AnTKV: Anchor Token-Aware Sub-Bit Vector Quantization for KV Cache in Large Language Models

Quantization has emerged as an effective and lightweight solution to reduce the memory footprint of the KV cache in Large Language Models (LLMs). Nevertheless, minimizing the performance degradation caused by ultra-low-bit KV cache quantization remains a significant challenge. We observe that quantizing the KV cache of different tokens has varying impacts on the quality of attention outputs. To systematically investigate this phenomenon, we perform forward error propagation analysis on attention and propose the Anchor Score (AnS) that quantifies the sensitivity of each token's KV cache to quantization-induced error. Our analysis reveals significant disparities in AnS across tokens, suggesting that preserving a small subset with full precision (FP16) of high-AnS tokens can greatly mitigate accuracy loss in aggressive quantization scenarios. Based on this insight, we introduce AnTKV, a novel framework that leverages Anchor Token-aware Vector Quantization to compress the KV cache. Furthermore, to support efficient deployment, we design and develop a triton kernel that is fully compatible with FlashAttention, enabling fast online Anchor Token selection. AnTKV enables LLaMA-3-8B to handle context lengths up to 840K tokens on a single 80GB A100 GPU, while achieving up to 3.5x higher decoding throughput compared to the FP16 baseline. Our experiment results demonstrate that AnTKV matches or outperforms prior works such as KIVI, SKVQ, KVQuant, and CQ under 4-bit settings. More importantly, AnTKV achieves significantly lower perplexity under ultra-low-bit quantization on Mistral-7B, with only 6.32 at 1-bit and 8.87 at 0.375-bit, compared to the FP16 baseline of 4.73.

Combining Self-attention and Dilation Convolutional for Semantic Segmentation of Coal Maceral Groups

The segmentation of coal maceral groups can be described as a semantic segmentation process of coal maceral group images, which is of great significance for studying the chemical properties of coal. Generally, existing semantic segmentation models of coal maceral groups use the method of stacking parameters to achieve higher accuracy. It leads to increased computational requirements and impacts model training efficiency. At the same time, due to the professionalism and diversity of coal maceral group images sampling, obtaining the number of samples for model training requires a long time and professional personnel operation. To address these issues, We have innovatively developed an IoT-based DA-VIT parallel network model. By utilizing this model, we can continuously broaden the dataset through IoT and achieving sustained improvement in the accuracy of coal maceral groups segmentation. Besides, we decouple the parallel network from the backbone network to ensure the normal using of the backbone network during model data updates. Secondly, DCSA mechanism of DA-VIT is introduced to enhance the local feature information of coal microscopic images. This DCSA can decompose the large kernels of convolutional attention into multiple scales and reduce 81.18% of parameters.Finally, we performed the contrast experiment and ablation experiment between DA-VIT and state-of-the-art methods at lots of evaluation metrics. Experimental results show that DA-VIT-Base achieves 92.14% pixel accuracy and 63.18% mIoU. Params and FLOPs of DA-VIT-Tiny are 4.95M and 8.99G, respectively. All of the evaluation metrics of the proposed DA-VIT are better than other state-of-the-art methods.

Styl3R: Instant 3D Stylized Reconstruction for Arbitrary Scenes and Styles

Stylizing 3D scenes instantly while maintaining multi-view consistency and faithfully resembling a style image remains a significant challenge. Current state-of-the-art 3D stylization methods typically involve computationally intensive test-time optimization to transfer artistic features into a pretrained 3D representation, often requiring dense posed input images. In contrast, leveraging recent advances in feed-forward reconstruction models, we demonstrate a novel approach to achieve direct 3D stylization in less than a second using unposed sparse-view scene images and an arbitrary style image. To address the inherent decoupling between reconstruction and stylization, we introduce a branched architecture that separates structure modeling and appearance shading, effectively preventing stylistic transfer from distorting the underlying 3D scene structure. Furthermore, we adapt an identity loss to facilitate pre-training our stylization model through the novel view synthesis task. This strategy also allows our model to retain its original reconstruction capabilities while being fine-tuned for stylization. Comprehensive evaluations, using both in-domain and out-of-domain datasets, demonstrate that our approach produces high-quality stylized 3D content that achieve a superior blend of style and scene appearance, while also outperforming existing methods in terms of multi-view consistency and efficiency.

