mlsanityData to 14 Oct 2026

Does the Selected Object Reach the Reader? Auditing Identity Handoffs in Grounded Language-Model Pipelines

arXiv 2609.04579 ↗First posted 4 Sep 2026cs.AI cs.CLcited by 0

Not on the map. Why?

Abstract (arXiv)

Grounded language-model pipelines can be divided into three stages: selecting an object, retrieving passages for it, and using that evidence to answer. If the selected object must reach the reader, losing it breaks the handoff. Benchmark recall checks the dataset-linked object, which can differ. We audit 600 HybridQA questions across three selector families. On 1,463 resolvable records where the selected object matches the dataset-traced passage, exact key lookup and exact title matching return the object every time. With every ranked rule given the same decoded selected title, body-only BM25 omits it on 389 records (26.6%) at cutoff five, while hybrid retrieval with reranking omits it on 14 (1.0%). The two identities differ on 329 of 1,792 resolvable records. With original-question rankings, their top-five checks disagree on 106 records (5.9%). Frozen reader comparisons associate the aligned object's presence with 28.6 to 31.0 points higher exact match. In a deliberately selected 64-item cohort, removing that passage sharply lowers exact match, while removing a similar-length comparison passage does not reproduce the drop. We release the Returned-Object Profile (ROP), an executable record of the target, returned-ID field, cutoff, membership rule, and complete expected population, with data and an offline replay.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 18 of its 35 references that are in the map. First posted on arXiv 4 Sep 2026.

Builds on

18 of 35 references.

References · 18

  1. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,297
  2. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,695
  3. 2022Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questionscited by 923
  4. 2019ERASER: A Benchmark to Evaluate Rationalized NLP Modelscited by 530
  5. 2022Precise Zero-Shot Dense Retrieval without Relevance Labelscited by 517
  6. 2021ColBERTv2: Effective and Efficient Retrieval via Lightweight Late Interactioncited by 476
  7. 2023Enhancing Retrieval-Augmented Large Language Models with Iterative Retrieval-Generation Synergycited by 346
  8. 2020HybridQA: A Dataset of Multi-Hop Question Answering over Tabular and Textual Datacited by 264
  9. 2019PullNet: Open Domain Question Answering with Iterative Retrieval on Knowledge Bases and Textcited by 184
  10. 2020Open Question Answering over Tables and Textcited by 163
  11. 2024Are self-explanations from Large Language Models faithful?cited by 101
  12. 2020Efficient One-Pass End-to-End Entity Linking for Questionscited by 76
6 more
  1. 2019Knowledge Guided Text Retrieval and Reading for Open Domain Question Answeringcited by 59
  2. 2024RaFe: Ranking Feedback Improves Query Rewriting for RAGcited by 49
  3. 2021Open Domain Question Answering with A Unified Knowledge Interfacecited by 32
  4. 2021Multi-Row, Multi-Span Distant Supervision For Table+Text Questioncited by 12
  5. 2024Entity Retrieval for Answering Entity-Centric Questionscited by 10
  6. 2026AgenticRAGTracer: A Hop-Aware Benchmark for Diagnosing Multi-Step Retrieval Reasoning in Agentic RAGcited by 5

Built on by

Not cited yet.