Gen AI Output Verification for Hallucination Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large language models (LLMs) and other generative AI systems often produce output results that include hallucinations or factual inaccuracies, which can lead to misinformation and loss of user trust.
Innovation Solution
Implementing a set-based comparison technique to quantify accuracy by comparing AI-generated outputs with a context-specific data set or source of truth, using similarity measures and thresholds to filter out inaccuracies and ensure factual correctness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If Gen AI models produce output based on input query and context data, then the model provides responsive and context-aware answers, but the output includes hallucinations and factual inaccuracies
Solution Approach 1:
The patent introduces an intermediary verification system that acts as a mediator between the Gen AI model output and the final response. This verification system compares the generated output against the original context data to identify and filter hallucinations, allowing the system to maintain both context-awareness and factual accuracy simultaneously.
Solution Approach 2:
The patent implements a feedback mechanism where the output of the Gen AI model is fed back into a verification process that checks for consistency with the context data. This feedback loop identifies discrepancies and hallucinations, enabling the system to correct inaccuracies while preserving the beneficial context-aware responses.
2Reliability
If the system implements verification to reduce hallucinations, then factual accuracy improves, but computational overhead increases
Solution Approach 1:
The patent applies partial verification by focusing computational resources on verifying only the critical factual claims in the Gen AI output rather than performing exhaustive verification of the entire response. This selective approach maintains high factual accuracy for important information while reducing overall computational overhead.
Solution Approach 2:
The patent performs preliminary filtering of the context data and output generation before verification, organizing information in a way that facilitates more efficient comparison and validation. This preliminary structuring reduces the computational complexity of the subsequent verification process.
Data Source
AI summary
Methods, systems, and computer program products that address inaccuracies in generative artificial intelligence (Gen AI) output. Some implementations involve identifying hallucinations, e.g., identifying circumstances in which differences between Gen AI output and a source of truth are greater than an acceptable threshold. Some implementations identify hallucinations and other inaccuracies using a set-based comparison technique that quantifies or otherwise measures accuracy based on similarity to a known source of truth. Some implementations enable the filtering of hallucinations and other inaccuracies in Gen AI (e.g., LLM) outputs. Some implementations enable such identification and/or filtering of Gen AI output by utilizing predetermined or customizable accuracy and/or confidence thresholds. A variety of post-comparison actions may be initiated based on identifying and/or filtering Gen AI output inaccuracies.