fullscreen

eFinder

eFinder

Home › Article

Quantum circuits help AI overcome memory limitations with minimal new parameters

Phys · · 579 words · By Sam Jarman
headphones Listen to the eFinder podcast briefing
Ready to play
Daily briefing

What to know about Quantum circuits help AI overcome memory limitations with minimal new parameters

Researchers from Multiverse Computing have developed a hybrid AI system that integrates small quantum circuit blocks into pre-trained large language models. Initial tests on Llama 3.1 8B and SmolLM2 showed modest performance improvements in perplexity and accuracy with minimal additional parameters.

Propaganda risk 10%
Claims checked 7
Techniques found 0
Topics 0

Coverage spectrum

Coverage gap: Low Left coverage
Left0%
Center80%
Right20%

5 sources compared across this story cluster. This is an eFinder estimate from indexed source coverage, not an editorial rating.

What happened

June 7, 2026 report Quantum circuits help AI overcome memory limitations with minimal new parameters Sam Jarman Author Gaby Clark Scientific Editor Robert Egan Associate Editor For millions of people, chatbots powered by large language models (LLMs) are now a…

Why it matters

These AI systems are growing at a rapid pace, but scaling them up is becoming increasingly costly and resource-intensive.

Common ground

Through a new preprint on the arXiv server, a team led by Borja Aizpurua at Multiverse Computing in San Sebastián, Spain, has found a way to improve the performance of LLMs using quantum computing.

Perspective signals

No major persuasion pattern has been attached yet, so the source, headline, and evidence should carry most of the weight for readers.

Researchers from Multiverse Computing have developed a hybrid AI system that integrates small quantum circuit blocks into pre-trained large language models. Initial tests on Llama 3.1 8B and SmolLM2 showed modest performance improvements in perplexity and accuracy with minimal additional parameters.

open_in_new Read the original article: https://phys.org/news/2026-06-quantum-circuits-ai-memory-limitations.html

analyticsAnalysis

10%
Propaganda Score
confidence: 95%
Low risk. This article shows minimal use of propaganda techniques.

fact_checkClaims Checked

eFinder analyzed this article and checked 7 claims against available evidence, cross-references, web search, and Wikipedia. Here is what the fact-checking layer found.

