mlsanityData to 14 Oct 2026

OCC-RAG: Optimal Cognitive Core for Faithful Question Answering

arXiv 2606.00683 ↗First posted 30 May 2026cs.CLcited by 0

On the map: LLM reasoning → Knowledge- and tool-augmented reasoning

Abstract (arXiv)

Recent progress in the development of language models has been defined by scale, with each generation absorbing more of the world's knowledge into its weights. However, many practical applications benefit more from robust reasoning than from extensive parametric knowledge. In this setting, task-specialized small language models (SLMs) offer a principled design choice. We introduce Optimal Cognitive Core (OCC), a family of SLMs built around this premise. As a variant of OCC, we present OCC-RAG, optimized for faithful question answering (QA) grounded in the provided context. This task directly aligns with the OCC design approach, requiring multi-hop reasoning over supplied passages while ignoring memorized knowledge. To train OCC-RAG, we implement a novel pipeline for synthesizing multi-context, multi-hop QA data at scale, producing a corpus of over three million examples targeting multi-hop reasoning, strict context faithfulness, and calibrated abstention. We release OCC-RAG-0.6B and OCC-RAG-1.7B, both mid-trained on this corpus. The models produce structured reasoning traces with source citations grounded in literal quotes from the context. Through OCC-RAG, we demonstrate that compact, task-specialized SLMs can match or exceed general-purpose models 2 -- 6x their size across multi-hop reasoning (HotpotQA, MuSiQue, TAT-QA), faithfulness (ConFiQA), and refusal (MuSiQue-Un) benchmarks.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Knowledge- and tool-augmented reasoning, one of 17 directions in LLM reasoning. Builds on 26 of its 34 references that are in the map. First posted on arXiv 30 May 2026.

Builds on

26 of 34 references.

Closest ancestors · 2

Same direction, newest first

  1. 2026SIN-Bench: Tracing Native Evidence Chains in Long-Context Multimodal Scientific Interleaved Literature15 Jan
  2. 2025Even Small Reasoners Should Quote Their Sources: Introducing the Pleias-RAG Model Family25 Apr

Other references · 24

  1. 2018HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answeringcited by 3,754
  2. 2020Don't Stop Pretraining: Adapt Language Models to Domains and Taskscited by 1,565
  3. 2021MuSiQue: Multihop Questions via Single-hop Question Compositioncited by 916
  4. 2020It's Not Just Size That Matters: Small Language Models Are Also Few-Shot Learnerscited by 669
  5. 2021TAT-QA: A Question Answering Benchmark on a Hybrid of Tabular and Textual Content in Financecited by 408
  6. 2025Small Language Models are the Future of Agentic AIcited by 270
  7. 2025AbstentionBench: Reasoning LLMs Fail on Unanswerable Questionscited by 145
  8. 2024xLAM: A Family of Large Action Models to Empower AI Agent Systemscited by 111
  9. 2024FaithEval: Can Your Language Model Stay Faithful to Context, Even If "The Moon is Made of Marshmallows"cited by 86
  10. 2023TinyGSM: achieving >80% on GSM8k with small language modelscited by 70
  11. 2025Evaluating Step-by-step Reasoning Traces: A Surveycited by 63
  12. 2023Natural Language Reasoning, A Surveycited by 61
12 more
  1. 2024Context-DPO: Aligning Language Models for Context-Faithfulnesscited by 50
  2. 2024Characterizing LLM Abstention Behavior in Science QA with Context Perturbationscited by 26
  3. 2025Task Matters: Knowledge Requirements Shape LLM Responses to Context-Memory Conflictcited by 16
  4. 2026Task-Specific Efficiency Analysis: When Small Language Models Outperform Large Language Modelscited by 15
  5. 2025When Models Lie, We Learn: Multilingual Span-Level Hallucination Detection with PsiloQAcited by 14
  6. 2024Quantifying reliance on external information over parametric knowledge during Retrieval Augmented Generation (RAG) using mechanistic analysiscited by 7
  7. 2026When Silence Is Golden: Can LLMs Learn to Abstain in Temporal QA and Beyond?cited by 7
  8. 2025Wikontic: Constructing Wikidata-Aligned, Ontology-Aware Knowledge Graphs with Large Language Modelscited by 6
  9. 2025Will It Still Be True Tomorrow? Multilingual Evergreen Question Classification to Improve Trustworthy QAcited by 5
  10. 2025DRAGOn: Designing RAG On Periodically Updated Corpuscited by 5
  11. 2026Long-Document QA with Chain-of-Structured-Thought and Fine-Tuned SLMscited by 3
  12. 2025On Synthesizing Data for Context Attribution in Question Answeringcited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Evaluating Prompt Engineering Techniques for RAG in Small Language Models: A Multi-Hop QA Approach
  2. 2024MINTQA: A Multi-Hop Question Answering Benchmark for Evaluating LLMs on New and Tail Knowledge
  3. 2025Credible Plan-Driven RAG Method for Multi-Hop Question Answering
  4. 2026PAR2-RAG: Planned Active Retrieval and Reasoning for Multi-Hop Question Answering
  5. 2023Self-prompted Chain-of-Thought on Large Language Models for Open-domain Multi-hop Reasoning
  6. 2026Reasoning in Trees: Improving Retrieval-Augmented Generation for Multi-Hop Question Answering
  7. 2025Collab-RAG: Boosting Retrieval-Augmented Generation for Complex Question Answering via White-Box and Black-Box LLM Collaboration
  8. 2025REAP: Enhancing RAG with Recursive Evaluation and Adaptive Planning for Multi-Hop Question Answering