arXiv:2603.12414v1 Announce Type: new Abstract: State Space Models (SSMs) such as Mamba achieve linear-time sequence processing through input-dependent recurrence, but this mechanism introduces a critical safety vulnerability. We show that the spectral radius rho(A-bar) of the discretized transition operator governs effective memory horizon: when an adversary drives rho toward zero through gradient-based Hidden State Poisoning, memory collapses from millions of tokens to mere dozens, silently destroying reasoning capacity without triggering output-level alarms. We prove an Evasion Existence Theorem […]

Ver mais

Recent advances in large language models (LLMs) have made it feasible to use them as automated debugging tutors, but it remains unclear how much can be gained by moving from single-model tutors to multi-agent councils with separated roles. We study this question in an offline simulation on 200 debugging cases drawn from an online judge, spanning 20 problems split into course-style and contest-style challenge tracks. We compare four single-model tutors based on current frontier models with four councils […]

Ver mais

Landslides represent a major geohazard with severe impacts on human life, infrastructure, and ecosystems, underscoring the need for accurate and timely detection approaches to support disaster risk reduction. This study proposes a modular, multi-model framework that fuses Sentinel-2 optical imagery with Sentinel-1 Synthetic Aperture Radar (SAR) data, for robust landslide detection. The methodology leverages multi-encoder vision transformers, where each data modality is processed through separate lightweight pretrained encoders, achieving strong performance in landslide detection. In addition, the integration […]

Ver mais

Personalized Federated Learning (PFL) aims to deliver effective client-specific models under heterogeneous distributions, yet existing methods suffer from shallow prototype alignment and brittle server-side distillation. We propose HEART-PFL, a dual-sided framework that (i) performs depth-aware Hierarchical Directional Alignment (HDA) using cosine similarity in the early stage and MSE matching in the deep stage to preserve client specificity, and (ii) stabilizes global updates through Adversarial Knowledge Transfer (AKT) with symmetric KL distillation on clean and adversarial proxy data. Using […]

Ver mais

Representation learning models exhibit a surprising stability in their internal representations. Whereas most prior work treats this stability as a single property, we formalize it as two distinct concepts: statistical identifiability (consistency of representations across runs) and structural identifiability (alignment of representations with some unobserved ground truth). Recognizing that perfect pointwise identifiability is generally unrealistic for modern representation learning models, we propose new model-agnostic definitions of statistical and structural near-identifiability of representations up to some error tolerance $ε$. […]

Ver mais

The “Perfect Decision” Trap We’re entering the era where AI doesn’t just answer questions — it selects actions. Supply chain routing. Credit risk. Fraud detection. Treatment planning. Portfolio optimisation. The pitch is always the same: “Give the model data and objectives, and it will find the best move.” And in a narrow, mathematical sense, it can. But here’s the catch: optimisation is a superpower and a liability. Because if a system can optimise perfectly, it can also optimise […]

Ver mais

I have launched many businesses and built multiple teams. If there is one super power that I think helped me in my entrepreneurial journey, it is the subtle art of delegating. As I am now using AI at work and in personal life and building an AI-first product, I see some people simply not getting anywhere with AI. Quality outputs or meaningful interactions seem to escape them. At a closer look though, I discovered that AI success and […]

Ver mais

The rapid proliferation of large language model (LLM) powered multi-agent systems creates a non-trivial combinatorial optimization problem: routing heterogeneous tasks to the most cost-effective model tier while maintaining quality guarantees. Current production systems rely on static lookup tables, which over-provision expensive models and waste computational budget. We formalize the LLM Cascade Routing Problem (LCRP) as a Quadratic Unconstrained Binary Optimization (QUBO) problem and solve it using the Quantum Approximate Optimization Algorithm (QAOA). We benchmark QAOA against greedy heuristics […]

Ver mais

To address the difficulty of maintaining stable structural modeling in intelligent decision systems under complex dynamic environments and the sensitivity of decision behavior to state uncertainty, this study proposes an intelligent decision modeling method based on structure aware learning. The method constructs structured state representations that map environmental observations, state evolution relationships, and decision processes into a unified representation space. This design enables explicit modeling of internal structural constraints and dynamic dependencies within the system. During decision modeling, […]

Ver mais