🔥 KVarN: Variance-Normalized KV-Cache Quantization Mitigates Error Accumulation in Reasoning Tasks

💡 The paper introduces KVarN, a new method for quantizing KV-cache in large language models to reduce error accumulation during autoregressive decoding. The problem addressed is that test-time scaling, which improves reasoning in large language models, becomes memory-bottlenecked during long-horizon decoding due to the growing KV-cache. Existing KV-cache quantization methods are not effective in this setting because they are evaluated under prefill-like settings, where errors behave differently than in autoregressive decoding. In autoregressive decoding, quantization errors accumulate across timesteps, primarily due to incorrect token scales.

The KVarN method addresses this issue by applying a Hadamard rotation followed by a dual-scaling variance normalization across both axes of the K and V matrices. This combination fixes outlying token-scale errors and substantially reduces error accumulation. The method is calibration-free, meaning it does not require any additional calibration steps.

The results show that KVarN establishes a new state-of-the-art for KV-cache quantization on generative benchmarks, including MATH500, AIME24, and HumanEval, at 2-bit precision. This means that KVarN is able to achieve better performance than existing methods while using less memory. The KVarN method is also available for implementation in large language models, providing a practical solution to the problem of error accumulation in autoregressive decoding. Overall, the paper contributes a new and effective method for quantizing KV-cache in large language models, which can improve the performance and efficiency of these models in reasoning tasks.

📅 Published on Jun 2

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2606.03458
• PDF: https://arxiv.org/pdf/2606.03458
• Project Page: https://github.com/huawei-csl/KVarN

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

#KVCacheQuantization #AutoregressiveDecoding #LargeLanguageModels #ErrorAccumulationMitigation #QuantizationMethodsForReasoningTasks

🔥 Role-Agent: Bootstrapping LLM Agents via Dual-Role Evolution

💡 The paper introduces the Role-Agent framework, which aims to improve the performance of Large Language Model agents by addressing the limitations of inefficient interaction feedback and static training environments. The problem with current Large Language Model agents is that their learning is hindered by the lack of effective feedback and the inability to adapt to changing environments, resulting in limited generalization.

To address this issue, the Role-Agent framework enables a single Large Language Model to function as both the agent and the environment, allowing for a bootstrapped co-evolution process. This framework consists of two components: World-In-Agent and Agent-In-World. The World-In-Agent component uses the Large Language Model as the agent to predict future states after each action, and the alignment between predicted and actual states is used as a reward to encourage environment-aware reasoning.

The Agent-In-World component analyzes failure modes from failed trajectories and retrieves tasks with similar failure patterns, thereby reshaping the training data distribution for targeted practice. This allows the Large Language Model to focus on improving its performance in areas where it is struggling.

The results of the experiments show that the Role-Agent framework consistently improves performance, with an average gain of over 4 percent over strong baselines. This demonstrates the effectiveness of the Role-Agent framework in improving the performance of Large Language Model agents by enabling them to adapt to changing environments and focus on targeted practice. Overall, the Role-Agent framework provides a novel approach to improving the performance of Large Language Model agents, and its results have significant implications for the development of more effective and adaptive language models.

📅 Published on Jun 9

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2606.10917
• PDF: https://arxiv.org/pdf/2606.10917

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

#LargeLanguageModels #AgentEnvironmentInteraction #DualRoleEvolution #BootstrappedLearning #CoEvolutionaryAlgorithms

🔥 LMCache: An Efficient KV Cache Layer for Enterprise-Scale LLM Inference

💡 The paper presents LMCache, an efficient key-value cache layer for large language model inference at the enterprise scale. The problem addressed is the traditional storage of key-value caches in GPU memory, which limits cache reuse across different queries and inference engines. As the total key-value cache stored by users grows rapidly, exceeding the capacity of GPU memory, there is a need to move caches outside GPU devices.

The authors propose LMCache as a solution, which extracts and stores key-value caches generated by modern large language model engines out of the GPU memory and shares them across engines and queries. LMCache supports cache offloading and prefill-decode disaggregation, allowing for cross-engine and GPU cache transfer. The key contributions of LMCache include highly optimized key-value cache data movement, a modular cache connector component that decouples LMCache from the evolution of inference engines, and a control API for flexible cache orchestration across different layers.

The evaluation of LMCache shows significant improvements in throughput, with up to 15 times improvement when combined with a large language model engine. The adoption of LMCache in enterprise settings provides valuable insights, such as the benefits of fetching key-value caches from remote storage and the impact of context truncation on prefix cache hit ratio. Overall, LMCache is presented as an efficient and open-source key-value caching solution that addresses the need for efficient cache management in large language model inference.

