Artificial Intelligence
5 papers, newest first.
-
GradReason: LLMs as Meta-Learners that Replace Training Epochs with Training-Data-Free Weight Updates
Ali Mohammadi Ruzbahani, Abbas Yazdinejad, Hadis KarimipourStandard training recipes make a fixed number of passes over the training data, even when an epoch mostly adds mini-batch noise. We study whether a large language model (LLM) that reads the weight and gradient history of a run can replace some of these epochs with updates that compute no gradient and read no training image. In GradReason, an LLM agent is called every ten epochs. In a restricted sandbox it analyses recent weight snapshots, epoch-mean gradients and validation metrics, then submits a Python program that computes new weights and a plan marking each of the next ten epochs as data or virtual. A virtual epoch is kept only if it passes a forward-only check on 5,000 held-out validation images. In 120 runs on CIFAR-10 and CIFAR-100 with three architectures, Claude Opus 5.5 used 32.8\% fewer data epochs than the 80-epoch recipe at 0.31 percentage points lower test accuracy (95\% CI $-0.72$ to $+0.09$; 0.94 points on ResNet-18), on the recipe's own frontier of data against accuracy. Unprompted, Opus averaged recent weight snapshots in every run. This choice produced the savings: a fixed rule with the same acceptance check but extrapolations from prior work saved 5.4\% of the data epochs, and the same rule with averaging added saved 25.0\%. The LLM's contribution is to find a working update within a single run; once found, a fixed rule implements it equally well. An Opus run cost 0.68 USD, but at this scale the agent's latency exceeded the GPU time it saved.
-
PIA: Agentic Physics-Informed Runtime Anomaly Detection
Ali Mohammadi Ruzbahani, Abbas Yazdinejad, Hadis KarimipourRuntime anomaly detectors for cyber-physical systems are either learned as black boxes from normal data, which miss stealthy manipulations that keep every signal inside its normal range, or derived by hand from physical models, which requires expert effort for every plant. We introduce physics-informed agents (PIA), in which a large language model agent writes the detector itself as a set of physical residuals, such as pump characteristics, tank mass balances, and controller logic, and refines every candidate against normal operating data through an evaluation workbench before deployment. The result is a compact, human-readable detector that runs in milliseconds and names the component behind each alarm. On the BATADAL water distribution attack benchmark, PIA with Claude Opus 5.5 reaches an F1 of 0.878 on the test attacks against 0.504 for the strongest of six statistical and deep baselines, keeps detecting attacks concealed by replayed and offset sensor readings, and attributes every detected test attack to the attacked component. Ablations show that testing hypotheses against data is decisive: agents that write residuals from the plant description alone lose up to 0.28 F1, whereas agents that see only anonymized signals recover most of the physical structure.
-
FedReason: Large Language Models as Weight-Space Meta-Aggregators for Heterogeneous Federated Learning
Ali Mohammadi Ruzbahani, Abbas Yazdinejad, Hadis KarimipourFederated learning servers combine client models with a fixed aggregation rule, and each rule in common use is designed for one kind of heterogeneity. We study whether a large language model (LLM) that inspects the client models can choose the rule itself, without any training on the server. In FedReason, the LLM is called every ten rounds. It analyses the weights by running code in a restricted server-side sandbox, sees only the printed results, may score a few candidates on a 500-image proxy set, and submits a short Python aggregation program. The server applies the program until the next call and falls back to FedAvg whenever its output is invalid. Full weight tensors never leave the server. We evaluated FedReason with Claude Opus 5.5 on CIFAR-10 and CIFAR-100 with ResNet-18 under Dirichlet label skew, width-heterogeneous sub-models and Byzantine clients, in 168 runs against 14 baselines. Across 12 paired label-skew runs it beat FedAvg by 1.26 percentage points on average (10 wins), and a control that spends the same per-round proxy budget on choosing among standard rules by 1.08 points. It was the strongest server-side rule on CIFAR-100 but did not beat SCAFFOLD, which changes client training. With three of ten clients malicious, it excluded all sign-flipping clients from the first round and stayed about one point below FLTrust without training a root model. Without proxy feedback, accuracy fell from 73.8\% to 38.8\% in one seed. A 60-round run cost 0.79 USD in API fees on average.
-
Multi-Agent Daily Planning System (MADPS): A GAIA-Based Architecture for Dynamic Scheduling
Shuvam Agarwala, Ali Mohammadi RuzbahaniThis document outlines the GAIA methodology design for a single-user Multi-Agent Daily Planning System (MADPS) engineered to generate, adapt, and personalize daily schedules. The architecture consists of three core autonomous components: the User Profile Agent (UPA), the Task Management Agent (TMA), and the Planning & Adaptation Agent (PAA). MADPS is designed to maximize priority satisfaction under constraints, minimize re-planning latency and schedule churn during task disruptions, and personalize schedules by learning user energy curves and preference weights. To support these objectives, the system employs FIPA-style interaction protocols and a dual-strategy PAA workflow, utilizing rapid, greedy local repairs for minor schedule disruptions and Genetic Algorithm (GA)-driven global optimization for system-wide impacts. Proposed schedules are evaluated using a comprehensive fitness function that carefully weights priority satisfaction, energy alignment, transition minimization, and hard constraint penalties to ensure feasible and highly optimized daily plans.
-
Decentralizing Cybersecurity across Artificial Intelligence, Internet of Things, Blockchain, Machine Learning, and Agentic AI: A Systematic Literature Review
Rezvan AbouzadehCybersecurity was built around a perimeter, a trusted central server, and a human analyst who reads the alerts. None of the three holds up well in deployments where billions of constrained devices generate data at the edge, models are trained across organizational borders, ledgers record who did what, and autonomous software agents act at machine speed on behalf of people. This article reviews how decentralization is being used to respond to that shift. We code and synthesize 216 studies, 94 of them drawn from one research programme on federated, blockchain-based, and autonomous defence and 122 from the wider literature on mechanisms, attacks, and standards, across five technology families: the Internet of Things, machine learning, federated and other decentralized learning, blockchain, and agentic artificial intelligence. Four findings stand out. Decentralized learning is the most mature of the strands and now has a credible answer to poisoning and inference threats, though at a cost in complexity that few papers measure. Blockchains contribute auditability and identity more reliably than they contribute detection. Agentic AI is the least settled strand, with evidence of attack running well ahead of evidence of defence. And very few studies combine more than two families, so the integrated architectures the field keeps proposing are mostly untested. We distil three reusable algorithms and a layered reference architecture, and we close with a research agenda built around accountability, evaluation, and post-quantum readiness.