- Rating
- 2
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
To address the shortcomings of existing hypergraph-based RAGs in modeling multi-level relationships, this paper proposes an Asymmetric Indexing-based Hierarchical Hyper-RAG. This method first constructs a three-layer nested hypergraph, then uses Asymmetric Indexing to fuse high-level context with low-level factual information, and finally retrieves and expands relevant evidence progressively through Query Alignment and Coarse-to-Fine Retrieval. Extensive experiments demonstrate the improved answer quality and high online retrieval efficiency of this proposed method.
Critical strengths
-
This paper constructs a three-layer nested hypergraph and integrates Asymmetric Indexing, Query Alignment, and Coarse-to-Fine Evidence Retrieval into a unified framework, enabling retrieval to simultaneously leverage high-level semantic context and fine-grained factual information.
-
The paper evaluates across eight document corpora, five LLM backbones, and three different evidence depths, demonstrating consistent improvements over the evaluated baselines.
Critical weaknesses
- Asymmetric Indexing follows the existing principles of Hierarchical Message Passing [1][2], and its information fusion mechanism is mainly a simple similarity-weighted aggregation. Furthermore, the ablation experiment only compared Asymmetric Indexing with the case without cross-layer propagation, without demonstrating its advantages over other information propagation strategies.
- Cue-activated hierarchical evidence recall mainly includes Topic Activation, Coarse-to-Fine Retrieval, Multi-granularity Sentence Reranking, and Entity-mediated Evidence Expansion. The core ideas of these operations have appeared in previous studies [3][4][5]. However, they have not been compared with .
- Page 2, Figure 1: “Sentences Representaton” is misspelled; it should be “Sentence Representation”.
- L272: “We develop a evidence recall” has an incorrect article; it should be “an”.
- L294: “We uses these signals” has an agreement error.
- L436, L453, L510: “Coarse-to-fine evidence retrieval” is misspelled.
[1] Teng, Zhiyang, and Yue Zhang. "Head-lexicalized bidirectional tree lstms." Transactions of the Association for Computational Linguistics 5 (2017): 163-177.
[2] Zhong, Zhiqiang, Cheng-Te Li, and Jun Pang. "Hierarchical message-passing graph neural networks: Z. Zhong et al." Data Mining and Knowledge Discovery 37.1 (2023): 381-408.
[3] Tao, Wenyu, et al. "Treerag: Unleashing the power of hierarchical storage for enhanced knowledge retrieval in long documents." Findings of the Association for Computational Linguistics: ACL 2025. 2025.
[4] Huang, Haoyu, et al. "Retrieval-Augmented Generation with Hierarchical Knowledge." EMNLP (Findings). 2025.
[5] Zhu, Xiangrong, et al. "Knowledge graph-guided retrieval augmented generation." Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.
Questions for authors
- Could you further elaborate on the key differences between Asymmetric Indexing, Cue-activated hierarchical evidence recall, and existing methods, as well as the innovations of this paper?
- "w/o Coarse-to-Fine Evidence Retrieval". Could you conduct a more detailed ablation experiment by removing the following three modules respectively: Topic-scoped Chunk Recall, Sentence Reranking, Entity-mediated Evidence Expansion?
- Could you add baselines to compare with other information propagation and retrieval methods, further demonstrating the effectiveness of Asymmetric Indexing and Cue-activated hierarchical evidence recall?
