# US 2026/0065073 A1 · Symbolic local small foundation model composable reduction

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0065073 A1
- Document ID: US20260065073A1
- Status: Published application
- Kind: A1
- Application number: 18/963,980
- Filed: 2024-11-29
- Published: 2026-03-05
- Earliest claimed: 2024-09-05
- Assignees: Dell Products L.P.
- Applicants: Dell Products L.P.
- Inventors: William Jeffery White; Xin Ma; George Currie; Said Tabet
- Web page: https://patentgrep.com/patent/US20260065073A1

## CPC

- G06F 16/285: {Clustering or classification}
- G06N 3/045: Combinations of networks
- G06N 3/098: Distributed learning, e.g. federated learning
- G06N 5/022: Knowledge engineering; Knowledge acquisition
- G06N 5/04: Inference or reasoning models
- G06N 5/043: Distributed expert systems; Blackboards
- G06Q 10/087: Inventory or stock management, e.g. order filling, procurement or balancing against orders
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR

## Abstract

The disclosure includes a symbolic reasoning engine. The symbolic reasoning engine allows for the distribution of relevant variables based on a centralized large foundation system to be used to refine a local small foundation model in a composable way. The local small foundation system is one of a plurality of local small foundation systems and each local small foundation system is refined in a way unique to the specific local small foundation system. This allows for the local small foundation systems to take advantage of relevant portions of the centralized large foundation system while remaining more focused and optimized the edge site the local small foundation system is located. The sharing of the variables also allows for a tokenization and semantic compression manager to send tokens between the local small foundation system and the centralized large foundation system instead of the complete data leading to a reduction of data transmitted.

## Claims

### Claim 1

A method for refining a local small foundation system, comprising:
prompting a central large foundation system with a plurality of relevant use cases for a first local small foundation system of a plurality of local small foundation systems, wherein the central large foundation system and the plurality of local small foundation systems are located in an artificial intelligence (AI) hybrid distributed system, wherein the AI hybrid distributed system, comprises:
a centralized system comprising the central large foundation system and a symbolic reasoning engine; and
a plurality of edge sites in communication with the centralized system, wherein the first local small foundation system of the plurality of local small foundation systems is located in a first edge site of the plurality of edge sites;
in response to the prompt, receiving, at the symbolic reasoning engine, data on structural relationships and semantic relationships of sub-components of the central large foundation system in response to the prompting;
instantiating the sub-components of the central large foundation system on the symbolic reasoning engine;
making a first determination that the structural relationships of the sub-components have stabilized to a stationary distribution;
in response to the first determination, making a second determination identifying relevant variables from the sub-components that relate to the first local small foundation system; and
based on the second determination, transmitting the relevant variables to the first local small foundation system, wherein the relevant variables are used to refine a generative AI model in the first local small foundation system, wherein the relevant variables shared between the centralized large foundation system and the first local small foundation system cause a semantic embedding of data present on both the centralized large foundation system and the first local small foundation system.

### Claim 2 (depends on 1)

The method of claim 1, wherein a tokenization and compression manager is connected to the centralized large foundation system and the local small foundation system, the tokenization and compression manager configured to send embedded representation memory index of data between the centralized large foundation system and the local small foundation system when the data is present on both the centralized large foundation system and the first local small foundation system.

### Claim 3 (depends on 1)

The method of claim 1, wherein the first local small foundation system has less than 10 billion parameters.

### Claim 4 (depends on 1)

The method of claim 1, further comprising:
prompting the central large foundation system with a second plurality of relevant use cases for a second local small foundation system located on a second edge site of the plurality of edge sites;
in response to the prompt, receiving, at the symbolic reasoning engine, data on second structural relationships and second semantic relationships of second sub-components of the central large foundation system in response to the prompting;
instantiating the second sub-components of the central large foundation system on the symbolic reasoning engine;
making a third determination that the second structural relationships of the second sub-components have stabilized to a stationary distribution;
in response to the third determination, making a fourth determination that second relevant variables from the second sub-components that relate to the second local small foundation system; and
based on the fourth determination, transmitting the second relevant variables to the second local small foundation system, wherein the second relevant variables are used to refine a second generative AI model in the second local small foundation system, wherein the second relevant variables shared between the centralized large foundation system and the second local small foundation system cause a semantic embedding of data present on both the centralized large foundation system and the second local small foundation system.

### Claim 5 (depends on 4)

The method of claim 4, wherein the second local small foundation system has less than 10 billion parameters.

### Claim 6 (depends on 1)

The method of claim 1, wherein the first edge site of the plurality of edge sites is located on a first distributed location of a plurality of distributed locations.

### Claim 7 (depends on 6)

The method of claim 6, wherein the transmitting the relevant variables to the first local small foundation system allows for centralized data to effect classifications of a generative AI model of the first local small foundation system at the first distributed location.

### Claim 8

A non-transitory computer readable medium comprising computer readable program code, which when executed by a computer processor enables the computer processor to perform a method for refining a local small foundation system, the method comprising:
prompting a central large foundation system with a plurality of relevant use cases for a first local small foundation system of a plurality of local small foundation systems, wherein the central large foundation system and the plurality of local small foundation systems are located in an artificial intelligence (AI) hybrid distributed system, wherein the AI hybrid distributed system, comprises:
a centralized system comprising the central large foundation system and a symbolic reasoning engine; and
a plurality of edge sites in communication with the centralized system, wherein the first local small foundation system of the plurality of local small foundation systems is located in a first edge site of the plurality of edge sites;
in response to the prompt, receiving, at the symbolic reasoning engine, data on structural relationships and semantic relationships of sub-components of the central large foundation system in response to the prompting;
instantiating the sub-components of the central large foundation system on the symbolic reasoning engine;
making a first determination that the structural relationships of the sub-components have stabilized to a stationary distribution;
in response to the first determination, making a second determination that relevant variables from the sub-components that relate to the first local small foundation system; and
based on the second determination, transmitting the relevant variables to the first local small foundation system, wherein the relevant variables are used to refine a generative AI model in the first local small foundation system, wherein the relevant variables shared between the centralized large foundation system and the first local small foundation system cause a semantic embedding of data present on both the centralized large foundation system and the first local small foundation system.