📅 Published on Oct 8, 2025

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2510.09665
• PDF: https://arxiv.org/pdf/2510.09665
• Project Page: https://huggingface.co/collections/dvps/dvps-scientific-watch

🤖 Models citing this paper:
• https://huggingface.co/enfinity7B/apac

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

#LargeLanguageModels #LLMInference #KVCacheOptimization #EnterpriseScaleAI #GPUAcceleratedInference

🔥 Foundations of Large Language Models

💡 The book Foundations of Large Language Models provides a comprehensive overview of the fundamental concepts underlying large language models. The book is structured into four main chapters, each focusing on a key area: pre-training, generative models, prompting techniques, and alignment methods. The authors aim to provide a foundational understanding of large language models, rather than a comprehensive coverage of all cutting-edge technologies. The book is intended for college students, professionals, and practitioners in natural language processing and related fields, serving as a reference for anyone interested in large language models.

The problem addressed by the book is the need for a clear understanding of the foundational concepts of large language models, which are becoming increasingly important in natural language processing. The method used to address this problem is a structured approach, dividing the topic into four key areas and exploring each in depth. The results of this approach are a book that provides a solid foundation for understanding large language models, which can be used as a reference by students, professionals, and practitioners in the field.

Overall, the book provides a foundational understanding of large language models, covering key areas such as pre-training, generative models, prompting techniques, and alignment methods, and is intended to serve as a reference for those interested in this topic. The book does not aim to cover all cutting-edge technologies, but rather provides a solid foundation for understanding the underlying concepts of large language models.

📅 Published on Jan 16, 2025

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2501.09223
• PDF: https://arxiv.org/pdf/2501.09223

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

#LargeLanguageModels #NaturalLanguageProcessing #PreTrainingMethods #GenerativeModels #LanguageModelAlignment

🔥 FastContext: Training Efficient Repository Explorer for Coding Agents

💡 The paper introduces FastContext, a dedicated exploration subagent designed to improve the efficiency of repository exploration in large language model coding agents. The problem addressed is that repository exploration is a major bottleneck in coding agents, consuming a substantial token budget and polluting the agent's context with irrelevant code snippets.

The method involves separating repository exploration from code solving using specialized exploration models. FastContext is invoked on demand and issues parallel tool calls to return concise file paths and line ranges as focused context. The exploration models used in FastContext are powered by 4B-30B parameters and are bootstrapped from strong reference-model trajectories. They are then refined with task-grounded rewards for broad first-turn search, multi-turn evidence gathering, and precise citation generation.

The results show that integrating FastContext into a coding agent improves end-to-end resolution rates by up to 5.5 percent while reducing coding-agent token consumption by up to 60 percent, with minimal overhead. The paper demonstrates that repository exploration can be effectively handled by specialized models, separate from the code solving process. The code and data for FastContext are made available, allowing for further research and development in this area. Overall, the paper presents a significant contribution to the field of coding agents and software engineering, providing a more efficient and effective approach to repository exploration.

📅 Published on Jun 12

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2606.14066
• PDF: https://arxiv.org/pdf/2606.14066
• Project Page: https://huggingface.co/microsoft/FastContext-1.0-4B-SFT

🤖 Models citing this paper:
• https://huggingface.co/microsoft/FastContext-1.0-4B-SFT
• https://huggingface.co/microsoft/FastContext-1.0-4B-RL

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

#EfficientRepositoryExploration #CodingAgents #LargeLanguageModels #RepositoryExplorationSubagents #SpecializedExplorationModels

🔥 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

🔥 Enhancing Financial Sentiment Analysis via Retrieval Augmented Large Language Models

💡 The paper addresses the challenge of financial sentiment analysis, which is crucial for investment decision-making. Traditional natural language processing models are limited by their size and training data, resulting in poor generalization and effectiveness. Large Language Models, despite their superior performance in various NLP tasks, also face challenges in financial sentiment analysis due to the discrepancy between their pre-training objective and the task of predicting sentiment labels. Additionally, the concise nature of financial news often lacks sufficient context, which can compromise the reliability of Large Language Models' sentiment analysis.

To overcome these challenges, the authors propose a retrieval-augmented Large Language Model framework. This framework consists of two modules: an instruction-tuned Large Language Model module that ensures the model behaves as a predictor of sentiment labels, and a retrieval-augmentation module that retrieves additional context from reliable external sources. This approach enables the model to leverage external context to improve its sentiment analysis capabilities.

The authors evaluate their framework against traditional models and other Large Language Models, such as ChatGPT and LLaMA. The results show that their approach achieves a significant performance gain, with improvements in accuracy and F1 score ranging from 15% to 48%. This demonstrates the effectiveness of the proposed retrieval-augmented Large Language Model framework in enhancing financial sentiment analysis. Overall, the paper contributes to the development of more accurate and reliable financial sentiment analysis models, which can inform better investment decisions.

