✨Shaping capabilities with token-level data filtering
📝 Summary:
Token filtering during pretraining effectively reduces unwanted language model capabilities while maintaining alignment, becoming more effective at larger scales and tolerating noisy labels with suffi...
🔹 Publication Date: Published on Jan 29
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.21571
• PDF: https://arxiv.org/pdf/2601.21571
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #DataScience #MachineLearning #HuggingFace #Research
📝 Summary:
Token filtering during pretraining effectively reduces unwanted language model capabilities while maintaining alignment, becoming more effective at larger scales and tolerating noisy labels with suffi...
🔹 Publication Date: Published on Jan 29
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.21571
• PDF: https://arxiv.org/pdf/2601.21571
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #DataScience #MachineLearning #HuggingFace #Research
Media is too big
VIEW IN TELEGRAM
✨LoL: Longer than Longer, Scaling Video Generation to Hour
📝 Summary:
Researchers addressed sink-collapse in autoregressive video generation, a failure mode where content reverts to a sink frame due to a RoPE and multi-head attention conflict. Their training-free multi-head RoPE jitter enables real-time, streaming video generation up to 12 hours without quality decay.
🔹 Publication Date: Published on Jan 23
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.16914
• PDF: https://arxiv.org/pdf/2601.16914
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #DataScience #MachineLearning #HuggingFace #Research
📝 Summary:
Researchers addressed sink-collapse in autoregressive video generation, a failure mode where content reverts to a sink frame due to a RoPE and multi-head attention conflict. Their training-free multi-head RoPE jitter enables real-time, streaming video generation up to 12 hours without quality decay.
🔹 Publication Date: Published on Jan 23
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.16914
• PDF: https://arxiv.org/pdf/2601.16914
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #DataScience #MachineLearning #HuggingFace #Research
✨Latent Adversarial Regularization for Offline Preference Optimization
📝 Summary:
GANPO uses latent-space regularization via adversarial divergence minimization to improve language model preference optimization. It offers more robust structural feedback than token-level methods, performing better under distributional shift and noise with minor overhead.
🔹 Publication Date: Published on Jan 29
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.22083
• PDF: https://arxiv.org/pdf/2601.22083
• Github: https://github.com/enyijiang/GANPO
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #DataScience #MachineLearning #HuggingFace #Research
📝 Summary:
GANPO uses latent-space regularization via adversarial divergence minimization to improve language model preference optimization. It offers more robust structural feedback than token-level methods, performing better under distributional shift and noise with minor overhead.
🔹 Publication Date: Published on Jan 29
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.22083
• PDF: https://arxiv.org/pdf/2601.22083
• Github: https://github.com/enyijiang/GANPO
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #DataScience #MachineLearning #HuggingFace #Research
✨Flow-based Extremal Mathematical Structure Discovery
📝 Summary:
FlowBoost is a new generative framework for discovering extremal geometric structures. It uses flow-matching, policy optimization, and local search. This closed-loop approach efficiently finds new best results for geometric optimization problems, outperforming prior methods like AlphaEvolve.
🔹 Publication Date: Published on Jan 25
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.18005
• PDF: https://arxiv.org/pdf/2601.18005
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #DataScience #MachineLearning #HuggingFace #Research
📝 Summary:
FlowBoost is a new generative framework for discovering extremal geometric structures. It uses flow-matching, policy optimization, and local search. This closed-loop approach efficiently finds new best results for geometric optimization problems, outperforming prior methods like AlphaEvolve.
🔹 Publication Date: Published on Jan 25
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.18005
• PDF: https://arxiv.org/pdf/2601.18005
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #DataScience #MachineLearning #HuggingFace #Research
❤1
✨Reinforcement Learning from Meta-Evaluation: Aligning Language Models Without Ground-Truth Labels
📝 Summary:
Reinforcement Learning from Meta-Evaluation RLME trains language models without ground truth labels. It uses an evaluators judgments on natural language meta questions as reward. RLME achieves comparable accuracy and efficiency to label based methods, broadening RL application for LLM training.
🔹 Publication Date: Published on Jan 29
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.21268
• PDF: https://arxiv.org/pdf/2601.21268
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #DataScience #MachineLearning #HuggingFace #Research
📝 Summary:
Reinforcement Learning from Meta-Evaluation RLME trains language models without ground truth labels. It uses an evaluators judgments on natural language meta questions as reward. RLME achieves comparable accuracy and efficiency to label based methods, broadening RL application for LLM training.
