QELM: Quantum-Enhanced Language Model

An experimental hybrid quantum/classical NLP framework exploring token encoding, parameterized quantum circuits, quantum attention-like blocks, measurement, and classical post-processing.

Research boundary

QELM is experimental research software. It explores hybrid quantum/classical language modeling without claiming to replace commercial LLMs or outperform validated production systems.

Architecture

How a token moves through the hybrid pipeline

The pipeline makes each stage visible: token preparation, circuit behavior, measurement, post-processing, and the limits of the current experiment.

How a token moves through QELM

Token encoding

Text is reduced into token maps and compact encodings before entering the quantum/classical path.

Token encoding

Parametric quantum circuits

Quantum attention-like blocks

Measurement and post-processing

Training gradients

Feature matrix

Experimental capabilities and public repo facts

Multi-block quantum transformer architecture

Experimental

Core QELM research direction.

Multi-head quantum attention

Experimental

Quantum attention-like modeling concept.

Sub-bit encoding

Experimental

Compact token representation direction.

Entropy-mixed gates

Experimental

Gate-mixing research toggle.

Parameter-shift gradient training

Implemented concept

Training method described in repo brief.

Pauli twirling and zero-noise extrapolation

Noise mitigation

Aer/IBM path support where available.

QELMChatUI.py

Interface

Loads .qelm models and token maps.