IterCOMP: Reasoning-aware Adaptive Prompt Compression for Multi-hop Question Answering
On the map: Retrieval-augmented generation → Knowledge-graph and structured-knowledge RAG
also in 1 more
Abstract (arXiv)
Multi-hop question answering requires complex reasoning across multiple evidence segments, which often overwhelms retrieval-augmented generation systems with lengthy and noisy contexts, thereby undermining both efficiency and accuracy. While existing prompt compression methods attempt to address this issue, they are typically designed for single-turn queries and fail to capture interdependent reasoning steps. We propose IterCOMP, a unified, training-free prompt compression framework that incorporates multi-hop reasoning within an iterative compression loop. IterCOMP decomposes documents into evidence segments, evaluates question answerability, and generates targeted follow-up questions to iteratively integrate essential evidence, producing a compact, reasoning-oriented prompt. Experiments on MusiQue, 2WikiMultiHopQA, and HotpotQA demonstrate that IterCOMP achieves substantial improvements in Exact Match and F1 scores while reducing the token budget, outperforming existing baselines and exhibiting robustness as reasoning complexity increases.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Knowledge-graph and structured-knowledge RAG, one of 15 directions in Retrieval-augmented generation. Builds on 34 of its 45 references that are in the map. First posted on arXiv 12 Jul 2026.
Builds on
34 of 45 references.
Closest ancestors · 4
Same direction, newest first
- 2025Mitigating Lost-in-Retrieval Problems in Retrieval Augmented Multi-Hop Question Answering20 Feb
- 2024SiReRAG: Indexing Similar and Related Information for Multihop Reasoning9 Dec
- 2024BRIEF: Bridging Retrieval and Inference for Multi-hop Reasoning via Compression20 Oct
- 2023Enhancing Retrieval-Augmented Large Language Models with Iterative Retrieval-Generation Synergy24 May
Other references · 30
- 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,297
- 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,695
- 2018HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answeringcited by 3,754
- 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,455
- 2023Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflectioncited by 1,839
- 2020Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Stepscited by 1,282
- 2022Measuring and Narrowing the Compositionality Gap in Language Modelscited by 1,075
- 2024M3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillationcited by 1,073
- 2022Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questionscited by 923
- 2023Large Language Models Can Be Easily Distracted by Irrelevant Contextcited by 841
- 2021Did Aristotle Use a Laptop? A Question Answering Benchmark with Implicit Reasoning Strategiescited by 743
- 2023LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compressioncited by 453
18 more
- 2023Learning to Compress Prompts with Gist Tokenscited by 390
- 2023Adapting Language Models to Compress Contextscited by 356
- 2024LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compressioncited by 348
- 2023FreshLLMs: Refreshing Large Language Models with Search Engine Augmentationcited by 318
- 2024MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queriescited by 257
- 2024xRAG: Extreme Context Compression for Retrieval-augmented Generation with One Tokencited by 125
- 2024CompAct: Compressing Retrieved Documents Actively for Question Answeringcited by 64
- 2024Learning to Compress Prompt in Natural Language Formatscited by 55
- 2024Improving Retrieval-Augmented Generation in Medicine with Iterative Follow-up Questionscited by 54
- 2024Prompt Compression with Context-Aware Sentence Encoding for Fast and Improved LLM Inferencecited by 53
- 2023Discrete Prompt Compression with Reinforcement Learningcited by 42
- 2024MoreHopQA: More Than Multi-hop Reasoningcited by 37
- 2023The ART of LLM Refinement: Ask, Refine, and Trustcited by 29
- 2024Retaining Key Information under High Compression Ratios: Query-Guided Compressor for LLMscited by 24
- 2024LLMs Know What They Need: Leveraging a Missing Information Guided Framework to Empower Retrieval-Augmented Generationcited by 17
- 2025More Documents, Same Length: Isolating the Challenge of Multiple Documents in RAGcited by 16
- 2024From Reading to Compressing: Exploring the Multi-document Reader for Prompt Compressioncited by 7
- 2024Using LLMs to Investigate Correlations of Conversational Follow-up Queries with User Satisfactioncited by 6
Built on by
Citing papers per year: 2026: 1.
1 paper.
Similar, not cited
Close in wording, no citation either way
- 2026The Reasoning Bottleneck in Graph-RAG: Structured Prompting and Context Compression for Multi-Hop QA
- 2026PAR2-RAG: Planned Active Retrieval and Reasoning for Multi-Hop Question Answering
- 2026CIRAG: Construction-Integration Retrieval and Adaptive Generation for Multi-hop Question Answering
- 2026Retrieval--Reasoning Processes for Multi-hop Question Answering: A Four-Axis Design Framework and Empirical Trends
- 2025Memory-Aware and Uncertainty-Guided Retrieval for Multi-Hop Question Answering
- 2025StepChain GraphRAG: Reasoning Over Knowledge Graphs for Multi-Hop Question Answering
- 2026MEGRAG: Multi-Granular Evidence Graphs for Answer-Aware Multi-Hop RAG
- 2024Tree of Reviews: A Tree-based Dynamic Iterative Retrieval Framework for Multi-hop Question Answering