✨End-to-End Test-Time Training for Long Context

📝 Summary:
This paper proposes End-to-End Test-Time Training TTT-E2E for long-context language modeling, treating it as continual learning. It uses a standard Transformer, learning at test time and improving initialization via meta-learning. TTT-E2E scales well and offers constant inference latency, being m...

🔹 Publication Date: Published on Dec 29, 2025

🔹 Paper Links:
• arXiv Page: https://arxivlens.com/PaperView/Details/end-to-end-test-time-training-for-long-context-6176-bf8fd7e6
• PDF: https://arxiv.org/pdf/2512.23675
• Github: https://github.com/test-time-training/e2e

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

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#TestTimeTraining #LongContext #LanguageModels #Transformers #ContinualLearning

✨Memory Bank Compression for Continual Adaptation of Large Language Models

📝 Summary:
Memory-augmented continual learning for LLMs faces growing memory bank issues. MBC compresses these banks via codebook optimization and an online resetting mechanism, using Key-Value Low-Rank Adaptation. It reduces bank size to 0.3 percent while maintaining high accuracy.

🔹 Publication Date: Published on Jan 2

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.00756
• PDF: https://arxiv.org/pdf/2601.00756
• Github: https://github.com/Thomkat/MBC

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

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#LLMs #ContinualLearning #MemoryCompression #MachineLearning #DeepLearning

✨MemOS: A Memory OS for AI System

📝 Summary:
MemOS is a memory operating system for LLMs that unifies plaintext, activation-based, and parameter-level memories. It treats memory as a system resource, using MemCubes for efficient storage, retrieval, and enabling continual learning and personalized modeling.

🔹 Publication Date: Published on Jul 4, 2025

🔹 Paper Links:
• arXiv Page: https://arxivlens.com/PaperView/Details/memos-a-memory-os-for-ai-system-4846-c5e0c676
• PDF: https://arxiv.org/pdf/2507.03724
• Project Page: https://memos.openmem.net/
• Github: https://github.com/MemTensor/MemOS

🔹 Models citing this paper:
• https://huggingface.co/kagvi13/HMP

✨ Datasets citing this paper:
• https://huggingface.co/datasets/MemTensor/MemOS_eval_result

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

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#AI #LLMs #MemoryOS #ContinualLearning #SystemDesign

✨CLARE: Continual Learning for Vision-Language-Action Models via Autonomous Adapter Routing and Expansion

📝 Summary:
CLARE enables robots to continually learn new tasks without forgetting, using lightweight adapters. It autonomously expands these adapters and dynamically routes them, ensuring high performance without needing task labels or storing past data.

🔹 Publication Date: Published on Jan 14

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.09512
• PDF: https://arxiv.org/pdf/2601.09512
• Project Page: https://tum-lsy.github.io/clare/
• Github: https://github.com/utiasDSL/clare

✨ Datasets citing this paper:
• https://huggingface.co/datasets/continuallearning/libero_10_image_task_0

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

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#ContinualLearning #Robotics #AI #MachineLearning #VLAModels

✨Knowledge is Not Enough: Injecting RL Skills for Continual Adaptation

📝 Summary:
LLMs struggle to apply new knowledge effectively via SFT alone. PaST combines SFT with injecting a domain-agnostic Skill Vector, derived from RL, to efficiently transfer reasoning skills. This novel framework significantly improves performance in question answering and tool-use tasks.

🔹 Publication Date: Published on Jan 16

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.11258
• PDF: https://arxiv.org/pdf/2601.11258

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

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#LLM #ReinforcementLearning #ContinualLearning #AI #MachineLearning

✨Continual GUI Agents

📝 Summary:
The Continual GUI Agents framework addresses performance degradation in dynamic UI environments. It introduces GUI-Anchoring in Flux GUI-AiF, a reinforcement fine-tuning method with novel anchoring rewards that stabilize learning across shifting UI domains and resolutions, outperforming existing ...

🔹 Publication Date: Published on Jan 28

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2601.20732
• PDF: https://arxiv.org/pdf/2601.20732

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

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#ContinualLearning #ReinforcementLearning #AIAgents #HumanComputerInteraction #MachineLearning

✨XSkill: Continual Learning from Experience and Skills in Multimodal Agents

📝 Summary:
XSkill is a dual-stream framework for continual learning in multimodal agents. It extracts and retrieves knowledge from visual observations, consolidating experiences and skills. This improves tool use efficiency, reasoning, and zero-shot generalization.

🔹 Publication Date: Published on Mar 12

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.12056
• PDF: https://arxiv.org/pdf/2603.12056
• Project Page: https://xskill-agent.github.io/xskill_page/
• Github: https://github.com/XSkill-Agent/XSkill

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

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#ContinualLearning #MultimodalAI #AIagents #MachineLearning #Robotics

✨Simple Recipe Works: Vision-Language-Action Models are Natural Continual Learners with Reinforcement Learning

📝 Summary:
Contrary to established belief, simple sequential fine-tuning with low-rank adaptation is highly effective for continual reinforcement learning in large Vision-Language-Action models. It achieves excellent plasticity and avoids catastrophic forgetting, often outperforming complex methods.

🔹 Publication Date: Published on Mar 12

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2603.11653
• PDF: https://arxiv.org/pdf/2603.11653

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

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#ReinforcementLearning #ContinualLearning #VLAmodels #AI #MachineLearning

✨Brainstacks: Cross-Domain Cognitive Capabilities via Frozen MoE-LoRA Stacks for Continual LLM Learning

📝 Summary:
Brainstacks enables continual multi-domain fine-tuning of large language models through modular adapter stacks with MoE-LoRA, residual boosting, and outcome-based routing that discovers transferable c...

🔹 Publication Date: Published on Apr 1

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2604.01152
• PDF: https://arxiv.org/pdf/2604.01152
• Project Page: https://huggingface.co/papers?q=null-space%20projection
• Github: https://github.com/achelousace/brainstacks

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

For more data science resources:
✓ https://xn--r1a.website/DataScienceT

#LLM #ContinualLearning #MoELoRA #DeepLearning #AIResearch

🔥 From RAG to Memory: Non-Parametric Continual Learning for Large Language Models

💡 The paper discusses the challenges of continual learning in large language models and how current methods such as retrieval-augmented generation have limitations in mimicking human long-term memory. The authors propose a new framework called HippoRAG 2 which builds upon previous work and enhances it with deeper passage integration and more effective online use of a large language model. This approach improves performance across factual, sense-making, and associative memory tasks, addressing the deterioration in performance seen in previous methods that tried to augment vector embeddings with structures like knowledge graphs. The results show that HippoRAG 2 outperforms standard retrieval-augmented generation comprehensively, achieving a 7 percent improvement in associative memory tasks over the state-of-the-art embedding model, while also exhibiting superior factual knowledge and sense-making memory capabilities. The work contributes to non-parametric continual learning for large language models, paving the way for more effective and human-like memory capabilities in artificial intelligence systems.

📅 Published on Feb 20, 2025

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2502.14802
• PDF: https://arxiv.org/pdf/2502.14802

🤖 Models citing this paper:
• https://huggingface.co/muthuk1/graphrag-inference-hackathon

📊 Datasets citing this paper:
• https://huggingface.co/datasets/osunlp/HippoRAG_2
• https://huggingface.co/datasets/g7haha/HippoRAG_2

━━━━━━━━━━━━━━━━━━━━━━━━
📢 By: https://xn--r1a.website/PaperNexus

#ContinualLearning #LargeLanguageModels #NonParametricLearning #RetrievalAugmentedGeneration #LongTermMemory