How does Alignment Enhance LLMs' Multilingual Capabilities? A Language Neurons Perspective

Multilingual Alignment is an effective and representative paradigm to enhance LLMs' multilingual capabilities, which transfers the capabilities from the high-resource languages to the low-resource languages. Meanwhile, some researches on language-specific neurons reveal that there are language-specific neurons that are selectively activated in LLMs when processing different languages. This provides a new perspective to analyze and understand LLMs' mechanisms more specifically in multilingual scenarios. In this work, we propose a new finer-grained neuron identification algorithm, which detects language neurons~(including language-specific neurons and language-related neurons) and language-agnostic neurons. Furthermore, based on the distributional characteristics of different types of neurons, we divide the LLMs' internal process for multilingual inference into four parts: (1) multilingual understanding, (2) shared semantic space reasoning, (3) multilingual output space transformation, and (4) vocabulary space outputting. Additionally, we systematically analyze the models before and after alignment with a focus on different types of neurons. We also analyze the phenomenon of ''Spontaneous Multilingual Alignment''. Overall, our work conducts a comprehensive investigation based on different types of neurons, providing empirical results and valuable insights for better understanding multilingual alignment and multilingual capabilities of LLMs.

Can Compressed LLMs Truly Act? An Empirical Evaluation of Agentic Capabilities in LLM Compression

Post-training compression reduces the computational and memory costs of large language models (LLMs), enabling resource-efficient deployment. However, existing compression benchmarks only focus on language modeling (e.g., perplexity) and natural language understanding tasks (e.g., GLUE accuracy), ignoring the agentic capabilities - workflow, tool use/function call, long-context understanding and real-world application. We introduce the Agent Compression Benchmark (ACBench), the first comprehensive benchmark for evaluating how compression impacts LLMs' agentic abilities. ACBench spans (1) 12 tasks across 4 capabilities (e.g., WorfBench for workflow generation, Needle-in-Haystack for long-context retrieval), (2) quantization (GPTQ, AWQ) and pruning (Wanda, SparseGPT), and (3) 15 models, including small (Gemma-2B), standard (Qwen2.5 7B-32B), and distilled reasoning LLMs (DeepSeek-R1-Distill). Our experiments reveal compression tradeoffs: 4-bit quantization preserves workflow generation and tool use (1%-3% drop) but degrades real-world application accuracy by 10%-15%. We introduce ERank, Top-k Ranking Correlation and Energy to systematize analysis. ACBench provides actionable insights for optimizing LLM compression in agentic scenarios. The code can be found in https://github.com/pprp/ACBench.

Efficient Reasoning via Chain of Unconscious Thought

Large Reasoning Models (LRMs) achieve promising performance but compromise token efficiency due to verbose reasoning processes. Unconscious Thought Theory (UTT) posits that complex problems can be solved more efficiently through internalized cognitive processes. Inspired by UTT, we propose a new reasoning paradigm, termed Chain of Unconscious Thought (CoUT), to improve the token efficiency of LRMs by guiding them to mimic human unconscious thought and internalize reasoning processes. Concretely, we first prompt the model to internalize the reasoning by thinking in the hidden layer. Then, we design a bag of token-efficient strategies to further help models reduce unnecessary tokens yet preserve the performance. Our work reveals that models may possess beneficial unconscious thought, enabling improved efficiency without sacrificing performance. Extensive experiments demonstrate the effectiveness of CoUT. Remarkably, it surpasses CoT by reducing token usage by 47.62% while maintaining comparable accuracy, as shown in Figure 1. The code of CoUT is available at this link: https://github.com/Rohan-GRH/CoUT

SLearnLLM: A Self-Learning Framework for Efficient Domain-Specific Adaptation of Large Language Models

