# Simplified Sparse Attention via Gist Tokens

arXiv: https://arxiv.org/abs/2604.20920 · HTML: https://mlsanity.com/paper/2604.20920/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Placed in KV cache optimization, one of 12 directions in Efficient LLM inference and serving.
- Builds on 37 of its 43 references that are in the map.
- First posted on arXiv 22 Apr 2026.

## Placed in

- Efficient LLM inference and serving / [KV cache optimization](https://mlsanity.com/inference/kv-cache-optimization/)

## Builds on (most cited first, 30 of 37)

- [Mamba: Linear-Time Sequence Modeling with Selective State Spaces](https://mlsanity.com/paper/2312.00752/) (2023)
- [FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness](https://mlsanity.com/paper/2205.14135/) (2022)
- [Longformer: The Long-Document Transformer](https://mlsanity.com/paper/2004.05150/) (2020)
- [TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension](https://mlsanity.com/paper/1705.03551/) (2017)
- [Efficiently Modeling Long Sequences with Structured State Spaces](https://mlsanity.com/paper/2111.00396/) (2021)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention](https://mlsanity.com/paper/2006.16236/) (2020)
- [SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering](https://mlsanity.com/paper/2405.15793/) (2024)
- [Generating Long Sequences with Sparse Transformers](https://mlsanity.com/paper/1904.10509/) (2019)
- [LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding](https://mlsanity.com/paper/2308.14508/) (2023)
- [GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints](https://mlsanity.com/paper/2305.13245/) (2023)
- [Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Steps](https://mlsanity.com/paper/2011.01060/) (2020)
- [The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale](https://mlsanity.com/paper/2406.17557/) (2024)
- [STaR: Bootstrapping Reasoning With Reasoning](https://mlsanity.com/paper/2203.14465/) (2022)
- [H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models](https://mlsanity.com/paper/2306.14048/) (2023)
- [RWKV: Reinventing RNNs for the Transformer Era](https://mlsanity.com/paper/2305.13048/) (2023)
- [DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models](https://mlsanity.com/paper/2512.02556/) (2025)
- [Mem0: Building Production-Ready AI Agents with Scalable Long-Term Memory](https://mlsanity.com/paper/2504.19413/) (2025)
- [Retentive Network: A Successor to Transformer for Large Language Models](https://mlsanity.com/paper/2307.08621/) (2023)
- [Gated Delta Networks: Improving Mamba2 with Delta Rule](https://mlsanity.com/paper/2412.06464/) (2024)
- [Quest: Query-Aware Sparsity for Efficient Long-Context LLM Inference](https://mlsanity.com/paper/2406.10774/) (2024)
- [LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compression](https://mlsanity.com/paper/2310.06839/) (2023)
- [Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention](https://mlsanity.com/paper/2502.11089/) (2025)
- [Learning to Compress Prompts with Gist Tokens](https://mlsanity.com/paper/2304.08467/) (2023)
- [RepoCoder: Repository-Level Code Completion Through Iterative Retrieval and Generation](https://mlsanity.com/paper/2303.12570/) (2023)
- [Adapting Language Models to Compress Contexts](https://mlsanity.com/paper/2305.14788/) (2023)
- [LongLoRA: Efficient Fine-tuning of Long-Context Large Language Models](https://mlsanity.com/paper/2309.12307/) (2023)
- [MoBA: Mixture of Block Attention for Long-Context LLMs](https://mlsanity.com/paper/2502.13189/) (2025)
- [BookSum: A Collection of Datasets for Long-form Narrative Summarization](https://mlsanity.com/paper/2105.08209/) (2021)
- [RedPajama: an Open Dataset for Training Large Language Models](https://mlsanity.com/paper/2411.12372/) (2024)

## Built on by (most cited first, 6 of 6)

- [MARCH: Scaling Recurrent Memory with Content-Routed State Anchors](https://mlsanity.com/paper/2608.12435/) (2026)
- [Safin-1: Safety from Within through Memory-Native State Evolution](https://mlsanity.com/paper/2609.00092/) (2026)
- [Block Sparse Attention with Log-Linear Complexity](https://mlsanity.com/paper/2609.31093/) (2026)
- [The Evolution of Attention in Large Language Models: Mechanisms, Trade-offs, and Emerging Trends](https://mlsanity.com/paper/2609.39661/) (2026)
- [Look Back, Think Ahead: Visual Memory on Demand for Efficient Multimodal Reasoning](https://mlsanity.com/paper/2610.12060/) (2026)
- [Kwai Summary Attention V2 Technical Report](https://mlsanity.com/paper/2610.18654/) (2026)

## Abstract (arXiv)

Sparse attention can reduce the cost of long-context inference, but most variants introduce new architectural components. We introduce Simplified Sparse Attention (SSA), a simpler approach to sparse attention that requires no architectural changes. Concretely, we first perform continued pretraining on sequences interleaved with gist tokens. We optimize the standard next-token loss as usual, but the gist tokens use an attention mask to restrict what parts of the context the language model can attend to; this teaches the model to pack each chunk's important information into the gist tokens. At inference time, SSA scores chunks via attention between the current query and the small set of gist tokens, selectively unfolding the top-k chunks by reintroducing their corresponding raw tokens. Since the query is scored only against the gist tokens, we avoid the memory-bandwidth cost associated with naive scoring against the full KV cache, without requiring the auxiliary KV cache approach used by sparse attention methods. On LongBench, SSA consistently outperforms compression and inference-time sparse-attention baselines under the same compression ratio. More strikingly, in retrieval-augmented generation, SSA can even outperform full attention after continued pretraining by over 5.7 points. We attribute this to the ability of SSA's selective unfolding, which concentrates attention on the query-relevant chunks and effectively filters out noise. SSA further extends to a hierarchical gist-of-gist variant (H-SSA) that achieves log-linear decoding complexity while maintaining or improving accuracy at high compression ratios up to 32x. The code is available at https://github.com/yuzhenmao/simplified-sparse-attention/.