To discover new physics, AI may need to 'unlearn' the old one

Phys · Jun 10, 2026 · 727 words · By SISSA Medialab

headphones Listen to the eFinder podcast briefing

Ready to play

Daily briefing

What to know about To discover new physics, AI may need to 'unlearn' the old one

The article discusses a study published in the Journal of Cosmology and Astroparticle Physics regarding the use of transfer learning in AI to search for new physics beyond the standard cosmological model. It explains how pretraining AI on simpler models can reduce computational costs but may also lead to 'negative transfer,' where prior knowledge hinders the detection of new physical phenomena.

Propaganda risk 0%

Claims checked 10

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

To discover new physics, AI may need to 'unlearn' the old one Sadie Harley Scientific Editor Robert Egan Associate Editor A study in the Journal of Cosmology and Astroparticle Physics explores how a machine-learning strategy known as transfer learning could…

Why it matters

Artificial intelligence is widely used in cosmology to analyze the universe.

Common ground

But testing theories beyond the standard cosmological model, known as ΛCDM, remains extremely computationally demanding.

Perspective signals

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

Follow-up questions

What concrete event or decision sits underneath the headline: To discover new physics, AI may need to 'unlearn' the old one?
What evidence would most clearly confirm or weaken the claim that transfer learning—a technique in which AI systems reuse knowledge acquired from one task to accelerate learning in another?
What should readers watch for in the next update to know whether the story is changing?

open_in_new Read the original article: https://phys.org/news/2026-06-physics-ai-unlearn.html

analyticsAnalysis

Propaganda Score

confidence: 100%

Low risk. This article shows minimal use of propaganda techniques.

fact_checkClaims Checked

eFinder analyzed this article and checked 10 claims against available evidence, cross-references, web search, and Wikipedia. Here is what the fact-checking layer found.

info Single Source 4

check_circle Corroborated 3

verified Verified By Reference 3

check_circle

Claim 1: “transfer learning—a technique in which AI systems reuse knowledge acquired from one task to accelerate learning in another”

CORROBORATED

Multiple independent sources (IBM, Meegle, and a research paper) define transfer learning as reusing knowledge from one task to accelerate learning in another.

travel_explore

web search NEUTRAL — Transfer learning is a machine learning technique in which knowledge gained through one task or dataset is used to improve model performance on another related ...
https://www.ibm.com/think/topics/transfer-learning

travel_explore

web search NEUTRAL — Transfer learning focuses on reusing knowledge from a pre-trained model or a previously learned task to accelerate learning in a new, related task. Continual ...
https://www.meegle.com/en_us/topics/transfer-learning/transf…

travel_explore

web search NEUTRAL — Oct 13, 2024 ... Transfer Learning focuses on leveraging knowledge from one task to accelerate learning in another, whereas Reinforcement Learning enables agents to learn ...
https://medium.com/@hassaanidrees7/transfer-learning-vs-60c9…

check_circle

Claim 2: “researchers first trained a neural network on simulations based on ΛCDM—this is known as pretraining—and then adapted it to more complex cosmological models”

CORROBORATED

Three separate web search snippets describe the exact methodology: pretraining a neural network on ΛCDM simulations and then adapting/fine-tuning it for more complex models.

travel_explore

web search NEUTRAL — In this case, researchers first trained a neural network on simulations based on ΛCDM—this is known as pretraining—and then adapted it to more complex cosmological models that include possible new phy…
https://phys.org/news/2026-06-physics-ai-unlearn.html

travel_explore

web search NEUTRAL — For this study, the team first trained a neural network using simulations based on ΛCDM. This initial training process, known as pretraining, gave the AI a foundation before it was exposed to more com…
https://scitechdaily.com/ai-learned-the-rules-of-the-univers…

travel_explore

web search NEUTRAL — Simulating universes with exotic physics is far more expensive than simulating the standard model. By first training a network on cheap standard-model simulations and then fine-tuning it on a much sma…
https://scienceblog.com/neuroedge/2026/06/10/to-find-new-phy…

verified

Claim 3: “On arXiv: DOI: 10.48550/arxiv.2510.19168”

VERIFIED BY REFERENCE

The provided evidence only defines what a DOI is in general; it does not confirm the specific DOI 10.48550/arxiv.2510.19168 for the mentioned study.

