SLM-FARL: Small Language Model Driven Federated Reinforcement Multi-Agentic Framework underlying 6G Edge Networks
Bibliographic record
Abstract
Emerging sixth-generation (6G) edge networks demand intelligent, scalable, and privacy-preserving learning systems that support real-time decision-making and natural language-driven control. In addition to training, these systems must also support federated unlearning (FU), the ability to selectively remove user data without full model retraining. However, existing federated learning (FL) and FU frameworks lack adaptability, require manual hyperparameter tuning, and are ill-suited for dynamic, resource-constrained environments. To address these challenges, we propose SLM-FARL, a hierarchical multi-agent deep reinforcement learning (MARL) framework that integrates small language models (SLMs) with FL and FU processes for autonomous, privacy-compliant learning aligned with user-level data removal demands. We implement a customized MAPPO algorithm to enable stable and adaptive policy updates across distributed SLM agents, orchestrated by a central LLM controller that supports human-in-the-loop interaction. To ensure real-time responsiveness and deployment efficiency, we incorporate SLM optimization techniques such as quantization and knowledge distillation, reducing model size and latency while maintaining performance. The proposed framework is evaluated on the UCI Adult dataset using 120 clients and demonstrates up to 15.78% higher FL-FU accuracy compared to baseline methods. The MAPPO Loss decreased by 90.12% indicates highly effective MARL convergence and Policy Entropy drop by 84.31% shows it confident policy decisions. The KD demonstrated an overall 17.56% improved performance over other model compression techniques. Thus, these metrics affirms the robustness and adapt-ability of our proposed SLM-FARL framework.
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 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| 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".