Research map / Architectures and efficiency / Attention and transformers
Autoencoding-Free Context Compression for LLMs via Contextual Semantic Anchors
Xin Liu, Runsong Zhao, Pengcheng Huang, Xinyu Liu, Junyi Xiao, Chunyang Xiao, Tong Xiao, Shengxiang Gao, Zhengtao Yu, JingBo Zhu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add bidirectional attention in the compressor encoder (ablation) | SAC · MRQA out-of-domain (BioASQ, TextbookQA avg) | F1: 45.11 -> 48.45 | Improved |
| add bidirectional attention in the compressor encoder (ablation) | SAC · MRQA in-domain (TriviaQA, HotpotQA avg) | F1: 63.62 -> 66.24 | Improved |
| add learnable anchor embedding to selected tokens (ablation) | SAC · MRQA in-domain (TriviaQA, HotpotQA avg) | F1: 64.58 -> 66.24 | Improved |
| add learnable anchor embedding to selected tokens (ablation) | SAC · MRQA out-of-domain (BioASQ, TextbookQA avg) | F1: 47.65 -> 48.45 | Improved |
| select anchors uniformly (middle of chunks) instead of randomly (ablation) | SAC · MRQA in-domain average, 5x | F1: 59.54 -> 63.63 | Improved |
| select anchors with LLMLingua-2 instead of uniform (ablation) | SAC · MRQA in-domain average, 5x | F1: 63.63 -> 63.26 | No clear effect |
| add autoencoding objective to pretraining (with LM) (ablation) | SAC · MRQA in-domain average, 5x | F1: 63.63 -> 62.04 | Hurt |
| pretrain SAC with AE only instead of LM only (ablation) | SAC · MRQA in-domain average, 15x | F1: 54.95 -> 49.93 | Hurt |
| add autoencoding objective to LM pretraining (ablation) | 500xCompressor · MRQA in-domain average, 5x/15x/51x | F1: 53.23 -> 55.26 (5x); 49.76 -> 49.35 (15x); 44.46 -> 43.19 (51x) | Mixed |
| replace EPL with SAC (anchors from context tokens) | Llama-3.2-3B compressor · MRQA in-domain average, 15x | F1: 68.60 -> 69.93 | Improved |
| replace EPL with SAC (anchors from context tokens) | Llama-3.1-8B compressor · MRQA out-of-domain average, 15x | F1: 59.88 -> 65.23 | Improved |
| replace EPL with SAC | Llama-3.2-1B compressor · MRQA in-domain average, 51x | F1: 43.22 -> 46.37 | Improved |
Closest papers by meaning
- In-context Autoencoder for Context Compression in a Large Language Model (ICLR 2024 · similarity 0.91)
- ChunkKV: Semantic-Preserving KV Cache Compression for Efficient Long-Context LLM Inference (NeurIPS 2025 · similarity 0.90)
- Compressing LLMs: The Truth is Rarely Pure and Never Simple (ICLR 2024 · similarity 0.89)
- Prompt Compression with Context-Aware Sentence Encoding for Fast and Improved LLM Inference (AAAI 2025 · similarity 0.89)
- Compressed Context Memory for Online Language Model Interaction (ICLR 2024 · similarity 0.88)
- Leveraging Attention to Effectively Compress Prompts for Long-Context LLMs (AAAI 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