check_circle Corroborated 4
info Single Source 2
verified Verified 1
verified
Claim 1: “The team also tested their platform on SmolLM2, a smaller 135-million-parameter model”
VERIFIED
Web search results confirm the existence of SmolLM2 with a 135-million-parameter version (0.1B), and the context of the research team testing their platform on it is supported by the overall narrative of the evidence.
menu_book
wikipedia NEUTRAL — Multiverse Computing is a quantum AI software company headquartered in San Sebastián, Spain, with offices in Madrid, Barcelona, Zaragoza, Paris, Munich, London, Milan, Toronto and San Francisco. The c…
https://en.wikipedia.org/wiki/Multiverse_Computing
travel_explore
web search NEUTRAL — In this video, I test out another local LLM SmolLM2 models (1.7B and 135M) available from ollama.com to see how they perform in extracting structured text u...
https://www.youtube.com/watch?v=5SNjdbICtkk
travel_explore
web search NEUTRAL — SmolLM2 builds on the success of its predecessor, SmolLM1, by offering enhanced capabilities while remaining lightweight. These models come in three configurations: 0.1B, 0.3B, and 1.7B parameters.
https://phdstudio.org/2024/11/01/smollm2-released-the-new-se…
+ 1 more evidence source
info
Claim 2: “GPT-5.5, for instance, is estimated to contain somewhere between two and five trillion parameters.”
SINGLE SOURCE
While web search results mention the existence of 'GPT-5.5' as a frontier model from April 2026, none of the provided evidence sources mention a parameter count of 'two to five trillion'.
travel_explore
web search NEUTRAL — openai/gpt-5.5. Compare.GPT-5.5 is OpenAI’s frontier model designed for complex professional workloads, building on GPT-5.4 with stronger reasoning, higher reliability, and improved token efficiency o…
https://openrouter.ai/openai/gpt-5.5
travel_explore
web search NEUTRAL — GPT-5.5 is OpenAI's April 2026 frontier model: 88.7% SWE-bench, 60% fewer hallucinations, Thinking and Pro variants. Full breakdown of features, pricing, access, and how to try it today.
https://apidog.com/blog/what-is-gpt-5-5/
travel_explore
web search NEUTRAL — GPT-5.5, Claude 4.7, Gemini: Хватит платить за все нейронки!Три недели я тестировал GPT, Claude, Gemini, Grok и DeepSeek на одинаковых задачах, чтобы понять
https://www.youtube.com/watch?v=2WWp799Reh8
check_circle
Claim 3: “Borja Aizpurua et al, Quantum-enhanced Large Language Models on Quantum Hardware via Cayley Unitary Adapters, arXiv (2026). DOI: 10.48550/arxiv.2605.05914”
CORROBORATED
The specific paper title, authors, year (2026), and DOI (10.48550/arxiv.2605.05914) are explicitly cited in multiple web search results.
menu_book
wikipedia NEUTRAL — Quantum programming refers to the process of designing and implementing algorithms that operate on quantum systems, typically using quantum circuits composed of quantum gates, measurements, and classi…
https://en.wikipedia.org/wiki/Quantum_programming
menu_book
wikipedia NEUTRAL — Microsoft Azure Quantum is a public cloud-based quantum computing platform developed by Microsoft, that offers quantum hardware, software, and solutions for developers to build quantum applications. I…
https://en.wikipedia.org/wiki/Microsoft_Azure_Quantum
menu_book
wikipedia NEUTRAL — A quantum computer is a real or theoretical computer that exploits quantum phenomena like superposition and entanglement in an essential way. It is widely believed that a quantum computer could perfor…
https://en.wikipedia.org/wiki/Quantum_computing
+ 3 more evidence sources
check_circle
Claim 4: “When Aizpurua's team applied this approach to Llama 3.1 8B, an eight-billion-parameter model developed by Meta, they achieved a 1.4% reduction in perplexity... while adding just 6,000 extra parameters.”
CORROBORATED
Multiple sources confirm the use of Meta's Llama 3.1 8B, the addition of approximately 6,000 parameters, and a reduction in perplexity (though the exact 1.4% figure is specifically mentioned in the context of the overall results in the search snippets).
menu_book
wikipedia NEUTRAL — Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd., doing business as DeepSeek, is a Chinese artificial intelligence (AI) company that develops large language models (LLMs).…
https://en.wikipedia.org/wiki/DeepSeek
menu_book
wikipedia NEUTRAL — In artificial intelligence, a foundation model (FM), also known as large x model (LxM, where "x" is a variable representing any text, image, sound, etc.), is a machine learning or deep learning model …
https://en.wikipedia.org/wiki/Foundation_model
menu_book
wikipedia NEUTRAL — Multiverse Computing is a quantum AI software company headquartered in San Sebastián, Spain, with offices in Madrid, Barcelona, Zaragoza, Paris, Munich, London, Milan, Toronto and San Francisco. The c…
https://en.wikipedia.org/wiki/Multiverse_Computing
+ 3 more evidence sources
info
Claim 5: “the quantum-enhanced model was able to correctly answer questions that two purely classical versions of the same model got wrong.”
SINGLE SOURCE
One arXiv snippet mentions a table showing representative MMLU questions where the CUA-enhanced Llama 3.1 8B answers correctly and the original fails. However, the claim specifically mentions 'SmolLM2' and 'two purely classical versions', which is not explicitly detailed in the other provided snippets.
travel_explore
web search NEUTRAL — Jan 16, 2026 ... To investigate this issue, we evaluate the training performance of LLMs in a federated learning setting using two aggregation methods: standard.
https://ceur-ws.org/Vol-4152/paper50.pdf
travel_explore
web search NEUTRAL — The assumptions that language models can correct each other's answers behind existing orchestration methods do not hold for. SLMs (Taubenfeld et al., 2024; ...
https://openreview.net/pdf?id=317bcKF4zv
travel_explore
web search NEUTRAL — May 7, 2026 ... Table 1 | | Representative MMLU questions where the CUA-enhanced Llama 3.1 8B answers correctly and the original fails. Both wins are verified ...
https://arxiv.org/html/2605.05914v1
check_circle
Claim 6: “the quantum components are executed on IBM's 156-qubit superconducting quantum processor.”
CORROBORATED
Two separate web search articles explicitly state that the quantum components were executed on IBM's 156-qubit superconducting quantum processor.
menu_book
wikipedia NEUTRAL — A large language model (LLM) is an AI model (typically a neural network) trained on a vast amount of text for natural language processing tasks, especially language generation. LLMs can typically gene…
https://en.wikipedia.org/wiki/Large_language_model
menu_book
wikipedia NEUTRAL — A large language model (LLM) is a type of machine learning model designed for natural language processing tasks such as language generation. LLMs are language models with many parameters, and are trai…
https://en.wikipedia.org/wiki/List_of_large_language_models
menu_book
wikipedia NEUTRAL — Retrieval-augmented generation (RAG) is a technique that enables large language models (LLMs) to retrieve and incorporate new information from external data sources. With RAG, LLMs first refer to a sp…
https://en.wikipedia.org/wiki/Retrieval-augmented_generation
+ 3 more evidence sources
check_circle
Claim 7: “Through a new preprint on the arXiv server, a team led by Borja Aizpurua at Multiverse Computing in San Sebastián, Spain, has found a way to improve the performance of LLMs using quantum computing.”
CORROBORATED
Multiple independent web search results confirm that a team led by Borja Aizpurua at Multiverse Computing in San Sebastián, Spain, published a preprint on arXiv regarding the use of quantum computing to improve LLM performance.
menu_book
wikipedia NEUTRAL — Román Orús Lacort (Barcelona, 1979) is a Spanish theoretical physicist who specializes in quantum information science and quantum tensor networks. He is Ikerbasque Research Professor at the Donostia I…
https://en.wikipedia.org/wiki/Román_Orús
travel_explore
web search NEUTRAL — Jun 7, 2026 ... Through a new preprint on the arXiv server, a team led by Borja Aizpurua at Multiverse Computing in San Sebastián, Spain, has found a way to ...
https://phys.org/news/2026-06-quantum-circuits-ai-memory-lim…
travel_explore
web search NEUTRAL — May 25, 2026 ... They say their work represents the first demonstration of "quantum enhancement" in a production-scale, pretrained large language model (LLM).
https://tech.yahoo.com/science/articles/scientists-trained-a…
+ 1 more evidence source
info Disclaimer: This analysis is generated by AI and should be used as a starting point for critical thinking, not as definitive truth. Claims are verified against publicly available sources. Always consult the original article and additional sources for complete context.