Update README.md

README.md

@@ -47,14 +47,18 @@ metrics:
 
 
 # MarCognity-AI  
-**The framework that teaches artificial intelligence how to think.**
+**A research framework for reflective and epistemically transparent AI systems**
 ---
 
 ## Table of Contents
 
 - [What is MarCognity-AI](#what-is-marcognity-ai)
 - [Origins](#origins)
+- [Community Recognition](#community-recognition)
 - [Vision](#vision)
+- [Limitations](#limitations)
+- [Research Status](#research-status)
+- [Observed Fracture](#observed-fracture)
 - [Why Use It](#why-use-it)
 - [Core Capabilities](#core-capabilities)
 - [Usage Examples](#usage-examples)
@@ -90,7 +94,7 @@ MarCognity wasn’t born in a lab.
 It was born from a curious, free, and determined mind.
 
 ---
-## Community Recognition
+## Early Community Interactions (Non-Endorsement)
 
 MarCognity-AI has already sparked resonance across major AI communities.
 
@@ -130,40 +134,49 @@ It integrates semantic memory, ethical auditing, and scientific validation to tr
 
 The official version of the code and the full research paper have been permanently archived on Zenodo and are citable using their Digital Object Identifier (DOI).
 
-| **MarCognity-AI** | [![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.17855185.svg)](https://doi.org/10.5281/zenodo.17855185) |
+| **MarCognity-AI** | [![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.17855185.svg)](https://doi.org/10.5281/zenodo.18435049) |
 |---|---|
-| **Permanent DOI** | `10.5281/zenodo.17855185` |
-| **Access Publication** | [Full Research Paper (PDF) & Code (Zenodo)](https://doi.org/10.5281/zenodo.17855185) |
+| **Permanent DOI** | `https://doi.org/10.5281/zenodo.18435049` |
+| **Access Publication** | [Full Research Paper (PDF) & Code (Zenodo)](https://doi.org/10.5281/zenodo.18435049) |
 
 
 
 ---
-
 ## Limitations
 
-LLM‑based metacognition collapses at the same fault line: 
-it confuses linguistic coherence with epistemic awareness.
+LLM‑based metacognition collapses at the same fault line: it confuses linguistic coherence with epistemic awareness.
 The model can say “this answer is unclear,” but not “I don’t know if what I’m saying is true.
 
 ---
-
 ## Research Status
-
-MarCognity-AI is an exploratory research framework. It is not a production-ready system. 
-Its purpose is to expose and study structural limits of LLM-based metacognition, 
+MarCognity-AI is an exploratory research framework.
+It is not a production-ready system.
+Its purpose is to expose and study structural limits of LLM-based metacognition,
 particularly the collapse between linguistic coherence and epistemic awareness.
+This line of investigation is ongoing.
 
 ---
 ## Observed Fracture
+
 During the development of MarCognity-AI, a recurring failure mode emerged:
 
-LLM-based metacognitive layers reliably optimize for linguistic coherence but fail to surface epistemic uncertainty as an explicit signal.
+LLM-based metacognitive layers reliably optimize for linguistic coherence
+but fail to surface epistemic uncertainty as an explicit signal.
 
-The system can evaluate how an answer is written, yet cannot account for whether what is being said is actually known, verifiable, or admissible.
+The system can evaluate how an answer is written,
+yet cannot account for whether what is being said is actually known,
+verifiable, or admissible.
 
-This collapse between coherence and awareness is not treated here as a bug to be fixed, but as a structural fracture to be studied.
+This collapse between coherence and awareness
+is not treated here as a bug to be fixed,
+but as a structural fracture to be studied.
 
 ---
+## Note for Readers
+The demo and cognitive journal in this repository are meant to expose a reproducible failure mode, not a solved system.
+
+---
+
 
 ## Why Use It
 
@@ -181,6 +194,9 @@ MarCognity isn’t just intelligent. It’s **aware**.
 ### 4. Reflective Cognitive Journal  
 Each response is accompanied by a detailed **metacognitive reflection**, saved in Markdown for analysis and reuse.
 
+### 5. Epistemic Transparency
+The Skeptical Agent exposes unsupported claims through a structured verification report, making epistemic uncertainty explicit.
+
 ---
 
 ## Core Capabilities
@@ -193,6 +209,7 @@ Each response is accompanied by a detailed **metacognitive reflection**, saved i
 - Version archiving in FAISS memory  
 - Ethical risk and linguistic bias analysis  
 - Reflective cognitive journal with Markdown export
+- Epistemic Verification Layer (Skeptical Agent): decomposes responses into claims and checks them against sources
 
 ---
 
@@ -210,6 +227,18 @@ Each response is accompanied by a detailed **metacognitive reflection**, saved i
 **Input:** “Compare the view of consciousness in philosophy and neuroscience.”  
 **Output:** Structured response + sources + cognitive visualization + reflective journal
 
+### Epistemic Verification Example
+Input: “Explain quantum entanglement.”
+Output:
+
+Generated response
+
+Claim-by-claim verification
+
+VERIFIED / EPISTEMIC FAILURE report
+
+Reasoning based on provided sources
+
 ---
 
 ## Try It Now  
@@ -248,14 +277,16 @@ It shows step-by-step how the agent generates, evaluates, and reflects on a resp
 
 ## Cognitive Architecture
 
-| Module                     | Function                                                  |
-|----------------------------|-----------------------------------------------------------|
-| Problem Classification     | Automatic recognition of input type                       |
-| Academic Prompting         | Structuring complex multidisciplinary queries             |
-| Scientific Retrieval       | Asynchronous querying of multiple open-access sources     |
-| Semantic Evaluation        | Response analysis with logical scoring                    |
-| FAISS Memory               | Archiving and comparison with past responses              |
-| Cognitive Visualization    | Uses HuggingFace’s transformers library and selected models for scientific content processing, ethical analysis, and cognitive representation |
+| Module                 | Function                                                                 |
+|------------------------|---------------------------------------------------------------------------|
+| Problem Classification | Automatic recognition of input type                                       |
+| Academic Prompting     | Structuring complex multidisciplinary queries                             |
+| Scientific Retrieval   | Asynchronous querying of multiple open‑access sources                     |
+| Semantic Evaluation    | Response analysis with logical and semantic scoring                       |
+| Skeptical Agent        | Sentence‑level claim verification against provided sources; flags unsupported statements and produces an epistemic report |
+| FAISS Memory           | Archiving and comparison with past responses, including reflective evaluations |
+| Cognitive Visualization| Scientific content processing, ethical analysis, and conceptual representation using selected transformer models |
+
 
 ---
 
@@ -282,4 +313,4 @@ It is a threshold.
 An invitation to rethink how artificial intelligence can reflect, improve, and act with awareness.
 This project is released under the Apache 2.0
 
-> _“Every response is a threshold. Every reflection, an act of agency.”_
+> _“Every response is a threshold. Every reflection, an act of agency.”_