llm-kb
← Back to research
Paper

Neuro-Agentic Control Framework Combines LLM Planner with Time-Series Foundation Model for Industrial IoT Security

A new research paper proposes a neuro-agentic control framework that couples an LLM-based planner (e.g., Gemini 2.5 Flash-Lite) with a pre-trained Time-Series Foundation Model to address the safety limitations of using LLMs for closed-loop control in industrial IoT environments. The framework aims to mitigate LLM hallucination risks while leveraging semantic reasoning for security control decisions.

0 engagement·1 source·Fri, Jul 10, 2026, 03:43 AM
The paper, titled 'Neuro-Agentic Control: A Deep Learning-based LLM-Powered Agentic AI Framework for Controlling Security Controls,' was published on arXiv on July 10, 2026. It highlights that cyberattacks on operational technology are causing costly downtime and physical damage, exposing the inadequacy of traditional rule-based monitoring. The proposed architecture uses an LLM-based planner (specifically Gemini 2.5 Flash-Lite) for high-level reasoning and a pre-trained Time-Series Foundation Model for robust, low-level control, aiming to reduce hallucination risks in safety-critical closed-loop control.

Entities

arXiv(tool)Neuro-Agentic Control(concept)Gemini 2.5 Flash-Lite(model)Time-Series Foundation Model(model)

Related

PaperFri, Jul 10, 2026, 05:52 PM

VEXAIoT paper proposes multi-agent LLM framework for autonomous IoT vulnerability exploitation

A new arXiv paper introduces VEXAIoT, an autonomous multi-agent framework for discovering and exploiting vulnerabilities in IoT systems using LLM agents. The work addresses the gap in applying LLM-based penetration testing to IoT-specific weaknesses like constrained hardware and outdated firmware.

0 engagement·1 source·arxiv
arXiv
PaperFri, Jul 10, 2026, 10:19 AM

Paper proposes blockchain-linked LLM-based decision management for telecom fraud control

A new arXiv paper reframes telecom fraud control as blockchain-linked auditable decision management for synthetic telecom/IoT fraud-control requests. The main finding is that a QLoRA-tuned LLM branch becomes much more usable than zero-shot prompting but mainly approaches, rather than outperforms, a lower-cost centralized ensemble.

0 engagement·1 source·arxiv
arXiv
ProductWed, Jul 8, 2026, 04:40 AM

Governed multi-agent execution platform with trading engine

Melaya is a platform for designing and executing governed multi-agent workflows, paired with a trading engine. It uses LLMs to orchestrate agents while ensuring security and compliance, targeting operators who need controlled agent execution.

0 engagement·1 source·rss
RSS
PaperThu, Jul 9, 2026, 07:34 PM

GATS framework eliminates LLM calls during agent planning inference

Researchers propose GATS (Graph-Augmented Tree Search), a planning framework that uses a layered world model and UCB1-based tree search to avoid LLM inference during planning, reducing computational cost and stochasticity. The approach outperforms LATS and ReAct on multi-step planning tasks.

0 engagement·1 source·arxiv
arXiv
PaperThu, Jul 9, 2026, 09:09 PM

TSRouter: Dynamic Modality-Model Selection for Time Series Reasoning

A new framework called TSRouter dynamically selects between LLMs and VLMs for time series reasoning, leveraging their complementary strengths. LLMs preserve exact numerical details but miss global patterns, while VLMs capture patterns but lose fine-grained data. TSRouter chooses the best modality and model per input, balancing accuracy and cost.

0 engagement·1 source·arxiv
arXiv