mlsanityData to 14 Oct 2026

Probe, Don't Prompt: A Hidden-State Probe for Metadata Filtering in Multi-Meta-RAG

arXiv 2607.03929 ↗First posted 4 Jul 2026cs.CL cs.AI cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Multi-Meta-RAG improves retrieval for multi-hop question answering by filtering a vector store on metadata (the news source) that it extracts from each query by prompting gpt-3.5-turbo. We show this proprietary, free-form extractor can be replaced by a local, deterministic probe trained on the hidden states of a small open-source language model. On all 2556 MultiHop-RAG queries the probe reaches 90.9% set-exact accuracy against 88.0% for a model-free substring baseline and 80.9% for GPT-3.5, a margin that comes entirely from null queries, on which GPT-3.5 never abstains; on non-null queries all three stay within about a point. Because the probe's output space is exactly the fixed 49-source vocabulary, it cannot drift outside the allow-list as the prompted model does. Three design choices make it work: selecting a shallow layer, mean pooling, and class-imbalance-aware multi-label training over the long tail of sources. A 135M-parameter model lands within ~1.5 points of a 1.5B one, so the filter is cheap to output: a partial forward pass through the first few layers plus one linear head, with no API. The code is available at https://github.com/mxpoliakov/Multi-Meta-RAG.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 16 of its 18 references that are in the map. First posted on arXiv 4 Jul 2026.

Builds on

16 of 18 references.

References · 16

  1. 2015Distilling the Knowledge in a Neural Networkcited by 10,324
  2. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,297
  3. 2024Qwen2.5 Technical Reportcited by 4,942
  4. 2020Dense Passage Retrieval for Open-Domain Question Answeringcited by 3,824
  5. 2016Understanding intermediate layers using linear classifier probescited by 1,353
  6. 2019BERT Rediscovers the Classical NLP Pipelinecited by 1,059
  7. 2021Probing Classifiers: Promises, Shortcomings, and Advancescited by 876
  8. 2019Designing and Interpreting Probes with Control Taskscited by 640
  9. 2018What you can cram into a single vector: Probing sentence embeddings for linguistic propertiescited by 457
  10. 2024LLM2Vec: Large Language Models Are Secretly Powerful Text Encoderscited by 404
  11. 2025SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Modelcited by 335
  12. 2025Layer by Layer: Uncovering Hidden Representations in Language Modelscited by 311
4 more
  1. 2024MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queriescited by 257
  2. 2017A Network Perspective on Stratification of Multi-Label Datacited by 37
  3. 2024Multi-Meta-RAG: Improving RAG for Multi-Hop Queries using Database Filtering with LLM-Extracted Metadatacited by 16
  4. 2025Optimizing Multi-Hop Document Retrieval Through Intermediate Representationscited by 5

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026PAR2-RAG: Planned Active Retrieval and Reasoning for Multi-Hop Question Answering
  2. 2025Investigating Retrieval-Augmented Generation Systems on Unanswerable, Uncheatable, Realistic, Multi-hop Queries
  3. 2024W-RAG: Weakly Supervised Dense Retrieval in RAG for Open-domain Question Answering
  4. 2025ComposeRAG: A Modular and Composable RAG for Corpus-Grounded Multi-Hop Question Answering
  5. 2025Question Decomposition for Retrieval-Augmented Generation
  6. 2026BridgeRAG: Training-Free Bridge-Conditioned Retrieval for Multi-Hop Question Answering
  7. 2026ConRAG: Consensus-Driven Multi-View Retrieval for Multi-Hop Question Answering
  8. 2025Retrieval-Augmented Generation: A Comprehensive Survey of Architectures, Enhancements, and Robustness Frontiers