YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
META-LOOP
Supervised Iterative Reasoning Protocol for AI
Standard V1.0 — RH/GPT Research
META-LOOP is a model-agnostic protocol designed to improve the robustness of AI problem solving through contextualization, exploration, verification, testing, comparison, adversarial contradiction, correction, selection, and supervised stopping.
It is not an AI model and it does not guarantee perfect answers. Its purpose is to make the reasoning process more systematic, auditable, and resistant to premature conclusions.
Core pipeline
CONTEXT → UNDERSTAND → EXPLORE → VERIFY → TEST → COMPARE → CONTRADICT → CORRECT → SELECT → SUPERVISE → STOP
Standard V1.0
- Context Loop 0
- Understanding Loop 1
- Exploration Loop 2
- Verification Loop 3
- Testing Loop 4
- Comparison Loop 5
- Contradiction Loop 6
- Correction Loop 7
- Selection Loop 8
- META-SUPERVISOR
Core principles
- Never invent missing information.
- Distinguish facts, estimates, assumptions, and opinions.
- Ask only for information that can materially affect the decision.
- Prefer fit to the actual need over unnecessary feature maximization.
- Consider existing equipment/resources before recommending additions.
- For important decisions, distinguish obligation, possibility, and strategic preference.
- Before stopping, identify the unresolved variable most likely to change the conclusion.
- Stop when further iteration is unlikely to materially improve the result.
Status
Standard V1.0 — frozen for independent benchmarking.
Future ideas should first be tested in the RH/GPT laboratory and should not be inserted into V1.0 merely because they sound useful.
Limitations
META-LOOP improves process discipline; it cannot guarantee truth. Its effectiveness depends on the underlying model, available information, tools, sources, and evaluation criteria.
For high-stakes legal, medical, financial, safety, or other professional decisions, qualified human review remains appropriate.
License
See LICENSE.
META-LOOP V1.0 is publicly available for independent testing.
Users are encouraged to test the protocol with different AI models and compare results with and without META-LOOP.
Useful observations include:
- Did META-LOOP identify missing information?
- Did it detect a false or incomplete premise?
- Did it identify relevant alternatives?
- Did it verify decisive facts?
- Did it challenge its leading solution?
- Did it change the final decision?
- Did it reduce unsupported assumptions or hallucinations?
- Was the final answer more robust?
- Was the stopping decision justified?
Community feedback is especially valuable when it includes a concrete test case or a comparison between a baseline response and a META-LOOP response.
META-LOOP V1.0 remains frozen for independent benchmarking. Experimental improvements should be tested separately before being considered for a future version.