3 comments
Here's our reply to the reviewers' question 2,3
2. Regarding the ablation experiments, what we did was indeed lacking, so in response to the reviewers' questions, we redesigned the corresponding ablation experiments. The experimental results are as follows:
| Ablated design | Comp. | Div. | Emp. | Log. | Read. | Rel. | Avg. | Δ (%) |
|---|---|---|---|---|---|---|---|---|
| Full Structure (AI-H²RAG) | 89.13 ± 0.87 | 79.44 ± 0.77 | 81.57 ± 0.85 | 85.67 ± 0.35 | 88.12 ± 0.30 | 92.53 ± 0.47 | 86.08 ± 0.39 | 0.0% |
| w/o Hierarchical Hypergraph Memory | 83.80 ± 1.37 | 71.90 ± 1.14 | 74.70 ± 0.69 | 80.60 ± 0.47 | 81.10 ± 0.31 | 85.61 ± 0.45 | 79.62 ± 0.57 | -7.5% |
| Ordinary Graph | 87.60 ± 0.79 | 74.80 ± 0.92 | 75.94 ± 0.89 | 82.70 ± 0.33 | 81.70 ± 0.22 | 91.10 ± 0.28 | 82.31 ± 0.43 | -4.4% |
| w/o Asymmetric Hypergraph Network | 85.00 ± 1.23 | 71.70 ± 1.33 | 75.70 ± 0.71 | 82.50 ± 0.27 | 80.20 ± 0.37 | 91.40 ± 0.45 | 81.08 ± 0.67 | -5.8% |
| Asymmetric Raw Aggregation | 85.80 ± 1.33 | 75.70 ± 0.87 | 74.30 ± 0.77 | 81.60 ± 0.42 | 80.42 ± 0.29 | 90.54 ± 0.78 | 81.39 ± 0.84 | -5.4% |
| Symmetric Final Aggregation | 85.60 ± 1.14 | 75.20 ± 1.56 | 77.26 ± 0.77 | 83.04 ± 0.68 | 85.10 ± 0.35 | 91.20 ± 0.54 | 82.90 ± 0.52 | -3.7% |
| w/o Query Alignment | 81.20 ± 0.87 | 71.40 ± 1.45 | 72.50 ± 0.90 | 80.90 ± 0.70 | 81.72 ± 0.60 | 91.20 ± 0.31 | 79.82 ± 0.64 | -7.3% |
| w/o Topic Routing | 85.50 ± 0.95 | 73.20 ± 0.63 | 75.82 ± 0.99 | 82.84 ± 0.57 | 81.34 ± 0.35 | 90.34 ± 0.31 | 81.51 ± 0.46 | -5.3% |
| w/o Chunk-level Coarse Retrieval | 83.80 ± 1.15 | 71.70 ± 1.36 | 74.96 ± 0.73 | 81.70 ± 0.84 | 81.20 ± 0.55 | 89.50 ± 0.75 | 80.48 ± 0.72 | -6.5% |
| w/o Sentence Reranking | 84.50 ± 1.67 | 74.30 ± 0.91 | 75.16 ± 0.71 | 81.94 ± 0.45 | 80.90 ± 0.47 | 90.60 ± 0.39 | 81.23 ± 0.44 | -5.6% |
| w/o Multi-hop Retrieval | 85.20 ± 0.84 | 73.90 ± 1.27 | 75.30 ± 0.54 | 83.00 ± 0.70 | 84.26 ± 0.43 | 90.80 ± 0.37 | 82.08 ± 0.45 | -4.6% |
This time, we conducted ablation experiments on 30 questions in each of stages 1, 2, and 3 for different groups. First, on the re-layered graph structure, including w/o Hierarchical Hypergraph Memory and Ordinary Graph. Then, on asymmetric retrieval, including w/o Asymmetric Hypergraph Network, Asymmetric Raw Aggregation, and Symmetric Final Aggregation. Finally, on the query, including w/o Query Alignment, w/o Topic Routing, w/o Chunk-level Coarse Retrieval, w/o Sentence Reranking, and w/o Multi-hop Retrieval.
3. This time we compared different levels of RAG on the commonly used public datasets HotpotQA, 2WikiMultiHopQA, and MuSiQue, and the experimental results are as follows:
| Method | hotpotqa F1 (%) | hotpotqa EM (%) | hotpotqa Recall@5 (%) | 2wikimultihopqa F1 (%) | 2wikimultihopqa EM (%) | 2wikimultihopqa Recall@5 (%) | musique F1 (%) | musique EM (%) | musique Recall@5 (%) |
|---|---|---|---|---|---|---|---|---|---|
| Ours | 72.76 | 60.80 | 85.40 | 60.00 | 54.70 | 74.21 | 45.04 | 35.90 | 54.20 |
| HyperGraphRAG | 70.30 | 58.50 | 77.45 | 54.49 | 49.40 | 62.28 | 38.58 | 28.50 | 45.67 |
| HiRAG | 58.58 | 49.10 | 81.40 | 47.49 | 41.70 | 72.00 | 33.33 | 26.00 | 49.20 |
| HGRAG | 72.10 | 60.60 | 86.65 | 59.26 | 53.80 | 73.25 | 40.40 | 32.40 | 56.39 |
| RAPTOR | 67.12 | 55.90 | 83.70 | 53.50 | 48.50 | 70.95 | 33.23 | 26.10 | 51.91 |
It can be seen that our method achieved good results in both F1 score and EM score, while the recall rate on HotpotQA and MuSiQue is slightly lower than HGRAG.
