Multi-turn dialogue response generation with autoregressive transformer models

This invention describes a method for training artificial intelligence models, specifically those with a "sequence-to-sequence" architecture that includes an encoder and a decoder. The training process involves using random padding on the input data to improve the model's learning and reduce redundancy. The claims detail how to prepare the training data for the decoder by adding special start and end markers.

Why it matters: The rapid evolution of large language models and AI training methods since 2023 means that while the described sequence-to-sequence architecture and training techniques remain foundational, the specific improvements detailed might now be standard practice or superseded by more advanced approaches.

Status: Not in forceListed as no longer active. The specific reason is not in the record we hold.
How hard to build: SpecializedSoftware logic, machine learning model training

3 AI reinvention directions for this patent

See all three

AI gives you a few directions you could take this. Pick one, and we check whether your version is different enough to patent, then write the filing.

Reinvent this with AI

Where this one came from

Every morning we go through millions of abandoned patents and pick out 15 worth your time. This was one of today’s.

This one

+10