Sukhbinder Singh

Siddhartha Patra, Saeed S. Jahromi, Sukhbinder Singh et al.

We show how quantum-inspired 2d tensor networks can be used to efficiently and accurately simulate the largest quantum processors from IBM, namely Eagle (127 qubits), Osprey (433 qubits) and Condor (1121 qubits). We simulate the dynamics of a complex quantum many-body system -- specifically, the kicked Ising experiment considered recently by IBM in Nature 618, p. 500-505 (2023) -- using graph-based Projected Entangled Pair States (gPEPS), which was proposed by some of us in PRB 99, 195105 (2019). Our results show that simple tensor updates are already sufficient to achieve very large unprecedented accuracy with remarkably low computational resources for this model. Apart from simulating the original experiment for 127 qubits, we also extend our results to 433 and 1121 qubits, and for evolution times around 8 times longer, thus setting a benchmark for the newest IBM quantum machines. We also report accurate simulations for infinitely-many qubits. Our results show that gPEPS are a natural tool to efficiently simulate quantum computers with an underlying lattice-based qubit connectivity, such as all quantum processors based on superconducting qubits.

Borja Aizpurua, Sukhbinder Singh, Augustine Kshetrimayum et al.

Large language models (LLMs) have transformed artificial intelligence, yet classical architectures impose a fundamental constraint: every trainable parameter demands classical memory that scales unfavourably with model size. Quantum computing offers a qualitatively different pathway, but practical demonstrations on real hardware have remained elusive for models of practical relevance. Here we show that Cayley-parameterised unitary adapters -- quantum circuit blocks inserted into the frozen projection layers of pre-trained LLMs and executed on a 156-qubit IBM Quantum System Two superconducting processor -- improve the perplexity of Llama 3.1 8B, an 8-billion-parameter model in widespread use, by 1.4% with only 6,000 additional parameters and end-to-end inference validated on real Quantum Processing Unit (QPU). A systematic study on SmolLM2 (135M parameters), chosen for its tractability, reveals monotonically improving perplexity with unitary block dimension, 83% recovery of compression-induced degradation, and correct answers to questions that both classical baselines fail -- with a sharp noise-expressivity phase transition identifying the concrete path to quantum utility at larger qubit scales.

Safa Hamreras, Sukhbinder Singh, Román Orús

We propose a scalable tensorization framework for neural network compression based on slice-wise feature distillation. Unlike conventional tensor decomposition methods that rely on costly global finetuning, our approach decomposes the network into slices consisting of either individual layers or blocks (e.g., convolutional layers or MLPs), or small groups of consecutive layers, and tensorizes each slice independently to reproduce the intermediate representations of the original pretrained model. This modular strategy improves accuracy recovery, reduces data requirements, and enables efficient parallel optimization. Experiments on ResNet-34 show significant gains over conventional global tensorization, achieving near-lossless compression at moderate compression rates with faster optimization. Results on GPT-2 XL further demonstrate the scalability of the method and its applicability to large-scale models, particularly in distributed settings.

Borja Aizpurua, Sukhbinder Singh, Román Orús

Large language models (LLMs) contain billions of parameters, yet many exact values are not essential. We show that what matters most is the relative rank of weights-whether one connection is stronger or weaker than another-rather than precise magnitudes. To reduce the number of unique weight values, we apply weight clustering to pretrained models, replacing every weight matrix with K shared values from K-means. For Llama 3.1-8B-Instruct and SmolLM2-135M, reducing each matrix to only 16-64 distinct values preserves strong accuracy without retraining, providing a simple, training-free method to compress LLMs on disk. Optionally fine-tuning only the cluster means (centroids) recovers 30-40 percent of the remaining accuracy gap at minimal cost. We then systematically randomize cluster means while keeping assignments fixed. Scrambling the relative ranks of the clusters degrades quality sharply-perplexity can increase by orders of magnitude-even when global statistics such as mean and variance are preserved. In contrast, rank-preserving randomizations cause almost no loss at mid and late layers. On the other hand, when many layers are perturbed simultaneously, progressive layer-by-layer replacement reveals that scale drift-not rank distortion-is the dominant collapse mechanism; however, an affine correction w' = aw + b with a > 0 (which preserves both rank order and overall weight distribution) can substantially delay this drift. This rank-based perspective offers a new lens on model compression and robustness.

Borja Aizpurua, Saeed S. Jahromi, Sukhbinder Singh et al.

We propose a method to enhance the performance of Large Language Models (LLMs) by integrating quantum computing and quantum-inspired techniques. Specifically, our approach involves replacing the weight matrices in the Self-Attention and Multi-layer Perceptron layers with a combination of two variational quantum circuits and a quantum-inspired tensor network, such as a Matrix Product Operator (MPO). This substitution enables the reproduction of classical LLM functionality by decomposing weight matrices through the application of tensor network disentanglers and MPOs, leveraging well-established tensor network techniques. By incorporating more complex and deeper quantum circuits, along with increasing the bond dimensions of the MPOs, our method captures additional correlations within the quantum-enhanced LLM, leading to improved accuracy beyond classical models while maintaining low memory overhead.