menu_book

wikipedia NEUTRAL — DOI or Doi most commonly refers to: Digital object identifier, an international standard for document identification United States Department of the Interior, an executive department of the U.S. gove…
https://en.wikipedia.org/wiki/DOI

menu_book

wikipedia NEUTRAL — A digital object identifier (DOI) is a persistent identifier, or persistent handle, used to uniquely identify various objects, standardized by the International Organization for Standardization (ISO).…
https://en.wikipedia.org/wiki/Digital_object_identifier

menu_book

wikipedia NEUTRAL — Doi (株式会社ドイ, Kabushiki Kaisha Doi) was a large Japanese retailer and distributor, best known outside Japan as the company that revived the Plaubel Makina 67 camera in the late 1970s and early 1980s. I…
https://en.wikipedia.org/wiki/Doi_(retailer)

verified

Claim 4: “ΛCDM successfully describes many properties of the universe—from its expansion to the distribution of galaxies”

VERIFIED BY REFERENCE

Wikipedia and NASA sources confirm that the ΛCDM model is the standard model of cosmology and describes the expansion of the universe and the distribution of matter/galaxies.

travel_explore

web search NEUTRAL — The Lambda-CDM, Lambda cold dark matter, or ΛCDM model is a mathematical model of the Big Bang theory with three major components: ordinary matter. It is the current standard model of Big Bang cosmolo…
https://en.wikipedia.org/wiki/Lambda-CDM_model

travel_explore

web search NEUTRAL — According to the current standard model of cosmology, the Lambda-CDM model, approximately 27% of the universe is dark matter and 68% is dark energy, with only a small fraction being the ordinary baryo…
https://en.wikipedia.org/wiki/Cold_dark_matter

travel_explore

web search NEUTRAL — In this picture, the infant universe is an extremely hot, dense, nearly homogeneous mixture of photons and matter, tightly coupled together as a plasma. An approximate graphical timeline of its theore…
https://lambda.gsfc.nasa.gov/education/graphic_history/univ_…

info

Claim 5: “The researchers observed this behavior in simulations involving massive neutrinos”

SINGLE SOURCE

The provided evidence for this claim contains only general definitions of the word 'negative' and unrelated businesses; there is no mention of neutrino simulations or 'negative transfer' in the context of the study.

travel_explore

web search NEUTRAL — This disambiguation page lists articles associated with the title Negative. If an internal link incorrectly led you here, you may wish to change the link to point directly to the intended article.
https://en.wikipedia.org/wiki/Negative

travel_explore

web search NEUTRAL — Negative is underwear that fits your life - effortless bras, underwear, thongs and loungewear. We’re not lacy, pink lingerie. We’re sheer, seamless and minimal.
https://negativeunderwear.com/

travel_explore

web search NEUTRAL — The meaning of NEGATIVE is marked by denial, prohibition, or refusal; also : marked by absence, withholding, or removal of something positive. How to use negative in a sentence.
https://www.merriam-webster.com/dictionary/negative

check_circle

Claim 6: “testing theories beyond the standard cosmological model, known as ΛCDM, remains extremely computationally demanding”

CORROBORATED

Multiple independent web search results explicitly state that testing theories beyond the ΛCDM model is extremely computationally demanding.

travel_explore

web search NEUTRAL — But testing theories beyond the standard cosmological model, known as ΛCDM, remains extremely computationally demanding. Although ΛCDM successfully describes many properties of the universe—from its e…
https://phys.org/news/2026-06-physics-ai-unlearn.html

travel_explore

web search NEUTRAL — For the standard ΛCDM model, those training sets already exist. For beyond-ΛCDM theories — the models that might contain new physics — they don’t, because nobody can afford to run that many.
https://www.startgaze.com/posts/2026-06-28-transfer-learning…

travel_explore

web search NEUTRAL — The base ΛCDM model is powerfully predictive, with little room for maneuver and only a few free parameters, so it is therefore easier to test than more baroque models. Theories beyond ΛCDM are often l…
https://www.academia.edu/85897426/Beyond_ΛCDM_Problems_solut…

info

Claim 7: “Certain effects produced by neutrino mass closely resemble variations associated with an existing ΛCDM parameter known as σ8, which describes how strongly matter clusters across the universe”

SINGLE SOURCE

While web search results mention neutrino-dark matter interactions and the S8 tension, none of the provided snippets explicitly confirm the specific claim that neutrino mass effects resemble variations of the σ8 parameter in the context of this AI study.

