MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202611059835 A) filed by Indian Institute Of Technology Delhi on May 11, 2026, for A Method And A System For Fine-Tuning A Pre-Trained Language Model.
Inventor includes Sivasubramani, Santhosh.
The application for the patent was published on July 17, 2026, under issue no. 29/2026.
Abstract: Disclosed herein is a method for fine tuning a pre trained language model. The method includes receiving (602), by a processor, a pre trained language model and preparing (604) a training dataset comprising first instruction–response pairs representing valid application patterns of a domain specific technique and second instruction–response pairs representing invalid application patterns with corresponding corrective mappings. The method further includes encoding (606) one or more state variables as numerical conditioning inputs associated with the instruction–response pairs. Further, the method includes modifying (608) the pre trained language model using a parameter efficient adaptation technique by attaching one or more trainable low rank matrices to a plurality of layers and training the low rank matrices while keeping the weights of the pre trained language model fixed. Furthermore, the method includes generating (610) a fine tuned language model configured to generate responses based on learned mappings between the state variables and the instruction–response patterns.
Disclaimer: Curated by HT Syndication.