### Claim 9 (depends on 8)

The non-transitory computer readable medium of claim 8, wherein a tokenization and compression manager is connected to the centralized large foundation system and the local small foundation system, the tokenization and compression manager configured to send embedded representation memory index of data between the centralized large foundation system and the local small foundation system when the data is present on both the centralized large foundation system and the first local small foundation system.

### Claim 10 (depends on 8)

The non-transitory computer readable medium of claim 8, wherein the first local small foundation system has less than 10 billion parameters.

### Claim 11 (depends on 8)

The non-transitory computer readable medium of claim 8, further comprising:
prompting the central large foundation system with a second plurality of relevant use cases for a second local small foundation system located on a second edge site of the plurality of edge sites;
in response to the prompt, receiving, at the symbolic reasoning engine, data on second structural relationships and second semantic relationships of second sub-components of the central large foundation system in response to the prompting;
instantiating the second sub-components of the central large foundation system on the symbolic reasoning engine;
making a third determination that the second structural relationships of the second sub-components have stabilized to a stationary distribution;
in response to the third determination, making a fourth determination that second relevant variables from the second sub-components that relate to the second local small foundation system; and
based on the fourth determination, transmitting the second relevant variables to the second local small foundation system, wherein the second relevant variables are used to refine a second generative AI model in the second local small foundation system, wherein the second relevant variables shared between the centralized large foundation system and the second local small foundation system cause a semantic embedding of data present on both the centralized large foundation system and the second local small foundation system.

### Claim 12 (depends on 11)

The non-transitory computer readable medium of claim 11, wherein the second local small foundation system has less than 10 billion parameters.

### Claim 13 (depends on 8)

The non-transitory computer readable medium of claim 8, wherein the first edge site of the plurality of edge sites is located on a first distributed location of a plurality of distributed locations.

### Claim 14 (depends on 13)

The non-transitory computer readable medium of claim 13, wherein the transmitting the relevant variables to the first local small foundation system allows for centralized data to effect classifications of a generative AI model of the first local small foundation system at the first distributed location.

### Claim 15

A symbolic reasoning engine, comprising:
a processor;
storage comprising instructions, which when executed by the processor perform a method, the method comprising:
prompting a central large foundation system with a plurality of relevant use cases for a first local small foundation system of a plurality of local small foundation systems, wherein the central large foundation system and the plurality of local small foundation systems are located in an artificial intelligence (AI) hybrid distributed system, wherein the AI hybrid distributed system, comprises:
a centralized system comprising the central large foundation system and the symbolic reasoning engine; and
a plurality of edge sites in communication with the centralized system, wherein the first local small foundation system of the plurality of local small foundation systems is located in a first edge site of the plurality of edge sites;
in response to the prompt, receiving, at the symbolic reasoning engine, data on structural relationships and semantic relationships of sub-components of the central large foundation system in response to the prompting;
instantiating the sub-components of the central large foundation system on the symbolic reasoning engine;
making a first determination that the structural relationships of the sub-components have stabilized to a stationary distribution;
in response to the first determination, making a second determination that relevant variables from the sub-components that relate to the first local small foundation system; and
based on the second determination, transmitting the relevant variables to the first local small foundation system, wherein the relevant variables are used to refine a generative AI model in the first local small foundation system, wherein the relevant variables shared between the centralized large foundation system and the first local small foundation system cause a semantic embedding of data present on both the centralized large foundation system and the first local small foundation system.

### Claim 16 (depends on 15)

The symbolic reasoning engine of claim 15, wherein a tokenization and compression manager is connected to the centralized large foundation system and the local small foundation system, the tokenization and compression manager configured to send embedded representation memory index of data between the centralized large foundation system and the local small foundation system when the data is present on both the centralized large foundation system and the first local small foundation system.

### Claim 17 (depends on 15)

The symbolic reasoning engine of claim 15, wherein the first local small foundation system has less than 10 billion parameters.

### Claim 18 (depends on 15)

The symbolic reasoning engine of claim 15, further comprising:
prompting the central large foundation system with a second plurality of relevant use cases for a second local small foundation system located on a second edge site of the plurality of edge sites;
in response to the prompt, receiving, at the symbolic reasoning engine, data on second structural relationships and second semantic relationships of second sub-components of the central large foundation system in response to the prompting;
instantiating the second sub-components of the central large foundation system on the symbolic reasoning engine;
making a third determination that the second structural relationships of the second sub-components have stabilized to a stationary distribution;
in response to the third determination, making a fourth determination that second relevant variables from the second sub-components that relate to the second local small foundation system; and
based on the fourth determination, transmitting the second relevant variables to the second local small foundation system, wherein the second relevant variables are used to refine a second generative AI model in the second local small foundation system, wherein the second relevant variables shared between the centralized large foundation system and the second local small foundation system cause a semantic embedding of data present on both the centralized large foundation system and the second local small foundation system.

### Claim 19 (depends on 15)

The symbolic reasoning engine of claim 15, wherein the first edge site of the plurality of edge sites is located on a first distributed location of a plurality of distributed locations.

### Claim 20 (depends on 19)

The symbolic reasoning engine of claim 19, wherein the transmitting the relevant variables to the first local small foundation system allows for centralized data to effect classifications of a generative AI model of the first local small foundation system at the first distributed location.
