Unlocking the Path Towards AI-Native Networks with Optimized Lightweight Large Language Models
Bibliographic record
Abstract
In the rapidly advancing era of Artificial Intelligence (AI), Large Language Models (LLMs) have emerged as a pivotal force, revolutionizing the automated creation of diverse content tailored to user preferences and intents. At the same time, Machine Learning (ML), particularly Deep Learning (DL), has achieved state-of-the-art (SoTA) performance in optimization and inference tasks across various domains, including telecommunications. However, the division of technological domains poses challenges to the integration of powerful AI/ML capabilities towards realizing the vision of “AI-native” networks, such as future 6G networks, aiming to offer ubiquitous intelligence across their infrastructure and service planes, seamlessly adapting and evolving to support new application classes. The paper proposes a novel out-of-the-box framework, TimesLM, for automated time series analytics leveraging Large Language Models (LLMs). It addresses the challenge of developing general-purpose solutions by utilizing lightweight LLMs as forecasters and optimizers. TimesLM offers zero-shot and few-shot inference capabilities, enabling confident uncertainty navigation for decision-making in AI-Native networks. Its architecture integrates global optimization with lightweight LLMs, followed by input enrichment and local prompt optimization. Evaluation results on diverse real-world datasets demonstrate improved accuracy and reduced latency, making TimesLM highly competitive against SoTA approaches, while reducing operational costs and carbon-footprint due to its lightweight LLMs. A limitations analysis is also presented. The proposed framework stands as a promising and sustainable step towards realizing the envisioned capabilities of AI-Native networks through efficient and effective time series analytics.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.004 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.003 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".