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

Page: https://stenobird.com/podcast/daily-paper-cast-7079649/flashmemory-deepseek-v4-lightning-index-ultra-long-context-via-lookahead-sparse-attention
Text version: https://stenobird.com/podcast/daily-paper-cast-7079649/flashmemory-deepseek-v4-lightning-index-ultra-long-context-via-lookahead-sparse-attention.md
Podcast: [Daily Paper Cast](https://stenobird.com/podcast/daily-paper-cast-7079649)
Published: 2026-06-10T04:34:31+00:00
Episode link: https://share.transistor.fm/s/f0cfc106
Audio file: https://media.transistor.fm/f0cfc106/f66b02e9.mp3
Processing state: not_requested
JSON: https://stenobird.com/v1/public/podcasts/daily-paper-cast-7079649/episodes/flashmemory-deepseek-v4-lightning-index-ultra-long-context-via-lookahead-sparse-attention
Duration seconds: 1307

## Resource

🤗 Upvotes: 46 | cs.LG, cs.AI Authors: Yan Wang, Qifan Zhang, Jiachen Yu, Tian Liang, Dongyang Ma, Xiang Hu, Zibo Lin, Chunyang Li, Zhichao Wang, Miao Peng, Nuo Chen, Jia Li, Yujiu Yang, Haitao Mi, Dong Yu Title: FlashMemory-DeepSeek-V4: Lightning Index Ultra-Long Context via Lookahead Sparse Attention Arxiv: http://arxiv.org/abs/2606.09079v2 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.

## Actions

- request_transcript: `POST https://stenobird.com/v1/public/podcasts/daily-paper-cast-7079649/episodes/flashmemory-deepseek-v4-lightning-index-ultra-long-context-via-lookahead-sparse-attention/transcription-requests` — Idempotently request low-priority transcript generation for this episode.
- read_markdown: `GET https://stenobird.com/podcast/daily-paper-cast-7079649/flashmemory-deepseek-v4-lightning-index-ultra-long-context-via-lookahead-sparse-attention.md` — Read the agent-friendly Markdown representation of this episode resource.

A page view does not enqueue transcription. Agents should invoke `request_transcript` explicitly when they need this episode processed.

## Transcript

Full transcripts are not published on public pages unless there is a clear rights basis.