🔹 Publication Date: Published on Jan 29
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.21268
• PDF: https://arxiv.org/pdf/2601.21268
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #DataScience #MachineLearning #HuggingFace #Research
❤1
✨EEG Foundation Models: Progresses, Benchmarking, and Open Problems
📝 Summary:
This paper benchmarks EEG foundation models, finding specialist models remain competitive. Linear probing is often insufficient, and larger models dont always improve generalization under current data conditions.
🔹 Publication Date: Published on Jan 25
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.17883
• PDF: https://arxiv.org/pdf/2601.17883
• Github: https://github.com/Dingkun0817/EEG-FM-Benchmark
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #DataScience #MachineLearning #HuggingFace #Research
📝 Summary:
This paper benchmarks EEG foundation models, finding specialist models remain competitive. Linear probing is often insufficient, and larger models dont always improve generalization under current data conditions.
🔹 Publication Date: Published on Jan 25
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.17883
• PDF: https://arxiv.org/pdf/2601.17883
• Github: https://github.com/Dingkun0817/EEG-FM-Benchmark
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #DataScience #MachineLearning #HuggingFace #Research
❤1
✨Stable Video Infinity: Infinite-Length Video Generation with Error Recycling
📝 Summary:
Stable Video Infinity SVI generates infinite-length videos with high temporal consistency and controllable storylines. It uses Error-Recycling Fine-Tuning, teaching the Diffusion Transformer to identify and correct its own errors by recycling self-generated errors.
🔹 Publication Date: Published on Oct 10, 2025
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2510.09212
• PDF: https://arxiv.org/pdf/2510.09212
• Project Page: https://stable-video-infinity.github.io/homepage/
• Github: https://github.com/vita-epfl/Stable-Video-Infinity
🔹 Models citing this paper:
• https://huggingface.co/vita-video-gen/svi-model
✨ Datasets citing this paper:
• https://huggingface.co/datasets/vita-video-gen/svi-benchmark
• https://huggingface.co/datasets/mzwydf/svi-benchmark
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#VideoGeneration #DiffusionModels #GenerativeAI #ComputerVision #AIResearch
📝 Summary:
Stable Video Infinity SVI generates infinite-length videos with high temporal consistency and controllable storylines. It uses Error-Recycling Fine-Tuning, teaching the Diffusion Transformer to identify and correct its own errors by recycling self-generated errors.
🔹 Publication Date: Published on Oct 10, 2025
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2510.09212
• PDF: https://arxiv.org/pdf/2510.09212
• Project Page: https://stable-video-infinity.github.io/homepage/
• Github: https://github.com/vita-epfl/Stable-Video-Infinity
🔹 Models citing this paper:
• https://huggingface.co/vita-video-gen/svi-model
✨ Datasets citing this paper:
• https://huggingface.co/datasets/vita-video-gen/svi-benchmark
• https://huggingface.co/datasets/mzwydf/svi-benchmark
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#VideoGeneration #DiffusionModels #GenerativeAI #ComputerVision #AIResearch
arXiv.org
Stable Video Infinity: Infinite-Length Video Generation with Error...
We propose Stable Video Infinity (SVI) that is able to generate infinite-length videos with high temporal consistency, plausible scene transitions, and controllable streaming storylines. While...
✨Live Avatar: Streaming Real-time Audio-Driven Avatar Generation with Infinite Length
📝 Summary:
Live Avatar enables real-time, high-fidelity, infinite-length avatar generation using a 14B-parameter diffusion model. It employs Timestep-forcing Pipeline Parallelism and the Rolling Sink Frame Mechanism to overcome limitations, achieving 20 FPS on 5 GPUs. This is the first practical system at t...
🔹 Publication Date: Published on Dec 4, 2025
🔹 Paper Links:
• arXiv Page: https://arxivexplained.com/papers/live-avatar-streaming-real-time-audio-driven-avatar-generation-with-infinite-length
• PDF: https://arxiv.org/pdf/2512.04677
• Project Page: https://liveavatar.github.io/
• Github: https://github.com/Alibaba-Quark/LiveAvatar
🔹 Models citing this paper:
• https://huggingface.co/Quark-Vision/Live-Avatar
✨ Spaces citing this paper:
• https://huggingface.co/spaces/ahm98alex/liveavatar-test
• https://huggingface.co/spaces/sdavignon/liveavatar
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#LiveAvatar #AvatarGeneration #RealtimeAI #DiffusionModels #GenerativeAI
📝 Summary:
Live Avatar enables real-time, high-fidelity, infinite-length avatar generation using a 14B-parameter diffusion model. It employs Timestep-forcing Pipeline Parallelism and the Rolling Sink Frame Mechanism to overcome limitations, achieving 20 FPS on 5 GPUs. This is the first practical system at t...
