Improving Parametric Knowledge Access in Reasoning Language Models

2026-02-25 · Melody Ma, John Hewitt arxiv

We study reasoning for accessing world knowledge stored in a language model's parameters. For example, recalling that Canberra is Australia's capital may benefit from thinking through major cities and the concept of purpose-built capitals. While reasoning language models are trained via reinforcement learning to produce reasoning traces on tasks such as mathematics, they may not reason well for accessing their own world knowledge. We first find that models do not generate their best world knowledge reasoning by default: adding a simple "think step-by-step" cue demonstrates statistically significant improvement in knowledge recall but not math. Motivated by this, we propose training models to reason over their parametric knowledge using world-knowledge question answering as a verifiable reward. After reinforcement learning on TriviaQA (+9.9%), performance also improves on Natural Questions, HotpotQA, SimpleQA, and StrategyQA by 4.2%, 2.1%, 0.6%, and 3.0%, respectively. Reasoning models are under-optimized for parametric knowledge access, but can be easily trained to reason better.

📄 PDF Abstract BibTeX arXiv:2602.22193

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement Learning Question Answering Natural Questions

From Storage to Access: Verifiable Activation of Parametric Knowledge in LLMs via Explicit Priming and Implicit Reasoning

2026-08-19 · Zuocheng Ying, Yang Yang, Yumou Wu, Chuanbo Zhu 외 arxiv

Although Large Language Models (LLMs) encode rich factual knowledge in their parameters, reliably recalling and verifying such knowledge remains a key bottleneck in factual question answering. Existing end-to-end methods…

Question Answering

SudoLM: Learning Access Control of Parametric Knowledge with Authorization Alignment

2024-10-18 · Qin Liu, Fei Wang, Chaowei Xiao, Muhao Chen

Existing preference alignment is a one-size-fits-all alignment mechanism, where the part of the large language model (LLM) parametric knowledge with non-preferred features is uniformly blocked to all the users. However, …

All Blocking Language Modeling Language Modelling+1

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks

2020-05-22 · NeurIPS 2020 12 · Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni 외

Large pre-trained language models have been shown to store factual knowledge in their parameters, and achieve state-of-the-art results when fine-tuned on downstream NLP tasks. However, their ability to access and precise…

Fact Verification Question Answering RAG Retrieval+3

Cross-Lingual Exploration for Parametric Knowledge

2026-06-23 · Elisha Diskind, Itamar Trainin, Uri Shaham, Leshem Choshen 외 arxiv

Parametric knowledge in Large Language Models is not equally accessible across languages. As a result, standard inference techniques often struggle to surface localized facts, leading to failures in cross-lingual knowled…

From Retrieval to Generation: Unifying External and Parametric Knowledge for Medical Question Answering

2025-10-21 · Lei Li, Xiao Zhou, Yingying Zhang, Xian Wu arxiv

Medical question answering (QA) requires extensive access to domain-specific knowledge. A promising direction is to enhance large language models (LLMs) with external knowledge retrieved from medical corpora or parametri…

Question Answering Answer Generation