# Discovering the Gems in Early Layers: Accelerating Long-Context LLMs with 1000x Input Token Reduction

arXiv: https://arxiv.org/abs/2409.17422 · HTML: https://mlsanity.com/paper/2409.17422/

## Key facts

- Cited by 39 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 22 of its 30 references that are in the map.
- First posted on arXiv 25 Sep 2024.

## Placed in

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

## Builds on (most cited first, 22 of 22)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness](https://mlsanity.com/paper/2205.14135/) (2022)
- [FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning](https://mlsanity.com/paper/2307.08691/) (2023)
- [Sparks of Artificial General Intelligence: Early experiments with GPT-4](https://mlsanity.com/paper/2303.12712/) (2023)
- [Emergent Abilities of Large Language Models](https://mlsanity.com/paper/2206.07682/) (2022)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone](https://mlsanity.com/paper/2404.14219/) (2024)
- [LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding](https://mlsanity.com/paper/2308.14508/) (2023)
- [SnapKV: LLM Knows What You are Looking for Before Generation](https://mlsanity.com/paper/2404.14469/) (2024)
- [H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models](https://mlsanity.com/paper/2306.14048/) (2023)
- [FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision](https://mlsanity.com/paper/2407.08608/) (2024)
- [Model Tells You What to Discard: Adaptive KV Cache Compression for LLMs](https://mlsanity.com/paper/2310.01801/) (2023)
- [LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compression](https://mlsanity.com/paper/2310.06839/) (2023)
- [MInference 1.0: Accelerating Pre-filling for Long-Context LLMs via Dynamic Sparse Attention](https://mlsanity.com/paper/2407.02490/) (2024)
- [LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compression](https://mlsanity.com/paper/2403.12968/) (2024)
- [Compressing Context to Enhance Inference Efficiency of Large Language Models](https://mlsanity.com/paper/2310.06201/) (2023)
- [LongWriter: Unleashing 10,000+ Word Generation from Long Context LLMs](https://mlsanity.com/paper/2408.07055/) (2024)
- [LongRAG: Enhancing Retrieval-Augmented Generation with Long-context LLMs](https://mlsanity.com/paper/2406.15319/) (2024)
- [LOOK-M: Look-Once Optimization in KV Cache for Efficient Multimodal Long-Context Inference](https://mlsanity.com/paper/2406.18139/) (2024)
- [ThinK: Thinner Key Cache by Query-Driven Pruning](https://mlsanity.com/paper/2407.21018/) (2024)

