Independent Researcher / Ouroboros Project
Wisdom Science studies how agents, embodied systems, and decision architectures improve through feedback, perturbation, evidence gates, recovery, and long-horizon experience. The papers, code, demos, and evidence records below show how each claim can be examined.
Problem Entry Points
These pages connect the research to questions people already ask: how to build reliable AI agents, ground hallucination claims, evaluate benchmarks, block unproven trades, and govern evidence in robotics.
A research map for ReflexBench, WisdomBench, proof-carrying action, and cognitive immunity, with runnable checks and the limits of each result.
A practical guide to reliable AI agents, deterministic fallbacks, refusal boundaries, evidence gates, and proof before autonomous action.
A grounded explanation of AI hallucination, self-certification failure, external evidence, unresolved states, and why confident answers are not enough.
A practical introduction to WisdomBench, a benchmark line for failure learning, recovery, repeated failures, and claim-bounded evaluation.
A bounded explanation of AI trading risk controls, no-proof no-trade gates, trace requirements, and why this is not financial advice.
A grounded page on robot learning from physical failure, embodied AI evidence, local proxies, and the boundary between simulation, shadow evidence, and real robot deployment.
A reader guide for moving from an AI claim to DOI records, GitHub source, Hugging Face mirrors, registries, boundaries, and counterexample routes.
Research Access
Use Zenodo for DOI records and fixed versions, GitHub for code and research discussion, Hugging Face for technical demos and registries, and this website for the connected research map and Chinese explanations.
Each resource states what can be checked and what remains unproven. Operational data and unpublished methods are not required to evaluate the claims made on these pages.
Find DOI records, fixed versions, artifact manifests, and citable research objects.
Inspect schemas and reproducible indices, report a counterexample, or discuss a concrete research question.
Open static demos, registries, and artifact pages without the surrounding editorial material.
Read Chinese introductions, long-form explanations, weekly updates, and links back to the underlying evidence.
Open Research
Useful criticism identifies the exact claim, the missing evidence, and the result that would change the conclusion. Corrections to a DOI, hash, manifest, baseline, or failed reproduction are recorded so later readers can follow what changed.
Public Archive
The public archive points to citable records, evidence maps, and challenge routes. GitHub indexes reviewable metadata; Zenodo remains the canonical file archive.
Systems Layer
The engineering layer organizes evidence trails, failure logs, workflow memory, social calibration, cognitive immunity, and closed-loop teaching into a practical system for research, operations, and reliable agent design.
Position
The next route is not only larger models. It is architecture, feedback, evidence, recovery, body-like subsystems, and the discipline to know when not to act.
This is a research archive, a systems map, and a public instrument panel for work that must remain falsifiable.
Contact
Independent Researcher, Ouroboros Project