🔹 Publication Date: Published on Dec 4, 2025
🔹 Paper Links:
• arXiv Page: https://arxivexplained.com/papers/live-avatar-streaming-real-time-audio-driven-avatar-generation-with-infinite-length
• PDF: https://arxiv.org/pdf/2512.04677
• Project Page: https://liveavatar.github.io/
• Github: https://github.com/Alibaba-Quark/LiveAvatar
🔹 Models citing this paper:
• https://huggingface.co/Quark-Vision/Live-Avatar
✨ Spaces citing this paper:
• https://huggingface.co/spaces/ahm98alex/liveavatar-test
• https://huggingface.co/spaces/sdavignon/liveavatar
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#LiveAvatar #AvatarGeneration #RealtimeAI #DiffusionModels #GenerativeAI
Arxivexplained
Live Avatar: Streaming Real-time Audio-Driven Avatar Generation with Infinite Length - Explained Simply
By Yubo Huang, Hailong Guo, Fangtai Wu et al.. # Live Avatar: Real-Time AI Avatars That Never Stop
**The Problem We've All Been Waiting to Solve**...
**The Problem We've All Been Waiting to Solve**...
👍1
🚀 Master Data Science & Programming!
Unlock your potential with this curated list of Telegram channels. Whether you need books, datasets, interview prep, or project ideas, we have the perfect resource for you. Join the community today!
🔰 Machine Learning with Python
Learn Machine Learning with hands-on Python tutorials, real-world code examples, and clear explanations for researchers and developers.
https://t.iss.one/CodeProgrammer
🔖 Machine Learning
Machine learning insights, practical tutorials, and clear explanations for beginners and aspiring data scientists. Follow the channel for models, algorithms, coding guides, and real-world ML applications.
https://t.iss.one/DataScienceM
🧠 Code With Python
This channel delivers clear, practical content for developers, covering Python, Django, Data Structures, Algorithms, and DSA – perfect for learning, coding, and mastering key programming skills.
https://t.iss.one/DataScience4
🎯 PyData Careers | Quiz
Python Data Science jobs, interview tips, and career insights for aspiring professionals.
https://t.iss.one/DataScienceQ
💾 Kaggle Data Hub
Your go-to hub for Kaggle datasets – explore, analyze, and leverage data for Machine Learning and Data Science projects.
https://t.iss.one/datasets1
🧑🎓 Udemy Coupons | Courses
The first channel in Telegram that offers free Udemy coupons
https://t.iss.one/DataScienceC
😀 ML Research Hub
Advancing research in Machine Learning – practical insights, tools, and techniques for researchers.
https://t.iss.one/DataScienceT
💬 Data Science Chat
An active community group for discussing data challenges and networking with peers.
https://t.iss.one/DataScience9
🐍 Python Arab| بايثون عربي
The largest Arabic-speaking group for Python developers to share knowledge and help.
https://t.iss.one/PythonArab
🖊 Data Science Jupyter Notebooks
Explore the world of Data Science through Jupyter Notebooks—insights, tutorials, and tools to boost your data journey. Code, analyze, and visualize smarter with every post.
https://t.iss.one/DataScienceN
📺 Free Online Courses | Videos
Free online courses covering data science, machine learning, analytics, programming, and essential skills for learners.
https://t.iss.one/DataScienceV
📈 Data Analytics
Dive into the world of Data Analytics – uncover insights, explore trends, and master data-driven decision making.
https://t.iss.one/DataAnalyticsX
🎧 Learn Python Hub
Master Python with step-by-step courses – from basics to advanced projects and practical applications.
https://t.iss.one/Python53
⭐️ Research Papers
Professional Academic Writing & Simulation Services
https://t.iss.one/DataScienceY
━━━━━━━━━━━━━━━━━━
Admin: @HusseinSheikho
Unlock your potential with this curated list of Telegram channels. Whether you need books, datasets, interview prep, or project ideas, we have the perfect resource for you. Join the community today!