Thank the authors for the responses. Some of my concerns have been addressed. After reading other peer reviews and the responses, I will maintain the original score.
Here's our reply to the reviewers' question 1
Thank you for your thoughtful review. We respond to each point below.
1. This article first introduces the concept of a hypergraph in the graph construction to build the corresponding hypergraph knowledge base, which is a key difference from RARTOR[1], HiRAG[2], and the Hierarchical Lexical Graph[3]. The reason for choosing a hypergraph is that, as shown by Hyper-rag[4] and hypergraphrag[5], a hypergraph can cover higher-order relationships and provide richer semantics. Next, regarding graph construction, unlike PAPTOR and HiRAG, which use GMM to construct upper-level graphs, and different from the Hierarchical Lexical Graph, which constructs upper-level statements by summarizing triples, we build from topic to chunk, sentence, and entity top-down based on the corresponding ownership relations. In this process, the same chunk may belong to multiple topics, the same sentence may belong to multiple paragraphs, and the same entity may belong to multiple sentences. To handle this, hypergraphs are undoubtedly the most suitable choice. Then, in constructing our retrieval network, regarding indexing, we consider that understanding a passage requires a combination of high-level context and concrete facts. Entities also need to be recognized under relevant semantics for better understanding. Therefore, the corresponding asymmetric information propagation is introduced. It is asymmetric because the vectors transmitted upward and downward are different versions. Parent nodes transmit the final representations fused with context downward, and child nodes transmit their original representations upward. For example, in the case of a “document chunk—sentence”: when a sentence → document chunk, the document chunk needs to know the local evidence it contains, so it aggregates the original vectors of each sentence. These vectors provide the content of the sentences themselves. When a document chunk → sentence, a single sentence may omit the object or describe only a local operation, so it receives the final vector of the document chunk. This chunk vector has already integrated the topic, the content of the chunk itself, and the original sentences inside the chunk, providing the sentence with a broader context. For instance, the original sentence reads, 'Adjust the right-side nut, and tighten after removing the gap.' From an upward perspective, this sentence contributes concrete operation steps to the document chunk. From a downward perspective, the document chunk informs the sentence that this nut belongs to the 'ball screw assembly' and the discussion concerns the 'axial gap.' For the same relation, the information transmitted on both sides serves different purposes. This asymmetric design allows local evidence to borrow from the upper-level context while also avoiding the borrowed context from flowing back as local evidence. Next, regarding the query, we did not directly use the original vectors for the corresponding matching. We also updated the query vectors through corresponding information propagation. At the retrieval level, indeed, our method is conceptually similar to RAPTOR, HiRAG, and Hierarchical Lexical Graph, which perform layer-by-layer retrieval starting from the topic or community level. However, this is a natural idea for constructing a hierarchical hypergraph structure. At the same time, the scoring we use for sentence re-ranking aligns with our design and is not the same as the currently existing scoring algorithms.
[1] Feng, Yifan, et al. "Hyper-RAG: combating LLM hallucinations using hypergraph-driven retrieval-augmented generation." Nature Communications 17, 5778 (2026).
[2] Luo, Haoran, et al. "HyperGraphRAG: Retrieval-Augmented Generation via Hypergraph-Structured Knowledge Representation." Advances in Neural Information Processing Systems 38 (2025): 152206-152234.
[3] Sarthi, Parth, et al. "RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval." International Conference on Learning Representations. 2024.
[4] Huang, Haoyu, et al. "Retrieval-Augmented Generation with Hierarchical Knowledge." Findings of the Association for Computational Linguistics: EMNLP 2025. 2025. 6044-6060.
[5] Ghassel, Abdellah, et al. "Hierarchical Lexical Graph for Enhanced Multi-Hop Retrieval." Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining. Vol. 2. 2025. 4457-4466.