FlashMemory-DeepSeek-V4: Lightning Index Ultra-Long Context via Lookahead Sparse Attention

• URL: http://arxiv.org/abs/2606.09079v2
• Date: Tue, 09 Jun 2026 17:04:45 GMT
• Title: FlashMemory-DeepSeek-V4: Lightning Index Ultra-Long Context via Lookahead Sparse Attention
• Abstract summary: Lookahead Sparse Attention (LSA) is powered by a Neural Memory Indexer built upon the DeepSeek-V4 architecture.<n>We instantiate this architecture via a backbone-free decoupled training strategy.<n>We show that FM-DS-V4 compresses the average physical KV cache footprint down to merely 13.5% of the full-context baseline.
• Score: 77.12062766962815
• License: http://creativecommons.org/publicdomain/zero/1.0/
• Abstract: Conventional LLMs keep the full KV cache loaded during decoding, causing a severe GPU memory bottleneck for ultra-long context serving. In this report, we propose Lookahead Sparse Attention (LSA), a novel inference paradigm powered by a Neural Memory Indexer built upon the DeepSeek-V4 architecture. Rather than passively attending to all historical tokens, LSA proactively predicts future context demands and preserves only the query-critical KV chunks in the GPU memory. Crucially, we instantiate this architecture via a backbone-free decoupled training strategy. By formulating the indexer as a standard dual-encoder architecture, we train it independently using standard retrieval training frameworks without ever loading the massive backbone model into GPU memory. We demonstrate that this "less is more" paradigm significantly maximizes serving efficiency while acting as an effective attention denoiser in tasks that rely on long-term global memory. Across primary long-context evaluation suites (e.g., LongBench-v2, LongMemEval, and RULER), FM-DS-V4 compresses the average physical KV cache footprint down to merely 13.5% of the full-context baseline, while consistently preserving or slightly elevating downstream accuracy (+0.6% absolute margin on average). Crucially, at extreme 500K scales, FlashMemory suppresses the physical KV cache overhead by over 90% without destabilizing the backbone's core reasoning capacities.

Related papers

• SeKV: Resolution-Adaptive KV Cache with Hierarchical Semantic Memory for Long-Context LLM Inference [37.37437232987781]
  SeKV is a resolution-adaptive semantic KV cache that organizes context into entropy-guided semantic spans.<n>We show that SeKV improves over the strongest semantic compression baseline by 5.9% on average.<n>It reduces GPU memory by 53.3% versus full KV caching at 128K context.
  arXiv  Detail & Related papers  (2026-06-30T05:18:02Z)
• KVDrive: A Holistic Multi-Tier KV Cache Management System for Long-Context LLM Inference [23.683284557561112]
  We present KVDrive, a holistic multi-tier KV cache management system spanning GPU memory, host DRAM, and SSD.<n> KVDrive tackles the problem from a systems perspective - jointly orchestrating cache placement, pipeline scheduling, and cross-tier coordination.<n>The system achieves up to 1.74x higher throughput compared to state-of-the-art works while preserving accuracy.
  arXiv  Detail & Related papers  (2026-05-18T08:54:16Z)
• Semantic-Aware Adaptive Visual Memory for Streaming Video Understanding [55.7992006853979]
  SAVEMem is a framework that brings semantic awareness into memory generation and lets the retrieval scope adapt per query.<n>We present SAVEMem, a training-free dual-stage framework that brings semantic awareness into memory generation and lets the retrieval scope adapt per query.
  arXiv  Detail & Related papers  (2026-05-08T15:40:40Z)
• DeltaKV: Residual-Based KV Cache Compression via Long-Range Similarity [50.52392445266824]
  We propose a residual-based KV cache compression framework motivated by long-range inter-token similarity and highly shared latent components in KV representations.<n>Instead of discarding tokens, DeltaKV encodes semantic residuals relative to retrieved historical references, preserving fidelity while substantially reducing storage.<n>Experiments show that DeltaKV reduces KV cache memory to 29% of the original while maintaining near-lossless accuracy on LongBench, SCBench, and AIME.
  arXiv  Detail & Related papers  (2026-02-08T15:14:36Z)
• Out of the Memory Barrier: A Highly Memory Efficient Training System for LLMs with Million-Token Contexts [68.79341332280062]
  Training Large Language Models (LLMs) on long contexts is severely constrained by prohibitive GPU memory overhead, not training time.<n>We introduce OOMB, a highly memory-efficient training system that directly confronts this barrier.<n>Our approach employs a chunk-recurrent training framework with on-the-fly activation recomputation, which maintains a constant activation memory footprint.
  arXiv  Detail & Related papers  (2026-02-02T13:52:40Z)
• HCAttention: Extreme KV Cache Compression via Heterogeneous Attention Computing for LLMs [13.013668526921778]
  Existing KV cache compression methods exhibit noticeable performance degradation when memory is reduced by more than 85%.<n>We propose HCAttention, a heterogeneous attention framework that integrates key quantization, value offloading, and dynamic KV eviction.<n> Experimental results on the LongBench benchmark demonstrate that our approach preserves the accuracy of full-attention model while shrinking the KV cache memory footprint to 25% of its original size.
  arXiv  Detail & Related papers  (2025-07-26T06:43:14Z)
• Paged Attention Meets FlexAttention: Unlocking Long-Context Efficiency in Deployed Inference [1.0175051111288864]
  We introduce a novel integration of PagedAttention with PyTorch's FlexAttention.<n>Our benchmarks on an NVIDIA L4 GPU demonstrate significantly reduced inference latency.<n>We open-source the full implementation and discuss its implications for future long-context model deployment.
  arXiv  Detail & Related papers  (2025-06-08T22:59:20Z)
• TailorKV: A Hybrid Framework for Long-Context Inference via Tailored KV Cache Optimization [21.229296254354878]
  Key-Value cache in generative large language models (LLMs) introduces substantial memory overhead.<n>Existing works mitigate this burden by offloading or compressing the KV cache.<n>We propose a hybrid compression method, TailorKV, which seamlessly integrates quantization and offloading.
  arXiv  Detail & Related papers  (2025-05-26T07:00:04Z)
• SCBench: A KV Cache-Centric Analysis of Long-Context Methods [61.025422435235456]
  We introduce SCBench, a benchmark for evaluating long-context methods from a KV cachecentric perspective.<n>We provide an extensive KV cache-centric analysis of eight categories long-context solutions, including Gated Linear RNNs and Mamba-Attention hybrids.<n>Our findings show that sub-O(n) memory methods suffer in multi-turn scenarios, while sparse encoding with O(n) memory and sub-O(n2) pre-filling perform robustly.
  arXiv  Detail & Related papers  (2024-12-13T17:59:52Z)
• ThinK: Thinner Key Cache by Query-Driven Pruning [63.13363917871414]
  Large Language Models (LLMs) have revolutionized the field of natural language processing, achieving unprecedented performance across a variety of applications.<n>This paper focuses on the long-context scenario, addressing the inefficiencies in KV cache memory consumption during inference.<n>We propose ThinK, a novel query-dependent KV cache pruning method designed to minimize attention weight loss while selectively pruning the least significant channels.
  arXiv  Detail & Related papers  (2024-07-30T17:59:08Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.