## Built on by (most cited first, 30 of 39)

- [FaithEval: Can Your Language Model Stay Faithful to Context, Even If "The Moon is Made of Marshmallows"](https://mlsanity.com/paper/2410.03727/) (2024)
- [TidalDecode: Fast and Accurate LLM Decoding with Position Persistent Sparse Attention](https://mlsanity.com/paper/2410.05076/) (2024)
- [MrT5: Dynamic Token Merging for Efficient Byte-level Language Models](https://mlsanity.com/paper/2410.20771/) (2024)
- [Circuit Complexity Bounds for RoPE-based Transformer Architecture](https://mlsanity.com/paper/2411.07602/) (2024)
- [Bypassing the Exponential Dependency: Looped Transformers Efficiently Learn In-context by Multi-step Gradient Descent](https://mlsanity.com/paper/2410.11268/) (2024)
- [The Computational Limits of State-Space Models and Mamba via the Lens of Circuit Complexity](https://mlsanity.com/paper/2412.06148/) (2024)
- [Looped ReLU MLPs May Be All You Need as Practical Programmable Computers](https://mlsanity.com/paper/2410.09375/) (2024)
- [HSR-Enhanced Sparse Attention Acceleration](https://mlsanity.com/paper/2410.10165/) (2024)
- [FastKV: Decoupling of Context Reduction and KV Cache Compression for Prefill-Decoding Acceleration](https://mlsanity.com/paper/2502.01068/) (2025)
- [Speculative Prefill: Turbocharging TTFT with Lightweight and Training-Free Token Importance Estimation](https://mlsanity.com/paper/2502.02789/) (2025)
- [Theoretical Constraints on the Expressive Power of mathsfRoPE-based Tensor Attention Transformers](https://mlsanity.com/paper/2412.18040/) (2024)
- [Efficient Prompt Compression with Evaluator Heads for Long-Context Transformer Inference](https://mlsanity.com/paper/2501.12959/) (2025)
- [Video Latent Flow Matching: Optimal Polynomial Projections for Video Interpolation and Extrapolation](https://mlsanity.com/paper/2502.00500/) (2025)
- [Theoretical Guarantees for High Order Trajectory Refinement in Generative Flows](https://mlsanity.com/paper/2503.09069/) (2025)
- [Recurrent Attention-based Token Selection for Efficient Streaming Video-LLMs](https://mlsanity.com/paper/2510.17364/) (2025)
- [Rethinking Key-Value Cache Compression Techniques for Large Language Model Serving](https://mlsanity.com/paper/2503.24000/) (2025)
- [CentroidKV: Efficient Long-Context LLM Inference via KV Cache Clustering](https://mlsanity.com/paper/2506.11418/) (2025)
- [Lossless KV Cache Compression to 2%](https://mlsanity.com/paper/2410.15252/) (2024)
- [Progressive Sparse Attention: Algorithm and System Co-design for Efficient Attention in LLM Serving](https://mlsanity.com/paper/2503.00392/) (2025)
- [FreqKV: Key-Value Compression in Frequency Domain for Context Window Extension](https://mlsanity.com/paper/2505.00570/) (2025)
- [SparseServe: Unlocking Parallelism for Dynamic Sparse Attention in Long-Context LLM Serving](https://mlsanity.com/paper/2509.24626/) (2025)
- [Advancing the Understanding of Fixed Point Iterations in Deep Neural Networks: A Detailed Analytical Study](https://mlsanity.com/paper/2410.11279/) (2024)
- [Unlocking the Theory Behind Scaling 1-Bit Neural Networks](https://mlsanity.com/paper/2411.01663/) (2024)
- [Neural Algorithmic Reasoning for Hypergraphs with Looped Transformers](https://mlsanity.com/paper/2501.10688/) (2025)
- [HAMburger: Accelerating LLM Inference via Token Smashing](https://mlsanity.com/paper/2505.20438/) (2025)
- [When Can We Solve the Weighted Low Rank Approximation Problem in Truly Subquadratic Time?](https://mlsanity.com/paper/2502.16912/) (2025)
- [SALE : Low-bit Estimation for Efficient Sparse Attention in Long-context LLM Prefilling](https://mlsanity.com/paper/2505.24179/) (2025)
- [Adaptive Layer Selection for Layer-Wise Token Pruning in LLM Inference](https://mlsanity.com/paper/2601.07667/) (2026)
- [StructKV: Preserving the Structural Skeleton for Scalable Long-Context Inference](https://mlsanity.com/paper/2604.06746/) (2026)
- [Theoretical Foundation of Flow-Based Time Series Generation: Provable Approximation, Generalization, and Efficiency](https://mlsanity.com/paper/2503.14076/) (2025)

## Abstract (arXiv)

Large Language Models (LLMs) have demonstrated remarkable capabilities in handling long context inputs, but this comes at the cost of increased computational resources and latency. Our research introduces a novel approach for the long context bottleneck to accelerate LLM inference and reduce GPU memory consumption. Our research demonstrates that LLMs can identify relevant tokens in the early layers before generating answers to a query. Leveraging this insight, we propose an algorithm that uses early layers of an LLM as filters to select and compress input tokens, significantly reducing the context length for subsequent processing. Our method, GemFilter, demonstrates substantial improvements in both speed and memory efficiency compared to existing techniques, such as standard attention and SnapKV/H2O. Notably, it achieves a 2.4× speedup and 30% reduction in GPU memory usage compared to SOTA methods. Evaluation on the Needle in a Haystack task shows that GemFilter significantly outperforms standard attention, SnapKV and demonstrates comparable performance on the LongBench challenge. GemFilter is simple, training-free, and broadly applicable across different LLMs. Crucially, it provides interpretability by allowing humans to inspect the selected input sequence. These findings not only offer practical benefits for LLM deployment, but also enhance our understanding of LLM internal mechanisms, paving the way for further optimizations in LLM design and inference. Our code is available at urlhttps://github.com/SalesforceAIResearch/GemFilter.