How Far Can Lightweight LLMs Go in Intent-to-SLO Translation?

Sabbir Ahmed; Hazem Abbas; Salimur Choudhury

cs.NINov 5, 2026

How Far Can Lightweight LLMs Go in Intent-to-SLO Translation?

Authors: Sabbir Ahmed, Hazem Abbas, Salimur Choudhury

Organizations: School of Computing, Queen’s University, Ontario, Canada

Abstract

Intent-based networking (IBN) promises QoS provisioning in which the network derives the configuration from operator-stated goals. Its first step, translating a natural-language intent into enforceable service-level objectives (SLOs) grounded in a live snapshot of the network state, increasingly relies on large language models (LLMs). These large models add cost and energy to the management loop and typically require multi-GPU, cloud-hosted infrastructure. Can lightweight open LLMs take over this step dependably? We answer this question with a five-dimensional evaluation of ten LLMs with <10B parameters from five families, under four prompting setups, on 716 placement intents. We evaluate translation performance, intent validation, hallucination, energy impact, and robustness to diverse phrasing. The best configurations translate intents with a strict F1 of 0.943 and reject up to 191 of 195 invalid intents. With few-shot reasoning, the strongest models hallucinate unsupported QoS requirements in only about one response in a hundred, whereas the zero-shot setup struggles. A translated intent costs 0.044-0.148 Wh of GPU energy without reasoning, and added energy cost for reasoning pays off for the weaker models. Furthermore, we show that realistic rephrasing lowers the rejection of invalid intents in 15 of 16 configurations for the best-performing models, leaving robust intent validation as a key open challenge for reliable IBN. Our code and data are available: https://github.com/ggck43/intent2slo.

Figures & tables

Panel of Figure 1 — Fig. 1: Evaluation framework. A lightweight LLM translates an operator intent and the state snapshot of the edge–cloud cluster into SLOs, or rejects the intent. Prompts are zero-shot ( zs ) or few-shot ( fs ), with or without chain-of-thought ( cot ). The output is assessed for translation performance (RQ1), intent validation (RQ2), hallucination (RQ3), energy and CO 2 e (RQ4), and robustness to operator phrasing (RQ5).

Orig.	For svc, please run it on the edge side, make sure the target node has at least 2 GB of memory, and pick the node with the highest storage capacity.
C1	We are retiring a couple of old machines at the end of the month, and the recommender is waiting for a new home. Please run svc on the edge side with a node that offers at least 2 GB of memory, and choose the one with the highest storage capacity. I’ll pass the result to the migration team tomorrow, so thank you for a quick reply.
C2	The recommender has moved between hosts twice since spring and the team wants it settled. Choose the node with the highest storage capacity for svc, make sure it has 2 GB of memory or more, and keep it on the edge side.
C3	Service: svc - Layr: edge side pls - Memory: node need at least 2 GB of mem - Storage: pick the node what have the highest storge capacity Migration is soon so asap if you can. Thx, i really apreciate it.

TABLE I: The three phrasing variants of one valid intent (RQ5).

		zs		fs		zs-cot		fs-cot		Mean
Family	Size	$F_{1}$	EM	$F_{1}$	EM	$F_{1}$	EM	$F_{1}$	EM	$F_{1}$
Qwen3.5	4B	0.918	0.820	0.910	0.774	0.892	0.772	0.929	0.845	0.912
	9B	0.921	0.835	0.922	0.802	0.942	0.879	0.930	0.823	0.929
Gemma 4	E2B	0.591	0.397	0.602	0.399	0.850	0.695	0.886	0.747	0.732
	E4B	0.726	0.403	0.872	0.733	0.883	0.754	0.943	0.879	0.856
Ministral 3	3B	0.408	0.159	0.714	0.493	0.479	0.236	0.847	0.702	0.612

TABLE II: Translation performance of the ten lightweight LLMs on the 521 valid intents; Mean is the

F_{1}

averaged over the four setups. Best value per column in bold, second best underlined.

Fig. 2: Strict

F_{1}

of the two best model families with fs-cot on explicit SLOs (thresholds stated in intent), implicit SLOs (values must be read from snapshot), and all SLOs. Errors lie mostly in implicit SLOs.

Fig. 3: Rejection rate $\rho$ of each model and setup on the 195 invalid intents (darker = more rejected). Examples and reasoning ( fs-cot ) reject most; whereas zero-shot setup and the Llama models reject least.

Fig. 4: Hallucination rate $\eta$ of each model and setup over all 716 intents (lighter = more trustworthy). Examples and reasoning ( fs-cot ) lower it for almost every model, down to about 1% in the best configurations.

