mlsanityData to 7 Oct 2026

Prompt and context compression

132 papers0 this weekHigh confidence in this grouping: the papers form one clear topicHigh confidence in this grouping: the papers form one clear topic · About

Keywords: compression, context, prompt, prompt compression, context compression

Key facts. Prompt and context compression: 132 papers on arXiv since 2022 (data through 7 Oct 2026). 72 of its 132 mapped papers (55%) cite “LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compression” (2024, arXiv 2403.12968). 67 of its 132 mapped papers (51%) cite “LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compression” (2023, arXiv 2310.06839). 3 new papers 24 Sep–7 Oct.

What’s new

24 Sep – 7 Oct 2026

3 new since 24 Sep

  1. Aperture: Merge-Consistent Rotary States for Compressed Tokens also fits: Audio and omni-modal token compression
  2. PC-SubMax: Efficient Prompt Compression via Regularized Submodular Maximization
  3. HasMem: Hard-Origin Adaptively Softened Memory for Long-Term LLM Agents also fits: Efficient reasoning and agent inference

Foundations

Most cited by papers here

  1. LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compression2024cited by 72 here
  2. LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compression2023cited by 67 here
  3. In-context Autoencoder for Context Compression in a Large Language Model2023cited by 30 here
  4. Prompt Compression and Contrastive Conditioning for Controllability and Toxicity Reduction in Language Models2022cited by 27 here
  5. xRAG: Extreme Context Compression for Retrieval-augmented Generation with One Token2024cited by 27 here

Lineage

The papers this direction builds on, oldest first.

Each paper on the line cites the one before.

  1. 2022Prompt Compression and Contrastive Conditioning for Controllability and Toxicity Reduction in Language Modelscited by 27 here
  2. 2023In-context Autoencoder for Context Compression in a Large Language Modelcited by 30 here
  3. 2023LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compressioncited by 67 here
  4. 2024LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compressioncited by 72 here
  5. 2024xRAG: Extreme Context Compression for Retrieval-augmented Generation with One Tokencited by 27 here
  6. 2024Context Embeddings for Efficient Answer Generation in RAGcited by 7 here
  7. 2024500xCompressor: Generalized Prompt Compression for Large Language Modelscited by 16 here
  8. 2025GMSA: Enhancing Context Compression via Group Merging and Layer Semantic Alignmentcited by 8 here
16 more it builds on, not in this chain
  1. 2023H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Modelscited by 8 here
  2. 2023Discrete Prompt Compression with Reinforcement Learningcited by 22 here
  3. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 20 here
  4. 2023Model Tells You What to Discard: Adaptive KV Cache Compression for LLMscited by 7 here
  5. 2023TCRA-LLM: Token Compression Retrieval Augmented Large Language Model for Inference Cost Reductioncited by 6 here
  6. 2024Long Context Compression with Activation Beaconcited by 15 here
  7. 2024LLoCO: Learning Long Contexts Offlinecited by 6 here
  8. 2024SnapKV: LLM Knows What You are Looking for Before Generationcited by 18 here
  9. 2024Retaining Key Information under High Compression Ratios: Query-Guided Compressor for LLMscited by 12 here
  10. 2024Characterizing Prompt Compression Methods for Long Context Inferencecited by 6 here
  11. 2024Fundamental Limits of Prompt Compression: A Rate-Distortion Framework for Black-Box Language Modelscited by 6 here
  12. 2024Prompt Compression with Context-Aware Sentence Encoding for Fast and Improved LLM Inferencecited by 21 here
  13. 2024TACO-RL: Task Aware Prompt Compression Optimization with Reinforcement Learningcited by 11 here
  14. 2024Perception Compressor: A Training-Free Prompt Compression Framework in Long Context Scenarioscited by 10 here
  15. 2024Prompt Compression for Large Language Models: A Surveycited by 14 here
  16. 2024EXIT: Context-Aware Extractive Compression for Enhancing Retrieval-Augmented Generationcited by 10 here

The 24 most cited of 156.

Browse

30 of 132, ranked 6–35

  1. 6Discrete Prompt Compression with Reinforcement Learning202322
  2. 7Prompt Compression with Context-Aware Sentence Encoding for Fast and Improved LLM Inference202421
  3. 8500xCompressor: Generalized Prompt Compression for Large Language Models202416
  4. 9Long Context Compression with Activation Beacon202415
  5. 10Prompt Compression for Large Language Models: A Survey202414
  6. 11Retaining Key Information under High Compression Ratios: Query-Guided Compressor for LLMs202412
  7. 12TACO-RL: Task Aware Prompt Compression Optimization with Reinforcement Learning202411
  8. 13Perception Compressor: A Training-Free Prompt Compression Framework in Long Context Scenarios202410
  9. 14EXIT: Context-Aware Extractive Compression for Enhancing Retrieval-Augmented Generation202410
  10. 15GMSA: Enhancing Context Compression via Group Merging and Layer Semantic Alignment20258
All 30
  1. 16Context Embeddings for Efficient Answer Generation in RAG20247
  2. 17Autoencoding-Free Context Compression for LLMs via Contextual Semantic Anchors20256
  3. 18LLoCO: Learning Long Contexts Offline20246
  4. 19Characterizing Prompt Compression Methods for Long Context Inference20246
  5. 20Fundamental Limits of Prompt Compression: A Rate-Distortion Framework for Black-Box Language Models20246
  6. 21A Silver Bullet or a Compromise for Full Attention? A Comprehensive Study of Gist Token-based Context Compression20246
  7. 22TCRA-LLM: Token Compression Retrieval Augmented Large Language Model for Inference Cost Reduction20236
  8. 23Prompt-SAW: Leveraging Relation-Aware Graphs for Textual Prompt Compression20245
  9. 24Position IDs Matter: An Enhanced Position Layout for Efficient Context Compression in Large Language Models20245
  10. 25From Reading to Compressing: Exploring the Multi-document Reader for Prompt Compression20245
  11. 26Selection-p: Self-Supervised Task-Agnostic Prompt Compression for Faithfulness and Transferability20245
  12. 27EFPC: Towards Efficient and Flexible Prompt Compression20254
  13. 28Understanding and Improving Information Preservation in Prompt Compression for LLMs20254
  14. 29DAC: A Dynamic Attention-aware Approach for Task-Agnostic Prompt Compression20254
  15. 30Adapting LLMs for Efficient Context Processing through Soft Prompt Compression20244
  16. 31Concise and Precise Context Compression for Tool-Using Language Models20244
  17. 32AdaComp: Extractive Context Compression with Adaptive Predictor for Retrieval-Augmented Large Language Models20244
  18. 33Parse Trees Guided LLM Prompt Compression20244
  19. 34Style-Compress: An LLM-Based Prompt Compression Framework Considering Task-Specific Styles20244
  20. 35COMI: Coarse-to-fine Context Compression via Marginal Information Gain20263