAnalysis #DataScience
1
MM-CRITIC: A Holistic Evaluation of Large Multimodal Models as Multimodal Critique

📝 Summary:
MM-CRITIC is a new benchmark evaluating Large Multimodal Models critique abilities across various dimensions and tasks. It uses expert-informed ground answers and GPT-4o for reliable scoring. This benchmark provides a comprehensive assessment of leading LMMs' critique capabilities.

🔹 Publication Date: Published on Nov 12

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2511.09067
• PDF: https://arxiv.org/pdf/2511.09067

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#LMMs #MultimodalAI #AIEvaluation #Benchmarking #AIResearch
Beyond Outlining: Heterogeneous Recursive Planning for Adaptive Long-form Writing with Language Models

📝 Summary:
This paper proposes an AI agent framework for adaptive long-form writing. It uses recursive task decomposition and dynamically integrates retrieval, reasoning, and composition, overcoming rigid outline-based methods. The framework consistently outperforms state-of-the-art approaches.

🔹 Publication Date: Published on Mar 11

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2503.08275
• PDF: https://arxiv.org/pdf/2503.08275
• Github: https://github.com/principia-ai/WriteHERE

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#AI #LanguageModels #LongformWriting #NLP #GenerativeAI
1
🤖🧠 Steel Browser: The Open-Source Browser API Powering AI Agents and Automation

🗓️ 16 Nov 2025
📚 AI News & Trends

The evolution of artificial intelligence has ushered in a new era of automation where AI agents can perform complex digital tasks with minimal human intervention. However, one of the biggest challenges for developers building these systems is browser automation managing sessions, proxies, cookies and debugging environments. This is where Steel Browser comes into play. Steel ...

#SteelBrowser #OpenSource #BrowserAutomation #AIAgents #WebScraping #DigitalAutomation
👍1🔥1
Transformer Explainer: Interactive Learning of Text-Generative Models

📝 Summary:
Transformer Explainer is an interactive web tool for non-experts to understand the GPT-2 model. It allows real-time experimentation with user input, visualizing how internal components predict text. This broadens access to education about modern generative AI.

🔹 Publication Date: Published on Aug 8, 2024

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2408.04619
• PDF: https://arxiv.org/pdf/2408.04619
• Project Page: https://poloclub.github.io/transformer-explainer/
• Github: https://github.com/helblazer811/ManimML

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

For more data science resources:
https://news.1rj.ru/str/DataScienceT

#AI #GenerativeAI #Transformers #AIeducation #ExplainableAI
❤‍🔥1👍1
🤖🧠 Skyvern: The Future of Browser Automation Powered by AI and Computer Vision

🗓️ 16 Nov 2025
📚 AI News & Trends

In today’s fast-evolving digital landscape, automation plays a crucial role in enhancing productivity, efficiency and innovation. Yet, traditional browser automation tools often struggle with complexity, maintenance and reliability. They rely heavily on DOM parsing, XPaths and rigid noscripts that easily break when websites change their layout. Enter Skyvern, an open-source, AI-driven browser automation platform developed ...

#Skyvern #BrowserAutomation #AIDriven #ComputerVision #OpenSource #WebAutomation
❤‍🔥11👍1
🤖🧠 OpenAI Evals: The Framework Transforming LLM Evaluation and Benchmarking

🗓️ 16 Nov 2025
📚 AI News & Trends

As large language models (LLMs) continue to reshape industries from education and healthcare to marketing and software development – the need for reliable evaluation methods has never been greater. With new models constantly emerging, developers and researchers require a standardized system to test, compare and understand model performance across real-world scenarios. This is where OpenAI ...

#OpenAIEvals #LLMEvaluation #Benchmarking #LargeLanguageModels #AIResearch #ModelEvaluation
1
🤖🧠 Context Engineering 2.0: Redefining Human–Machine Understanding

🗓️ 16 Nov 2025
📚 AI News & Trends

As artificial intelligence advances, machines are becoming increasingly capable of understanding and responding to human language. Yet, one crucial challenge remains how can machines truly understand the context behind human intentions? This question forms the foundation of context engineering, a discipline that focuses on designing, organizing and managing contextual information so that AI systems can ...

#ContextEngineering #AIEducation #HumanMachineUnderstanding #AIContext #NaturalLanguageProcessing #AIModels