LLM-Based Agentic Negotiation Framework for 6G Network Slicing
A research paper proposes a risk-aware framework for agentic negotiation in 6G autonomous networks, addressing the uncertainty neglect bias in large language model (LLM)-powered agents. The framework uses Digital Twins (DTs) to predict full latency distributions and applies Conditional Value-at-Risk (CVaR) from extreme value theory to shift decision-making from mean-based reasoning to tail-risk management. This approach aims to ensure robust resource allocation in 6G network slicing by building a statistically-grounded buffer against worst-case outcomes. The paper is available on arXiv under identifier 2511.19175.
Key facts
- The paper addresses uncertainty neglect bias in LLM-powered agents for 6G networks.
- It proposes a risk-aware framework for agentic negotiation in network slicing.
- Agents use Digital Twins to predict full latency distributions.
- Conditional Value-at-Risk (CVaR) from extreme value theory is used for evaluation.
- The framework shifts reasoning from mean to tail-risk for worst-case outcomes.
- The paper is available on arXiv with ID 2511.19175.
- The framework aims to ensure robust resource allocation in 6G.
- It builds a statistically-grounded buffer against extreme events.
Entities
Institutions
- arXiv