Skip to content
Log in
#small language model Open access

Are reasoning paradigms scale-aware? A cross-paradigm verification of prompting, retrieval, and knowledge-graph scaffolding for small language models

Zhenzhen Gu Jie Liu Xian Liu
Aug 2026 · Journal of King Saud University: Computer and Information Sciences · Vol 38 · 0 citations · 44 references

TL;DR

A scale-aware comparative study of reasoning enhancement for SLMs across three major families of methods: prompting-based reasoning, retrieval-based augmentation, and knowledge graph guided scaffolding shows that reasoning-enhancement strategies are not universally transferable across model scales under the evaluated settings.

Abstract

Recent reasoning-enhancement paradigms, including chain-based prompting, retrieval-augmented generation, and graph-guided reasoning, have substantially improved the performance of large language models on question answering and knowledge-intensive tasks. However, whether these paradigms remain effective when transferred to small language models (SLMs) is still insufficiently understood. This question is particularly important for real-world deployments, where smaller models are often preferred due to resource constraints, latency requirements, and privacy considerations. In this paper, we conduct a scale-aware comparative study of reasoning enhancement for SLMs across three major families of methods: prompting-based reasoning, retrieval-based augmentation, and knowledge graph guided scaffolding. Using a unified experimental protocol, we evaluate representative paradigms on diverse question answering benchmarks and examine their behavior from three complementary perspectives: effectiveness, efficiency, and robustness. Our results show that reasoning-enhancement strategies are not universally transferable across model scales under the evaluated settings. In particular, chain-based prompting can become unstable on ultra-small models and is associated with longer, harder-to-control reasoning traces, while simply increasing unstructured retrieved context often yields limited gains at substantially higher token and latency costs. By contrast, graph-organized and graph-guided support is associated with more stable improvements in the evaluated multi-hop and knowledge-intensive settings, although its benefit remains dependent on graph quality, task structure, query-time cost, and the specific construction choices. These observations suggest that the success of reasoning enhancement depends not only on the method itself but also on the interaction with model scale and evidence organization. Beyond benchmarking, our study offers practical guidance for designing more inspectable, traceable, and resource-aware question answering systems, while pointing to structured evidence as a promising direction for adapting reasoning support to smaller foundation models under appropriate deployment conditions.

DOI Semantic Scholar
Read PDF

Similar papers

#small language model Open access Aug 2026

LifeSciBench: Evaluating Language Models on Realistic, Expert-Level Tasks in the Life Sciences

LifeSciBench is introduced, a benchmark of 750 expert-authored tasks designed to evaluate whether language models can handle realistic life science research work, with each constituent task paired with a human expert-written rubric.

Amelia Liu, Andrew Ho, Anne Marie Droste et al. · 2 citations
#small language model Preprint Aug 2026

CritICL: Inference-Time Weak-to-Strong Generalization from Small Language Model Failure Modes

Experimental results show that CritICL consistently outperforms standard in-context learning and achieves performance competitive with or superior to test-time scaling methods, while requiring significantly fewer generations and lower token cost.

Yu-Fan Wu, Yinghui He, Zhengyi Hu et al. · 1 citation
#artificial intelligence Preprint Aug 2026

TestifAI: Tomography-Based Testing for Deep Learning Systems

TestifAI, a deep learning testing framework for efficient and accurate estimation of robustness against combinations of perturbations, is proposed and partial model tomography is introduced, a novel approach to reconstructing model behaviour in a multi-perturbation space from tests that apply only a small number of perturbations.

Arooj Arif, T. Hartung, E. Botoeva et al. · 1 citation
#small language model Sep 2026

Low Carbon Scheduling of Integrated Energy System Based on Large Language Model-Embedded Multi-Agent Reinforcement Learning

The complex multi-energy coupling characteristics inherent to integrated energy system (IES) present unprecedented challenges for the implementation of low-carbon scheduling. Existing optimization methods often exhibit limitations in system scalability, algorithm adaptivity, and carbon reduction efficacy for complex IES. This paper proposes a Large Language Model (LLM)-Embedded Multi-Agent Reinforcement Learning (LEMARL) to address the aforementioned issues. The proposed method integrates the global perception capability of LLMs with the dynamic optimization capability of MARL. Specifically, the LLM-Embedded module generates high-quality reward functions and policy frameworks from a global perspective, while the MARL module leverages these LLM-generated strategies for distributed interactive iterations—greatly enhancing computation efficiency and scalability. Simulation results demonstrate that LEMARL reduces carbon emissions by 7.76% and simultaneously decreases operating costs by 4.49% in a small-scale IES. Furthermore, LEMARL also exhibits superior applicability and scalability in large-scale IES of the IEEE 141-bus power grid integrated with 51-node thermal system.

Chen Xia, Tong Gou, Yinliang Xu et al. · 1 citation
#small language model Preprint Aug 2026

Apodex 1.1: Scaling Agentic Intelligence for Complex Work

General-purpose language models can reason and synthesize knowledge, but complex work also requires sustained interaction with files, information sources, and executable code, together with state maintenance, failure recovery, and verifiable delivery. We call this \emph{working capability}: sustained, verifiable progress toward a real-world objective. Apodex 1.1 develops this capability along two complementary dimensions. \emph{Environment Scaling} expands the diversity and verifiability of executable file, search, and code environments, while \emph{Agentic Coordination Scaling} trains agents to decompose long-horizon tasks, delegate parallel work, integrate asynchronous results, and replan. A shared execution harness and AgentOS maintain task state and provenance across tools and agents, and training turns environment trajectories and coordination traces into reliable behavior. Across complex professional work, finance, scientific research, mathematics, coding, and search, Apodex 1.1 reaches the leading performance band despite using a substantially smaller model than many frontier systems. The 35B-parameter Apodex 1.1 Mini further retains strong working capability in a locally deployable form. These results ground agentic intelligence in useful, verifiable work completed over time and advance our goal of building a \emph{Heavy-Duty Solver} for ambitious, long-running tasks.

Apodex Team B. An, B. Li, B. Wang et al. · 1 citation

Related blog posts