Learn Machine Learning with hands-on Python tutorials, real-world code examples, and clear explanations for researchers and developers.
https://t.iss.one/CodeProgrammer
Machine learning insights, practical tutorials, and clear explanations for beginners and aspiring data scientists. Follow the channel for models, algorithms, coding guides, and real-world ML applications.
https://t.iss.one/DataScienceM
This channel delivers clear, practical content for developers, covering Python, Django, Data Structures, Algorithms, and DSA – perfect for learning, coding, and mastering key programming skills.
https://t.iss.one/DataScience4
Python Data Science jobs, interview tips, and career insights for aspiring professionals.
https://t.iss.one/DataScienceQ
Your go-to hub for Kaggle datasets – explore, analyze, and leverage data for Machine Learning and Data Science projects.
https://t.iss.one/datasets1
The first channel in Telegram that offers free Udemy coupons
https://t.iss.one/DataScienceC
Advancing research in Machine Learning – practical insights, tools, and techniques for researchers.
https://t.iss.one/DataScienceT
An active community group for discussing data challenges and networking with peers.
https://t.iss.one/DataScience9
The largest Arabic-speaking group for Python developers to share knowledge and help.
https://t.iss.one/PythonArab
Explore the world of Data Science through Jupyter Notebooks—insights, tutorials, and tools to boost your data journey. Code, analyze, and visualize smarter with every post.
https://t.iss.one/DataScienceN
Free online courses covering data science, machine learning, analytics, programming, and essential skills for learners.
https://t.iss.one/DataScienceV
Dive into the world of Data Analytics – uncover insights, explore trends, and master data-driven decision making.
https://t.iss.one/DataAnalyticsX
Master Python with step-by-step courses – from basics to advanced projects and practical applications.
https://t.iss.one/Python53
Professional Academic Writing & Simulation Services
https://t.iss.one/DataScienceY
━━━━━━━━━━━━━━━━━━
Admin: @HusseinSheikho
Please open Telegram to view this post
VIEW IN TELEGRAM
❤2
✨The Well: a Large-Scale Collection of Diverse Physics Simulations for Machine Learning
📝 Summary:
The Well is a new 15TB collection of 16 diverse physics simulation datasets. It provides comprehensive data from various domains for benchmarking machine learning models in physical systems, addressing gaps in current standard datasets. A unified PyTorch interface aids usage.
🔹 Publication Date: Published on Nov 30, 2024
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2412.00568
• PDF: https://arxiv.org/pdf/2412.00568
• Github: https://github.com/PolymathicAI/the_well
✨ Datasets citing this paper:
• https://huggingface.co/datasets/polymathic-ai/rayleigh_benard
• https://huggingface.co/datasets/polymathic-ai/gray_scott_reaction_diffusion
• https://huggingface.co/datasets/polymathic-ai/turbulence_gravity_cooling
✨ Spaces citing this paper:
• https://huggingface.co/spaces/polymathic-ai/TheWell
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#MachineLearning #PhysicsSimulations #AIforScience #Datasets #PyTorch
📝 Summary:
The Well is a new 15TB collection of 16 diverse physics simulation datasets. It provides comprehensive data from various domains for benchmarking machine learning models in physical systems, addressing gaps in current standard datasets. A unified PyTorch interface aids usage.
🔹 Publication Date: Published on Nov 30, 2024
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2412.00568
• PDF: https://arxiv.org/pdf/2412.00568
• Github: https://github.com/PolymathicAI/the_well
✨ Datasets citing this paper:
• https://huggingface.co/datasets/polymathic-ai/rayleigh_benard
• https://huggingface.co/datasets/polymathic-ai/gray_scott_reaction_diffusion
• https://huggingface.co/datasets/polymathic-ai/turbulence_gravity_cooling
✨ Spaces citing this paper:
• https://huggingface.co/spaces/polymathic-ai/TheWell
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#MachineLearning #PhysicsSimulations #AIforScience #Datasets #PyTorch
arXiv.org
The Well: a Large-Scale Collection of Diverse Physics Simulations...
Machine learning based surrogate models offer researchers powerful tools for accelerating simulation-based workflows. However, as standard datasets in this space often cover small classes of...