Intent (abridged)	Ground-truth SLOs	Model (setup): hallucinated SLO	Why it is a hallucination
“Deploy svc … at least 50 GB of storage capacity and pick the node with the lowest current memory utilization” (valid)	storage cap. $\geq$ 50 GB; memory util. $\leq$ [lowest] 34.27 %	Qwen3.5-9B ( zs ): memory util. $\leq$ [lowest] 18.0 %	18.0 is the storage utilization of the correct node: right node, wrong metric
“I would like to place svc and keep CPU utilization below 40 %” (valid)	CPU util. $\leq$ 40 %	Qwen3.5-9B ( fs-cot ): placement = edge	a placement layer the intent does not name

TABLE III: Hallucinated SLOs of both kinds: an unsupported value (grounded from the wrong metric) and an unsupported requirement (an invented placement); svc = service

Fig. 5:

F_{1}

against GPU energy per valid intent (one request at a time), one panel per model (columns: families; larger model on top), one point per setup. Arrows add reasoning (dashed: zs

\rightarrow

zs-cot ; solid: fs

\rightarrow

fs-cot ). Steep arrows are reasoning worth its energy; flat ones are not.

Fig. 6: Effect of operator phrasing per model and prompting setup. Change, in percentage, between the variants C1–C3 (pooled) and the original intents, for strict $F_{1}$ on valid intents (left) and rejection rate on invalid intents (right).

Fig. 7: Change in rejection rate of invalid intents for each phrasing variant.

Intent-Based Networking

Intent-Based Networking translates high-level goals into validated network policies and uses feedback to manage services, resources, and performance.

See all 24 papers

Intent2Tc: Automated Intent-to-Traffic Control Translation with Language Models
Sep 25, 2026Andrea Masini, Sudipta Acharya, Paolo Bellavista +2Intent-Based Networking
University of Ottawa, Ottawa, ON, Canada · University of Bologna, Italy
AI-driven Intent-Based Networking Approach for Self-configuration of Next Generation Networks
Mar 24, 2026Md. Kamrul Hossain, Walid AljobyIntent-Based Networking Language Model-Based Control
Information and Computer Science Department, King Fahd University of Petroleum and Minerals, Dhahran 31261, Saudi Arabia · IRC for Intelligent Secure Systems, King Fahd University of Petroleum and Minerals, Dhahran 31261, Saudi Arabia
Uncertainty Signals for Network Intent Translation: Risk Ranking and Ambiguity Localization
Sep 3, 2026Ala' A. Alsamarneh, Omar AlhusseinIntent-Based Networking LLM Uncertainty Estimation
Department of Computer Science, College of Computing and Mathematical Sciences Khalifa University, Abu Dhabi, United Arab Emirates

Sep 25, 2026J/K move · Enter open · S save

Intent2Tc: Automated Intent-to-Traffic Control Translation with Language Models

Andrea Masini, Sudipta Acharya, Paolo Bellavista, Luca Foschini+1

University of Ottawa, Ottawa, ON, Canada · University of Bologna, Italy

Intent-Based Networking

Open paper AskPDF arXiv

Automated and highly usable Quality-of-Service (QoS) enforcement requires translating high-level service intents into deployable traffic-management policies. Although intent-based networking (IBN) has simplified policy specification, bridging the gap between business-level intents and executable network configurations remains complex, error-prone, and difficult to automate. This paper presents Intent2Tc, a closed-loop language-model-driven framework that translates business-level traffic-shaping intents into declarative sub-intents and subsequently into validated, executable Linux traffic control (tc) configurations. The framework integrates an Active Queue Management (AQM)-based digital twin (DT) semantic model, automated metadata extraction, critique-driven refinement, and Retrieval-Augmented Generation (RAG)-based knowledge reuse to improve semantic consistency and configuration reliability. We evaluate multiple open-source large language models (LLMs) and small language models (SLMs), together with Claude Sonnet-4.6, on 100 Request for Comments (RFC) 9315-compliant traffic-shaping intents. Across both translation stages, Intent2Tc achieves high semantic fidelity, configuration accuracy, and deployment readiness, with Claude Sonnet-4.6 reaching 0.98 semantic similarity, 1.0 semantic unit coverage, and 0.045 normalized edit distance. Furthermore, RAG reduces token consumption and inference latency while enabling compact models such as Phi-4-mini to approach the performance of substantially larger models. Linux tc serves as the target configuration platform, demonstrating the practical applicability of the proposed framework.

Abstract

Intent-Based Networking

Explore similar work

Intent2Tc: Automated Intent-to-Traffic Control Translation with Language Models

AI-driven Intent-Based Networking Approach for Self-configuration of Next Generation Networks

Uncertainty Signals for Network Intent Translation: Risk Ranking and Ambiguity Localization