Back to results
US 12,591,736 B2Granted
Fine-tuned large language models for capability controller
AssigneeSap SE
InventorSrinivasa Byaiah Ramachandra Reddy
Life of this patent
Estimating the term…
Abstract
In an example embodiment, capability definitions are leveraged to fine-tune an LLM, producing a fine-tuned LLM that is specifically designed to handle capability definitions, without the need to provide such capability definitions as input tokens to an LLM. This fine-tuned LLM has a deeper understanding of the capabilities of a system than generic LLMs.
Claims20
A system comprising:
- at least one hardware processor;
- and
- a computer-readable medium storing instructions that, when executed by the at least one hardware processor, cause the at least one hardware processor to perform operations comprising:
- accessing one or more capability definitions of services offered by a software architecture, the one or more capability definitions stored in a repository of the software architecture and each capability definition containing information about one or more capabilities of a corresponding service offered by the software architecture;
- automatically generating an instruction dataset for each of the one or more capability definitions;
- fine-tuning a large language model (LLM), located outside of the software architecture, to produce a fine-tuned special-purpose LLM, based on the instruction datasets;
- accessing conversation information regarding a conversation between a first user and an application of the system;
- sending the conversation information along with a prompt to the fine-tuned special-purpose LLM, the prompt instructing the fine-tuned special-purpose LLM to identify one or more capability definitions matching an intent of the first user as reflected in the conversation;
- receiving, from the fine-tuned special-purpose LLM, one or more matching capability definitions;
- and
- causing display, to the first user, of the one or more matching capability definitions.
8 more elements
A method comprising:
- accessing one or more capability definitions of services offered by a software architecture, the one or more capability definitions stored in a repository of the software architecture and each capability definition containing information about one or more capabilities of a corresponding service offered by the software architecture;
- automatically generating an instruction dataset for each of the one or more capability definitions;
- fine-tuning a large language model (LLM), located outside of the software architecture, to produce a fine-tuned special-purpose LLM, based on the instruction datasets;
- accessing conversation information regarding a conversation between a first user and an application of the system;
- sending the conversation information along with a prompt to the fine-tuned special-purpose LLM, the prompt instructing the fine-tuned special-purpose LLM to identify one or more capability definitions matching an intent of the first user as reflected in the conversation;
- receiving, from the fine-tuned special-purpose LLM, one or more matching capability definitions;
- and
- causing display, to the first user, of the one or more matching capability definitions.
5 more elements
A non-transitory machine-readable medium storing instructions which, when executed by one or more processors, cause the one or more processors to perform operations comprising:
- accessing one or more capability definitions of services offered by a software architecture, the one or more capability definitions stored in a repository of the software architecture and each capability definition containing information about one or more capabilities of a corresponding service offered by the software architecture;
- automatically generating an instruction dataset for each of the one or more capability definitions;
- fine-tuning a large language model (LLM), located outside of the software architecture, to produce a fine-tuned special-purpose LLM, based on the instruction datasets;
- accessing conversation information regarding a conversation between a first user and an application of the system;
- sending the conversation information along with a prompt to the fine-tuned special-purpose LLM, the prompt instructing the fine-tuned special-purpose LLM to identify one or more capability definitions matching an intent of the first user as reflected in the conversation;
- receiving, from the fine-tuned special-purpose LLM, one or more matching capability definitions;
- and
- causing display, to the first user, of the one or more matching capability definitions.
5 more elements
Dependencies are inferred from the claim text.