Chemical Reaction Protocol Generation Using Latent Similarity

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing chemical reaction protocols lack comprehensive and reproducible methods for generating detailed reaction protocols, leading to inefficiencies, safety concerns, and compromised reproducibility without a known protocol.

Innovation Solution

An AI system uses vector representations and encoders to process various input types of chemical reactions, creating a holistic vector embedding that queries a database to generate complete reaction protocols based on user input, incorporating mechanistic and electronic representations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If existing chemical reaction protocols are used without comprehensive generation methods, then simplicity is maintained, but reliability and reproducibility deteriorate

Engineering Contradiction:

Improvereproducibility of chemical reaction protocolsVSAvoidcomplexity of protocol generation system

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary system comprising encoders, vector embeddings, and LLMs that mediate between chemical reaction inputs and protocol generation. The encoder converts chemical reactions into vector representations, which then serve as intermediaries for the LLM to generate appropriate protocols, ensuring reliability without requiring direct complex rule-based systems.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces traditional mechanical rule-based protocol generation systems with an AI-driven approach using encoders and large language models. This substitution allows for more reliable and reproducible protocol generation by leveraging learned patterns from chemical literature rather than rigid pre-programmed rules.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If detailed reaction protocols are generated comprehensively, then reliability improves, but loss of time increases

Engineering Contradiction:

Improvecompleteness of reaction protocolVSAvoidtime for protocol generation

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent performs preliminary actions by pre-processing chemical reactions into vector embeddings using encoders before protocol generation. This preliminary conversion prepares the data in advance, allowing the LLM to generate complete protocols more efficiently without requiring time-consuming real-time analysis during the generation process.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent uses copying by retrieving similar existing protocols from a database using vector embeddings and copying their structures and formats. This allows the system to generate comprehensive protocols by adapting proven templates rather than creating entirely new ones, reducing generation time while maintaining completeness.

Inventive Principle:

Principle #26Copying

3Adaptability or versatility

If multiple input types are processed holistically, then adaptability improves, but device complexity increases

Engineering Contradiction:

Improveability to process various input typesVSAvoidcomplexity of processing system

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements universality by using a single encoder system that can process multiple input types (chemical reactions, mechanisms, outcomes) and convert them into unified vector representations. This multi-functional encoder handles diverse inputs through the same processing pipeline, improving adaptability without proportionally increasing complexity.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent merges different input processing streams into a unified vector embedding space. By combining electronic representations, mechanistic representations, and outcome data into a single holistic vector format, the system achieves versatile processing capability while consolidating complexity into one integrated representation layer rather than separate processing systems.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS20260057136A1Latent space similarity and LLM-based reductive and generative operations for generating chemical reaction protocols

Publication Date: 2026.02.26 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

Method and apparatus for receiving an electronic representation of a chemical reaction, receiving a textual prompt of a problem statement regarding at least one of a first reaction protocol used for the chemical reaction; and a reaction outcome of the chemical reaction; encoding the electronic representation into a first vector using a first encoder; converting the electronic representation of the chemical reaction into a mechanistic representation of a structural change of molecules of the chemical reaction; encoding the mechanistic representation into a second vector using a second encoder; generating a vector embedding using the first vector and the second vector; querying a database using the vector embedding to retrieve a second reaction protocol; and generating a response to the textual prompt using the second reaction protocol.