skills.rest
sovr610
sovr610COMMUNITY
Explore all skills by this author →
WebsiteGitHub

Dual-Process Reasoning (System 1/2 + Metacognitive Control)

Route System 1 predictions and System 2 refinement with deterministic metacognitive control.

Updated Feb 28, 2026
Search Tags:#routing#trace#calibration#determinism#interpretability#dual-process
One-click install
npx skills add https://github.com/sovr610/refffiy --skill dual-process-reasoning-system-1-2-metacognitive-control
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: Dual-Process Reasoning (System 1/2 + Metacognitive Control)
Source: https://github.com/sovr610/refffiy/tree/main/brain-ai-dev/skills/dual-process-reasoning
Command: npx skills add https://github.com/sovr610/refffiy --skill dual-process-reasoning-system-1-2-metacognitive-control

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires torch, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

Coordinate fast System 1 predictions and slow System 2 iterative refinement using a deterministic metacognitive router to ensure reproducible per-item routing and traceable reasoning across batches.

Core Features & Use Cases

  • System 1 fast predictor, System 2 iterative refinement, and deterministic routing that respects per-item independence.
  • Confidence calibration, novelty scoring, and a JSON-serializable ReasoningTrace/ReasoningOutput contract for interpretability and auditing.
  • Suitable for large-scale inference pipelines, model evaluation, and debugging/assessment of routing strategies with traceable provenance.

Quick Start

Instantiate a DualProcessFullConfig (e.g., .minimal()) and run DualProcessReasoner on a batch with return_details to obtain the final outputs and optional traces.

Frequently Asked Questions about Dual-Process Reasoning (System 1/2 + Metacognitive Control)

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I implement deterministic routing for fast and slow reasoning in inference pipelines?▼

It coordinates fast System 1 predictions and slow System 2 iterative refinement using a deterministic metacognitive router to ensure reproducible per-item routing and traceable reasoning across batches.

What is dual-process reasoning for large-scale model evaluation?▼

Dual-process reasoning for model evaluation separates fast System 1 predictions from slow System 2 iterative refinement, coordinated by a metacognitive router to provide confidence calibration, novelty scoring, and JSON-serializable reasoning traces.

How do I get JSON-serializable reasoning traces for interpretability and auditing?▼

You obtain JSON-serializable reasoning traces by running the configured reasoner on a batch with the return_details flag enabled, which outputs a structured ReasoningTrace and ReasoningOutput contract for interpretability and auditing.

Does this dual-process routing approach enforce per-item independence during batch processing?▼

Yes, the deterministic metacognitive routing explicitly enforces per-item independence and eliminates stochastic routing, ensuring reproducible inference and traceable provenance across all items in a batch.

Can I use this System 1 and System 2 router with PyTorch for model calibration?▼

Yes, it requires PyTorch as its dependency and supports model calibration. The modular configuration includes dedicated Calibration and System1Config components to execute confidence calibration and novelty scoring alongside routing.

What are the limitations of using a deterministic metacognitive router for inference?▼

The limitation of a deterministic metacognitive router is that it eliminates stochastic routing variations. This strict determinism, while ensuring reproducible traces, may limit exploratory routing strategies during model debugging and assessment.

Related Skills

Explore More Software Engineering Skills→
tenstorrenttenstorrent
341

ttnn

Convert PyTorch tensors and execute tensor operations on Tenstorrent MeshDevice topologies.

Official
Advanced
coreymaypraycoreymaypray
1

model-qa

Audit AI models with evaluation, calibration, and SHAP analysis.

Community
Advanced
silva2kandsilva2kand

agentic-image-loop

Automate iterative image generation, annotation, and refinement workflows.

Community
Intermediate
victorp1leonvictorp1leon

read-only-intent-delivery

Route read-only intents deterministically with traceable outcomes and non-mutation enforcement.

Community
Advanced