❤1
✨ASTRA: Automated Synthesis of agentic Trajectories and Reinforcement Arenas
📝 Summary:
ASTRA automates training tool-augmented language models for multi-step decision-making. It uses synthetic data and verifiable reinforcement learning, integrating SFT and online RL. This achieves state-of-the-art performance.
🔹 Publication Date: Published on Jan 29
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.21558
• PDF: https://arxiv.org/pdf/2601.21558
• Github: https://lianjiatech.github.io/astra.blog/
🔹 Models citing this paper:
• https://huggingface.co/Emperorizzis/ASTRA-14B-Thinking-v1
• https://huggingface.co/Emperorizzis/ASTRA-32B-Thinking-v1
✨ Datasets citing this paper:
• https://huggingface.co/datasets/Emperorizzis/ASTRA-SFT-1k
• https://huggingface.co/datasets/Emperorizzis/ASTRA-RL-1k
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #ReinforcementLearning #LanguageModels #MultiStepDecisionMaking #MachineLearning
📝 Summary:
ASTRA automates training tool-augmented language models for multi-step decision-making. It uses synthetic data and verifiable reinforcement learning, integrating SFT and online RL. This achieves state-of-the-art performance.
🔹 Publication Date: Published on Jan 29
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.21558
• PDF: https://arxiv.org/pdf/2601.21558
• Github: https://lianjiatech.github.io/astra.blog/
🔹 Models citing this paper:
• https://huggingface.co/Emperorizzis/ASTRA-14B-Thinking-v1
• https://huggingface.co/Emperorizzis/ASTRA-32B-Thinking-v1
✨ Datasets citing this paper:
• https://huggingface.co/datasets/Emperorizzis/ASTRA-SFT-1k
• https://huggingface.co/datasets/Emperorizzis/ASTRA-RL-1k
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #ReinforcementLearning #LanguageModels #MultiStepDecisionMaking #MachineLearning
✨DreamActor-M2: Universal Character Image Animation via Spatiotemporal In-Context Learning
📝 Summary:
DreamActor-M2 is a universal character animation framework. It uses in-context learning to fuse appearance and motion cues, along with self-bootstrapped data synthesis for RGB-driven animation. This approach overcomes motion injection tradeoffs and pose prior limitations, achieving superior fidel...
🔹 Publication Date: Published on Jan 29
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.21716
• PDF: https://arxiv.org/pdf/2601.21716
• Project Page: https://grisoon.github.io/DreamActor-M2/
• Github: https://grisoon.github.io/DreamActor-M2/
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#CharacterAnimation #AI #ComputerVision #DeepLearning #GenerativeAI
📝 Summary:
DreamActor-M2 is a universal character animation framework. It uses in-context learning to fuse appearance and motion cues, along with self-bootstrapped data synthesis for RGB-driven animation. This approach overcomes motion injection tradeoffs and pose prior limitations, achieving superior fidel...
🔹 Publication Date: Published on Jan 29
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.21716
• PDF: https://arxiv.org/pdf/2601.21716
• Project Page: https://grisoon.github.io/DreamActor-M2/
• Github: https://grisoon.github.io/DreamActor-M2/
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#CharacterAnimation #AI #ComputerVision #DeepLearning #GenerativeAI
✨DenseGRPO: From Sparse to Dense Reward for Flow Matching Model Alignment
📝 Summary:
DenseGRPO addresses sparse rewards in flow matching models by providing dense, step-wise rewards for intermediate denoising steps. It uses these rewards to adaptively calibrate exploration, improving alignment with human preferences in text-to-image generation.
🔹 Publication Date: Published on Jan 28
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.20218
• PDF: https://arxiv.org/pdf/2601.20218
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #MachineLearning #ReinforcementLearning #TextToImage #GenerativeAI
📝 Summary:
DenseGRPO addresses sparse rewards in flow matching models by providing dense, step-wise rewards for intermediate denoising steps. It uses these rewards to adaptively calibrate exploration, improving alignment with human preferences in text-to-image generation.
🔹 Publication Date: Published on Jan 28
🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.20218
• PDF: https://arxiv.org/pdf/2601.20218
==================================
For more data science resources:
✓ https://t.iss.one/DataScienceT
#AI #MachineLearning #ReinforcementLearning #TextToImage #GenerativeAI
✨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
📝 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
📝 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
📝 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
📝 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
📝 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