travel_explore

web search NEUTRAL — Jan 2, 2026 ... Here we present compelling evidence that DM–neutrino interactions can resolve the persistent structure growth parameter discrepancy, {S}_{8}={\ ...
https://www.nature.com/articles/s41550-025-02733-1

travel_explore

web search NEUTRAL — Aug 12, 2025 ... We show the neutrino signature kernels estimate the same cosmological parameters as the model with the SPT (Standard Perturbation Theory) ...
https://arxiv.org/pdf/2508.06759

travel_explore

web search NEUTRAL — Sep 2, 2025 ... neutrino degeneracy parameter is compatible with Big Bang nucleosynthesis data. The scalar index ns exceeds 1 slightly, which is compatible with ...
https://iopscience.iop.org/article/10.3847/2041-8213/adfc6d

info

Claim 8: “In some cases, transfer learning reduced the number of expensive simulations needed by more than a factor of 10”

SINGLE SOURCE

The provided evidence for this claim consists of general definitions of transfer learning or unrelated Wikipedia entries (football transfers, military transfers). No specific evidence in the provided snippets confirms the 'factor of 10' reduction for this specific study.

menu_book

wikipedia NEUTRAL — In the United States Armed Forces, a dignified transfer is a procedure honoring the return of the remains of a servicemember from the theater of operations where they have died in the service of the U…
https://en.wikipedia.org/wiki/Dignified_transfer

menu_book

wikipedia NEUTRAL — This is a list of Cypriot football transfers for the 2009–10 winter transfer window by club. Only transfers of clubs in the Cypriot First Division and Cypriot Second Division are included. The winter …
https://en.wikipedia.org/wiki/List_of_Cypriot_football_trans…

menu_book

wikipedia NEUTRAL — The following is a list of most expensive association football transfers, which details the highest transfer fees ever paid for players, as well as transfers which set new world transfer records. The …
https://en.wikipedia.org/wiki/List_of_most_expensive_associa…

+ 3 more evidence sources

verified

Claim 9: “Veena Krishnaraj, et al. Transfer Learning Beyond the Standard Model, Journal of Cosmology and Astroparticle Physics (2026)”

VERIFIED BY REFERENCE

The provided Wikipedia evidence for this claim is completely irrelevant, discussing general learning standards, machine learning, and transformers, but not the specific paper by Veena Krishnaraj.

menu_book

wikipedia NEUTRAL — Learning standards (also called academic standards, content standards and curricula) are elements of declarative, procedural, schematic, and strategic knowledge that, as a body, define the specific co…
https://en.wikipedia.org/wiki/Learning_standards

menu_book

wikipedia NEUTRAL — Machine learning (ML) is a field of study in artificial intelligence concerned with the development and study of statistical algorithms that can learn from data and generalize to unseen data, and thu…
https://en.wikipedia.org/wiki/Machine_learning

menu_book

wikipedia NEUTRAL — In deep learning, the transformer is a family of artificial neural network architectures based on the multi-head attention mechanism, in which text is converted to numerical representations called tok…
https://en.wikipedia.org/wiki/Transformer_(deep_learning)

info

Claim 10: “A study in the Journal of Cosmology and Astroparticle Physics explores how a machine-learning strategy known as transfer learning could dramatically reduce the computational cost of searching for new physics beyond the standard cosmological model”

SINGLE SOURCE

While web search results discuss machine learning in cosmology and the existence of the Journal of Cosmology and Astroparticle Physics, none of the provided evidence snippets explicitly confirm this specific study about transfer learning reducing computational costs. The evidence is too general to corroborate the specific claim.

menu_book

wikipedia NEUTRAL — Astroparticle Physics is a peer-reviewed scientific journal covering experimental and theoretical research in the interacting fields of cosmic ray physics, astronomy and astrophysics, cosmology, and p…
https://en.wikipedia.org/wiki/Astroparticle_Physics_(journal…

menu_book

wikipedia NEUTRAL — Astroparticle physics, also called particle astrophysics, is a branch of particle physics that studies elementary particles of astrophysical origin and their relation to astrophysics and cosmology. It…
https://en.wikipedia.org/wiki/Astroparticle_physics

menu_book

wikipedia NEUTRAL — The Journal of Cosmology and Astroparticle Physics is an online-only peer-reviewed scientific journal focusing on all aspects of cosmology and astroparticle physics. This encompasses theory, observati…
https://en.wikipedia.org/wiki/Journal_of_Cosmology_and_Astro…

+ 3 more evidence sources

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.