Research
One question, asked in four different fields.
How do you trust a model's output when you cannot check it directly? In domain adaptation that means choosing a hyper-parameter with no target labels. In reinforcement learning it means assigning credit from two demonstrations. In generative AI it means binding the output to a contract a solver can verify. Same question.
- 380
- 8
- 8
Selected papers
Every card opens an explainer written for someone outside the subfield, with the paper itself a click away.
43 of 45 downloadable
Peer-reviewed work, preprints and drafts. Where I am not first author, the paper page says so.
15 papers
- Cortex: A Fixed-Point Theory of Governed Coding Agents
- HyDRA: A Hybrid-Driven Reasoning Architecture for Verifiable Knowledge Graphs
- Primality Testing via Circulant Matrix Eigenvalue Structure: A Novel Approach Using Cyclotomic Field Theory
- Ringdown Bounds on UV-Regularized Black-Hole Cores
- Large Language Models Can Self-Improve At Web Agent Tasks
- SymbolicAI: A framework for logic-based approaches combining generative models and solvers
- Parameter Choice and Neuro-Symbolic Approaches for Deep Domain-Invariant Learning
- Addressing Parameter Choice Issues in Unsupervised Domain Adaptation by Aggregation
- A Dataset Perspective on Offline Reinforcement Learning
- Align-RUDDER: Learning From Few Demonstrations by Reward Redistribution
- Reactive Exploration to Cope with Non-Stationarity in Lifelong Reinforcement Learning
- InfODist: Online distillation with Informative rewards improves generalization in Curriculum Learning
- The balancing principle for parameter choice in distance-regularized domain adaptation
- XAI and Strategy Extraction via Reward Redistribution
- Artificial Intelligence, Market Power and India in a Multipolar World
Papers an agent produced end to end, published unedited and unreviewed. The process is the artefact, not the result.
30 papers
- A Contradiction-Aware Survey Framework for Multi-Objective Decision Support
- AutoTW-ASP: Automatic Low-Treewidth Encoding Synthesis and Backend Routing for Neurosymbolic ASP
- AutoTW-ASP: Automatic Low-Treewidth Rewrite Synthesis and Uncertainty-Aware Backend Routing for Exact Neurosymbolic ASP Training
- Benchmarking and Selecting State-of-the-Art Modern Fourier Transformation Methods
- Calibrated Hybrid Evaluation of Quantum Reservoir Classification Under Finite-Shot and Simulability Constraints
- Certified Regime Mapping for Quantum Reservoir Computing Under Parity-Constrained Evaluation
- Compositional Security Control for AI-Assisted Coding Workflows: A Threat-Model-Grounded Security–Productivity Frontier
- Conditional Constrained Routing and Metric Bridging for SymbolicAI Workflows Under CPU-Only Budgets
- Conservative Offline RL with Uncertainty-Aware Policy Improvement
- Contract-Governed Multi-Agent Graph Orchestration for Long-Horizon Autonomous Research Pipelines
- Curiosity-Conditioned Goal-Optimal Reinforcement Learning
- Dependence-Aware Multi-Head Activation Monitoring for Distribution Shift and OOD Reliability
- Detecting Physical and Procedural Bias in Lottery Draws: A Number-Theoretic and Statistical Study
- Do Sunspot Cycles Causally Affect Social Tensions and Population Harm in Developing Countries?
- Dual-Timescale Task-Agnostic Activations for Continual Learning: Stability Guarantees and Boundary-Case Evidence
- Durable Engraftment Modeling for Stem-Cell-Derived Islet Replacement in Type 1 Diabetes
- Entropy-Aware Memory Systems for Continual Learning: Balancing Neuroplasticity and Stability Under Stochastic Workloads
- Glucose-Responsive Insulin Design via Hybrid Machine Learning, Molecular Dynamics, and Pareto Selection
- Inner-Shell Raman X-Gate Tradeoffs for a Neutral ¹⁷¹Yb Nuclear Qubit at Fixed Optical Power
- Interference-Gated Dynamic Activation for Task-Agnostic Continual Learning: A Formal-Empirical Audit of Stability, Forgetting, and Failure Regimes
- Local-Energy Embedding for Critical Control in 3D Navier–Stokes: A Proof Program and Quantitative Criteria
- Material Signatures for Antineutrino-Based Detectability of Covert Fissile Production in Fusion Reactors
- Navier–Stokes Regularity via Critical Norm Tracking
- Noise-Biased Surface Code Thresholds Under Realistic Gate Sets
- Parity-Constrained Quantum Reservoir Computing for Image Classification: Formal Guarantees and Staged Simulation Evidence
- Parity-Locked Quantum Reservoir Computing for PCA-Encoded Image Classification: Robust Advantage, Entanglement Frontiers, and Operator–Dynamics Attribution
- Quantum Reservoir Computing Under Comparator Parity: Regime-Conditioned Advantage, Entanglement Effects, and Kernel-Null Boundaries
- Representation-Conditioned Synthesis for Multi-Objective Decision Support: Formal Comparability, Evidence Grading, and Boundary Diagnostics
- Simulability-Aware Quantum Reservoir Computing for Image Classification under Matched Readout Fairness
- Stability-Aware Bilevel Source Dataset Selection for Importance-Weighted Least Squares in Unsupervised Domain Adaptation
Patent
Agent-Based Neuro-Symbolic Methodologies for Scientific Discoveries and Workflow Automation
Patent pending
- US 2026/0212154 A1
- 19/032,587
- 21 January 2025
- 23 July 2026
- Marius-Constantin Dinu; Claudiu Leoveanu-Condrei
- ExtensityAI FlexCo, Wels (AT)
Talks
-
Reinforcement Learning — Where We Are and What's Next
-
KI-Startup-Founder über OpenAI, DeepSeek und Artificial General Intelligence
-
Extensity AI — The tool that automates research work
Affiliations & community
- AI Austria RL Community — team member; co-organiser of the International Reinforcement Learning Bootcamp (2nd edition, 2025). ↗
- Distinguished Expert, Global Advisory Council of the Indian Society of Artificial Intelligence and Law (ISAIL-GAC). ↗
- Founder of the "AI Is All You Need" research community on Discord, 450+ members; research outreach reaching 250,000+ views across X and LinkedIn. ↗
- Open-source maintainer: SymbolicAI, and Lighter (dependency injection for PyTorch).