arXiv Computation and Language October 23, 2026 By Aaditya Chauhan, Arun Agarwal

The Fusion Tax: A Diagnostic Analysis of When Hybrid Retrieval Hurts

Read the original on arXiv Computation and Language →

The paper investigates the costs of hybrid retrieval—combining a sparse lexical retriever with a dense neural one—across seven BEIR datasets and multiple fusion strategies. It finds that even the best fusion method can rank relevant documents worse than either component alone on a substantial fraction of queries, a phenomenon dubbed the "fusion tax." The tax is largely a ranking effect that preserves overall recall but harms individual query performance, especially when only one retriever covers a query.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Computation and Language.

arXiv Machine Learning

Aug 4

Do Static Embeddings Add Value to Hybrid Dutch Retrieval?

arXiv:2608. 02112v1 Announce Type: new Abstract: Embedding benchmarks measure standalone model quality, but they do not establish whether a low-cost retriever contributes complementary ranking information once lexical and transformer-based retrieval are already combined.

By Ant\'onio Pereira Barata

llms rag benchmarks

Hugging Face Trending Papers

Aug 16

Dense Expands, Sparse Anchors: Channel-Asymmetric Query Expansion for Hybrid Retrieval

LLM-based query expansion improves retrieval by generating document-like passages. In hybrid retrieval, however, most evaluations fuse fixed top-$L$ dense and sparse rankings. Because the cutoff contr...

llms

arXiv Computation and Language

Sep 30

MERGE: Multi-LLM Ensemble for Retrieval via Generative Enrichment

arXiv:2609.37574v1 Announce Type: cross Abstract: Large Language Models (LLMs) are increasingly used to enrich user queries in information retrieval (IR) so that a standard retriever such as BM25 can...

By Tzu-I Ho, Yung-Yu Shih, Shang-Yu Su, Dongzhe Wang, Yun-Nung Chen

llms benchmarks

arXiv Machine Learning

Jun 4

Training-Free Lexical-Dense Fusion for Conversational-Memory Retrieval

arXiv:2606. 04194v1 Announce Type: new Abstract: Retrieving the few past turns that answer a new query across long multi-session histories is the retrieval bottleneck behind long-term conversational memory (LoCoMo, LongMemEval).

By Christian Lysenst{\o}en

rag benchmarks safety

arXiv Machine Learning

Jun 30

Diagnosing and Mitigating Retrieval Bottlenecks in LLM-Based Cold-Start Recommendation

arXiv:2606. 29947v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used as rerankers in recommender systems, with the expectation that semantic understanding will help in cold-start and long-tail regimes.

By Zhe Dong (University of Maine at Presque Isle), Fang Qin (Stanford University), Manish Shah (Independent Researcher), Yicheng Wang (Independent Researcher)

llms rag benchmarks

arXiv Machine Learning

Aug 26

RetrievalFormer: A Dual-Encoder Transformer for Efficient Approximate Nearest Neighbor Retrieval and Cold-Item Recommendation

arXiv:2608.24079v1 Announce Type: cross Abstract: A shared search-and-recommendation index must score new items from features alone because search has no exploration slot. In a public log covering bo...

By Theodore Rogers, Joe Standerfer, Dmitrii Timoshenko, Haoxue Li, Zuhaib Akhtar, Soyoung Yang

llms

One email a morning, machine-written

Related stories

Do Static Embeddings Add Value to Hybrid Dutch Retrieval?

Dense Expands, Sparse Anchors: Channel-Asymmetric Query Expansion for Hybrid Retrieval

MERGE: Multi-LLM Ensemble for Retrieval via Generative Enrichment

Training-Free Lexical-Dense Fusion for Conversational-Memory Retrieval

Diagnosing and Mitigating Retrieval Bottlenecks in LLM-Based Cold-Start Recommendation

RetrievalFormer: A Dual-Encoder Transformer for Efficient Approximate Nearest Neighbor Retrieval and Cold-Item Recommendation