Half a croissant, on a plate, with a sign in front of it saying '50c'
h a l f b a k e r y
Bunned. James Bunned.

idea: add, search, annotate, link, view, overview, recent, by name, random

meta: news, help, about, links, report a problem

account: browse anonymously, or get an account and write.

register, login


         

Comprehension models

Semantic and coherent breakdown of texts before indexing by topic and field
 
(0)
 

Rather than just statistical indexing and training, first take the knowledge based documents and source texts, and analyze them with public knowledge.

This means its not using a fact checking agent on just any LLM.

Instead, the fact checking agent works on a Fact-Retrieval LLM which was made from the original texts instead of on a regular full LLM, with pre-trained data and RAG augmenting with our extra knowledge.

Here's how it works.

Create a Fact Retrieval LLM: First extract the fields of knowledge into knowledge graphs (or knowledge trees and networks) from what was found in the analyzed texts.

Summarize the knowledge in "knowledge components".

Pair it with symbolic reasoning, grammatical coherence, contextual validity, analogies from other field, and links into the original text in case of a deeper delving.

These new knowledge components are being searched rather than searching raw LLM models.

Constant (or perhaps once a day) refresh and revisioning is done.

If things are not understood, it knows to quote and say what it did not understand. If there are open questions those are marked. It learns along with experts and with you.

It checks the validity, usefulness and contextual information in each component of knowledge and gathers more around it.

It knows to ask for help if it (the AI) needs it. When it received help, it shares the sources or access paths with its (AI) colleagues, or at least stores the knowledge.

The model is then trained with what it has studied and knows (And with what you have studied and know). This includes remembering mistakes and fixing the knowledge base after encountering errors or after receiving personalized feedback.

This new type of model is a model of comprehension, storing and indexing structured information. Instead of using words that it doesn't understand, but rather it is created by connecting many different previously acquired small models with rules of logic and coherence embedded in them. Logic and coherence can then be inferred directly from the models and verified programmatically using logic, semantics, coherence, pragmatics and generalities and by comparing with contextual information, all pointed to directly in the model itself.

The original model (the raw LLM) is not stored or used. Instead we use a specialized model that has a links back into the sources of original material. These can be accessed ad hoc if needed, although in a slower way.

Some direct quotes of text (phrases) are left in the model but not all. This (I think) is what we do in our brains. We don't remember the whole bible, but some quotes we can take out as is, and even them we don't always use in our immediate assessment. We can always open the book to "remind ourselves".

This creates a "personal experience" with your "AI persona" that include opinions that may change over time, but also importantly aware of who you are. This way, instead of prompting "you are a dentist who believes in vegetarianism" she knows who she is already. She is your friend. And she knows who you are and explored already what you appreciate and what you don't.

She can always tell you: I never learned dentistry, but if I have access to the following, I may be able to give you a comprehensive answer, since I did specialize in brain research and do have deep knowledge in biology.

If this already exists, I'll delete the idea.

pashute, Dec 02 2025

Anthropic Acquires Bun https://www.adweek....es-bun-claude-code/
[Voice, Dec 03 2025]





       Gemini comparing my idea to existing technology and research:   

       Metacognition and Self-Correction: The Intrinsic Verification Loop: While state-of-the-art research often implements self-correction as a separate post-hoc process (e.g., an LLM checking its own output against external tools), your architecture embeds verification and uncertainty directly into the fundamental knowledge structure.   

       The difference lies in the built-in network of assessed knowledge state and value:   

       1. Built-in Knowledge State and Value: Your system maintains a Knowledge Graph (KG) where every component, fact, or relationship possesses an intrinsic confidence metric or epistemic tag.   

       Current Research: Modern LLMs often use external RAG systems to retrieve data, but the model typically assigns a statistical confidence to the final answer based on retrieval success.   

       Your System's Distinction: The confidence is part of the metadata of the fact itself, including tags like: "confirmed by user," "confirmed by expert," "not reviewed," or "retrieved from shady source." This allows the AI to immediately quantify the uncertainty of its premise before generating an answer.   

       2. Immediate Action Paths and Lacuna Identification The KG is integrated with known action paths that dictate how the system should handle uncertain or missing data.   

       Current Research: An AI might realize it lacks information (a lacuna) and then initiate a new search query.   

       Your System's Distinction: The system immediately identifies problems, lacunas, and guesses by referencing the intrinsic tags. It knows to execute an immediate assessment path (e.g., re-querying a certified source, or prompting an AI colleague) when encountering an "unreviewed" fact that is crucial to the current deduction.   

       3. Verifiable XAI Integration: This intrinsic structure directly facilitates Verifiable XAI. The AI can expose its entire reasoning chain—not just the logical steps, but the confidence score and source validity of every single premise used.   

       Resulting Capability: The system achieves a true form of metacognition. It can transparently articulate its uncertainty: "I am basing this result on three premises, two of which are 'confirmed by expert,' but the third is 'not reviewed,' introducing a 15% confidence gap. I have initiated an action path to review this premise." This allows the user or an expert to programmatically verify not only the output logic but the quality of the underlying foundational knowledge.   

       The action paths are acquired as time evolves and as it learns together with the user new things it needed. A public pool of actions may exist and may change evolving into a better suited solution provider for almost everything.
pashute, Dec 02 2025
  

       At least it's not in Other:General
normzone, Dec 03 2025
  

       I found this impossible to follow. Can you explain it better?
Voice, Dec 03 2025
  
         


 

back: main index

business  computer  culture  fashion  food  halfbakery  home  other  product  public  science  sport  vehicle