Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Using GPT to organize and access information, and generate questions. Long term goal is to make an agent-like research assistant.
| Date | Stars |
|---|---|
| 2026-07-31 | 692 |
| 2026-08-02 | 692 |
| 2026-08-06 | 692 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# A knowledge graph from GPT
## High-level description
This program is meant to create an external memory module for a language model, and ultimately provide
agent-like capabilities to a language model (long-term goal).
- The combined system would ideally gather information via a text interface, categorize and structure it, then identify gaps in knowledge,
or inconsistencies.
- The language model can be shown subsets of this information (chosen based on the structure of
the knowledge graph), and then choose to propose further questions to ask the environment as clarification,
building knowledge over time.
### Goals
This project aims to address a few of the major shortcomings of language models:
* Memory
* External memory solves the problem of lack of long-term learning from one or two examples.
* Logic
* Language models are generally unable to force structured responses,
but recalling logical arguments from memory could help.
* Interpretability
* By observing what memories are accessed, we can understand sources of resulting statements and information flow.
* Alternatively, by observing how the language model processes and categorizes information, we can understand the
inherent structure of the information learned by the raw language model.
* Developing agency
* Language models lack coherent agency, as language models generate both sides of a conversation.
* Structuring the language model as a component in a reinforcement learning system, with the goal
of categorizing and uncovering information, restores agency.
* Computational resource use for training
* Can we continuously improve the entire machine learning system (model and RL wrapper), without
continuously retraining the model parameters? Simply by recursively improving the memory (which is re-inserted
through the prompt).
* Language models could be trained for specialized sub-component tasks in the resulting global system.
* Bootstrapping capabilities
* Is a minimal level of reason and analogy sufficient to tackle arbitrarily complex processing of knowledge,
by breaking down ideas into minimal components, and treating each separately?
* There are probably opportunities here for bootstrapping and policy improvement of the language model,
through self-generated examples (as used below in extracting question embeddings, and
generating questions from examples of clusters of questions).
### Target uses long term
1. Database generation and parsing + question answering
* Summarize a research field or class notes or textbook
* Identify conflicting information and disputes,
or different explanations for the same topic or idea
2. Educational tool or personal learning tool
* Construct the agent to serve as a spaced-repetition flashcard assistant.
* Learn what knowledge the user has, and how quickly they forget it, then periodically reprompt them.
* Learn to suggest new information for the user to learn, tailored to their current knowledge and interests.
* Do everything with a flexible, natural language interface to pose questions and interpret responses.
* A structured eucational tool
* Fix the knowledge in the graph (distilled from experts), then use the knowledge structure and spaced-repetition
framework to understand a student's learning needs, and interface with them.
3. Hypothesis generation for scientific research
* Process entire scientific fields, including papers, textbooks, audio lectures, etc.
* Come up with novel ideas and research proposals
### Outline of the program
The program is designed as a wrapper for a language model in python.
The knowledge graph (stored in python) makes periodic calls to the language model when necessary.
A key feature is that the structure of the knowledge graph is fully-human interpretable.
Even the embeddings of information, and the facts themselves, are in natural language.
Moreover, all steps of the algorithm aExcerpt of 25,493 characters
Read on GitHub
12
4
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:f26a562c323ca4bc, name:knowledge graph