📅 Published on Oct 6, 2023

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2310.04027
• PDF: https://arxiv.org/pdf/2310.04027

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

#FinancialSentimentAnalysis #RetrievalAugmentedModels #LargeLanguageModels #NaturalLanguageProcessing #FinancialTextAnalysis

🔥 Efficient Guided Generation for Large Language Models

💡 The paper presents an efficient method for guiding large language model text generation using regular expressions and context-free grammars. The problem addressed is that guided generation can be impractical due to significant overhead. The authors propose an approach that adds minimal overhead to the token sequence generation process. This method makes guided generation feasible in practice. The approach is implemented in the open source Python library Outlines, providing a practical solution for efficient guided generation. The results indicate that the method is effective, allowing for guided generation with little to no overhead, which is a significant contribution to the field of natural language processing.

📅 Published on Jul 19, 2023

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2307.09702
• PDF: https://arxiv.org/pdf/2307.09702

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

#LargeLanguageModels #GuidedTextGeneration #RegularExpressions #ContextFreeGrammars #EfficientGenerationMethods

🔥 JetSpec: Breaking the Scaling Ceiling of Speculative Decoding with Parallel Tree Drafting

💡 The paper introduces JetSpec, a speculative decoding framework designed to improve the inference speed and acceptance rates of large language models. The problem addressed is the scaling limitation of speculative decoding, which accelerates autoregressive large language models by drafting multiple tokens and verifying them in parallel. However, increasing the draft budget only improves speed when acceptance remains high and drafting overhead stays low, creating a scaling ceiling.

The proposed JetSpec framework combines efficient forward drafting with causal conditioning to break this ceiling. It trains a causal parallel draft head over fused hidden states from the frozen target model, producing candidate trees whose scores align with the target model's autoregressive factorization. This approach enables JetSpec to convert larger draft budgets into longer accepted prefixes and higher end-to-end speedup.

The method is compared to bidirectional-head and tree-based speculative decoding baselines across various benchmarks, including math, coding, and chat tasks on dense and MoE models. The results show that JetSpec consistently outperforms these baselines, achieving significant speedup on different workloads. Specifically, JetSpec achieves up to 9.64x speedup on math tasks and 4.58x on open-ended conversational workloads, with further latency gains demonstrated through integration with virtual large language models under realistic serving loads.

Overall, the paper contributes a novel speculative decoding framework that breaks the scaling ceiling of prior methods, enabling faster and more efficient large language model inference. The code and models are made available for further research and development.

📅 Published on Jun 25

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2606.18394
• PDF: https://arxiv.org/pdf/2606.18394
• Project Page: https://jetspec-project.github.io/jetspec-web/

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

#SpeculativeDecoding #LargeLanguageModels #AutoregressiveModeling #ParallelTreeDrafting #CausalConditioning

🔥 Scaling the Horizon, Not the Parameters: Reaching Trillion-Parameter Performance with a 35B Agent

💡 The paper introduces Agents-A1, a 35 billion parameter Mixture-of-Experts Agentic Model that achieves performance comparable to trillion-parameter models by scaling the agent horizon instead of the parameters. The problem addressed is how to improve the performance of large language models on long-horizon tasks without increasing the number of parameters. The method used is a three-stage training approach, which includes supervised fine-tuning, domain-level teacher models, and multi-teacher distillation. The model is trained on a long-horizon knowledge-action infrastructure that connects external knowledge, actions, observations, and verifier outcomes, producing agentic trajectories with an average length of 45,000 tokens. The results show that Agents-A1 achieves strong and broad performance on long-horizon agent benchmarks, outperforming or matching the results of 1 trillion parameter models on several tasks, including SEAL-0, IFBench, HiPhO, FrontierScience-Olympiad, and MolBench-Bind. The paper provides a practical path for scaling the horizon using a smaller model that can reach or match the performance of larger models on long-horizon tasks.

📅 Published on Jun 29

🔗 Links:
• GitHub: https://github.com/huggingface
• arXiv: https://arxiv.org/abs/2606.30616
• PDF: https://arxiv.org/pdf/2606.30616
• Project Page: https://internscience.github.io/Agents-A1/

🤖 Models citing this paper:
• https://huggingface.co/InternScience/Agents-A1
• https://huggingface.co/InternScience/Agents-A1-FP8-dynamic
• https://huggingface.co/Abiray/Agents-A1-Q4_K_M-GGUF

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

#MixtureOfExperts #AgenticModels #LongHorizonTasks #LargeLanguageModels #ParameterEfficientTraining