When using supervised fine-tuning (SFT) to adapt large language models (LLMs) to specific domains, a significant challenge arises: should we use the entire SFT dataset for fine-tuning? Common practice often involves fine-tuning directly on the entire dataset due to limited information on the LLM's past training data. However, if the SFT dataset largely overlaps with the model's existing knowledge, the performance gains are minimal, leading to wasted computational resources. Identifying the unknown knowledge within the SFT dataset and using it to fine-tune the model could substantially improve the training efficiency. To address this challenge, we propose a self-learning framework for LLMs inspired by human learning pattern. This framework takes a fine-tuning (SFT) dataset in a specific domain as input. First, the LLMs answer the questions in the SFT dataset. The LLMs then objectively grade the responses and filter out the incorrectly answered QA pairs. Finally, we fine-tune the LLMs based on this filtered QA set. Experimental results in the fields of agriculture and medicine demonstrate that our method substantially reduces training time while achieving comparable improvements to those attained with full dataset fine-tuning. By concentrating on the unknown knowledge within the SFT dataset, our approach enhances the efficiency of fine-tuning LLMs.

Multimodal Online Federated Learning with Modality Missing in Internet of Things

The Internet of Things (IoT) ecosystem generates vast amounts of multimodal data from heterogeneous sources such as sensors, cameras, and microphones. As edge intelligence continues to evolve, IoT devices have progressed from simple data collection units to nodes capable of executing complex computational tasks. This evolution necessitates the adoption of distributed learning strategies to effectively handle multimodal data in an IoT environment. Furthermore, the real-time nature of data collection and limited local storage on edge devices in IoT call for an online learning paradigm. To address these challenges, we introduce the concept of Multimodal Online Federated Learning (MMO-FL), a novel framework designed for dynamic and decentralized multimodal learning in IoT environments. Building on this framework, we further account for the inherent instability of edge devices, which frequently results in missing modalities during the learning process. We conduct a comprehensive theoretical analysis under both complete and missing modality scenarios, providing insights into the performance degradation caused by missing modalities. To mitigate the impact of modality missing, we propose the Prototypical Modality Mitigation (PMM) algorithm, which leverages prototype learning to effectively compensate for missing modalities. Experimental results on two multimodal datasets further demonstrate the superior performance of PMM compared to benchmarks.

FlowKV: Enhancing Multi-Turn Conversational Coherence in LLMs via Isolated Key-Value Cache Management

Large Language Models (LLMs) are increasingly deployed in multi-turn conversational applications, where the management of the Key-Value (KV) Cache presents a significant bottleneck. The linear growth of the KV Cache with dialogue history imposes substantial computational costs, and existing eviction strategies often degrade performance by repeatedly compressing early conversational context, leading to information loss and context forgetting. This paper introduces FlowKV, a novel \textbf{multi-turn isolation mechanism} for KV Cache management, which can be applied to any KV Cache compression method without training. FlowKV's core innovation is a multi-turn isolation mechanism that preserves the accumulated compressed KV cache from past turns. Compression is then strategically applied only to the newly generated KV pairs of the latest completed turn, effectively preventing the re-compression of older context and thereby mitigating catastrophic forgetting. Our results demonstrate that FlowKV consistently and significantly outperforms baseline strategies in maintaining instruction-following accuracy and user preference retention from 10.90\% to 75.40\%, particularly in later conversational turns.

SSR: Speculative Parallel Scaling Reasoning in Test-time

Large language models (LLMs) have achieved impressive results on multi-step mathematical reasoning, yet at the cost of high computational overhead. This challenge is particularly acute for test-time scaling methods such as parallel decoding, which increase answer diversity but scale poorly in efficiency. To address this efficiency-accuracy trade-off, we propose SSR (Speculative Parallel Scaling Reasoning), a training-free framework that leverages a key insight: by introducing speculative decoding at the step level, we can accelerate reasoning without sacrificing correctness. SSR integrates two components: a Selective Parallel Module (SPM) that identifies a small set of promising reasoning strategies via model-internal scoring, and Step-level Speculative Decoding (SSD), which enables efficient draft-target collaboration for fine-grained reasoning acceleration. Experiments on three mathematical benchmarks-AIME 2024, MATH-500, and LiveMathBench - demonstrate that SSR achieves strong gains over baselines. For instance, on LiveMathBench, SSR improves pass@1 accuracy by 13.84% while reducing computation to 80.5% of the baseline FLOPs. On MATH-500, SSR reduces compute to only 30% with no loss in accuracy.