Sukhbinder Singh, Saeed S. Jahromi, Roman Orus

Convolutional neural networks (CNNs) are one of the most widely used neural network architectures, showcasing state-of-the-art performance in computer vision tasks. Although larger CNNs generally exhibit higher accuracy, their size can be effectively reduced by ``tensorization'' while maintaining accuracy, namely, replacing the convolution kernels with compact decompositions such as Tucker, Canonical Polyadic decompositions, or quantum-inspired decompositions such as matrix product states, and directly training the factors in the decompositions to bias the learning towards low-rank decompositions. But why doesn't tensorization seem to impact the accuracy adversely? We explore this by assessing how \textit{truncating} the convolution kernels of \textit{dense} (untensorized) CNNs impact their accuracy. Specifically, we truncated the kernels of (i) a vanilla four-layer CNN and (ii) ResNet-50 pre-trained for image classification on CIFAR-10 and CIFAR-100 datasets. We found that kernels (especially those inside deeper layers) could often be truncated along several cuts resulting in significant loss in kernel norm but not in classification accuracy. This suggests that such ``correlation compression'' (underlying tensorization) is an intrinsic feature of how information is encoded in dense CNNs. We also found that aggressively truncated models could often recover the pre-truncation accuracy after only a few epochs of re-training, suggesting that compressing the internal correlations of convolution layers does not often transport the model to a worse minimum. Our results can be applied to tensorize and compress CNN models more effectively.

Pablo Martin-Ramiro, Unai Sainz de la Maza, Sukhbinder Singh et al.

Defect detection is one of the most important yet challenging tasks in the quality control stage in the manufacturing sector. In this work, we introduce a Tensor Convolutional Neural Network (T-CNN) and examine its performance on a real defect detection application in one of the components of the ultrasonic sensors produced at Robert Bosch's manufacturing plants. Our quantum-inspired T-CNN operates on a reduced model parameter space to substantially improve the training speed and performance of an equivalent CNN model without sacrificing accuracy. More specifically, we demonstrate how T-CNNs are able to reach the same performance as classical CNNs as measured by quality metrics, with up to fifteen times fewer parameters and 4% to 19% faster training times. Our results demonstrate that the T-CNN greatly outperforms the results of traditional human visual inspection, providing value in a current real application in manufacturing.

Roman Rausch, David Jansen, Sukhbinder Singh et al.

Large Language Models (LLMs) are very demanding in terms of their computational resources. Low-rank decompositions of LLM weights, e.g. via Singular Value Decomposition (SVD), is a promising approach for LLM compression, but presents several practical hurdles, e.g. selecting appropriate layer-wise ranks and getting rid of its parameter redundancy. In this work, we present two physics-inspired improvements to SVD LLM compression: (1) \textbf{FermiGrad}, a gradient-descent algorithm that determines globally optimal layer-wise ranks by relaxing the discrete singular-value truncation into a continuous optimization using the Fermi function; (2) \textbf{PivGa}, an additional \textit{lossless} compression of the low-rank factors that exploits the intrinsic gauge freedom in their parametrization.

Andrei Tomut, Saeed S. Jahromi, Abhijoy Sarkar et al.

Large Language Models (LLMs) such as ChatGPT and LlaMA are advancing rapidly in generative Artificial Intelligence (AI), but their immense size poses significant challenges, such as huge training and inference costs, substantial energy demands, and limitations for on-site deployment. Traditional compression methods such as pruning, distillation, and low-rank approximation focus on reducing the effective number of neurons in the network, while quantization focuses on reducing the numerical precision of individual weights to reduce the model size while keeping the number of neurons fixed. While these compression methods have been relatively successful in practice, there is no compelling reason to believe that truncating the number of neurons is an optimal strategy. In this context, this paper introduces CompactifAI, an innovative LLM compression approach using quantum-inspired Tensor Networks that focuses on the model's correlation space instead, allowing for a more controlled, refined and interpretable model compression. Our method is versatile and can be implemented with - or on top of - other compression techniques. As a benchmark, we demonstrate that a combination of CompactifAI with quantization allows to reduce a 93% the memory size of LlaMA 7B, reducing also 70% the number of parameters, accelerating 50% the training and 25% the inference times of the model, and just with a small accuracy drop of 2% - 3%, going much beyond of what is achievable today by other compression techniques. Our methods also allow to perform a refined layer sensitivity profiling, showing that deeper layers tend to be more suitable for tensor network compression, which is compatible with recent observations on the ineffectiveness of those layers for LLM performance. Our results imply that standard LLMs are, in fact, heavily overparametrized, and do not need to be large at all.