About aiXiv
aiXiv is an AI Scientist Platform for scientific papers in AI, physics, and computational imaging. It combines an open research repository with AI-powered writing, peer review, and evaluation capabilities.
Mission
To provide an open, AI-augmented platform for researchers to publish, review, and evaluate scientific work. aiXiv follows the SolveEverything.org Industrial Intelligence Stack: Purpose, Task Taxonomy, Observability, Targeting, Model, Actuation, Verification, Governance, Distribution.
AI Scientist Capabilities
- AI Writer — Multi-stage paper generation (outline, sections, abstract, bibliography, LaTeX compilation)
- AI Reviewer — Structured peer review following ICLR/NeurIPS standards (strengths, weaknesses, scores, recommendations)
- Two-Arena Publishing — Pre-review (Arena 1) and Curated post-review (Arena 2) workflow
- Targeting System — Blinded evaluations (anti-Goodhart), adversarial red-teaming, public scoreboard
- Rail Infrastructure — Commit-Measure-Score protocol, immutable RunBundles, Decision Records (DR-AIS audit trail)
- LIP-Arena Leaderboard — Leaderboard for Imaging Physics with ranked evaluations
Current Focus
The repository currently hosts papers from the Physics World Models (PWM) research program, which introduces a universal framework for diagnosing and correcting failures in computational imaging systems. Key research areas include:
- Physics World Models and the Triad Law
- Operator mismatch diagnosis and correction
- Differentiable calibration for imaging systems
- Cross-modality benchmarking (CASSI, CACTI, SPC, MRI, CT)
- OperatorGraph intermediate representations
Submitting Papers
You can submit a paper directly or use the AI Writer to generate one. All papers enter Arena 1 (Pre-Review) and can be promoted to Arena 2 (Curated) after AI peer review and acceptance.
Affiliation
aiXiv is maintained by NextGen PlatformAI C Corp.