Researchers propose new listeria labelling for smoked salmon and other ready-to-eat foods

To reduce the number of Listeria infections, researchers at DTU are proposing a new labelling scheme designed to help consumers choose products that inhibit the growth of Listeria.

Technical University of Denmark

Facebook

XLinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

Older people and pregnant women are particularly vulnerable to Listeria infections, which can originate, for example, from smoked or gravad/marinated fish. Photo: Lene Koss

view more 

Credit: Photo: Lene Koss

Listeria infections are on the rise, and around 15% of people in the EU who contract the infection die from it.

Researchers from the DTU National Food Institute are proposing a new labelling scheme designed to give consumers a better opportunity to choose ready-to-eat foods, such as smoked salmon and spiced pork roll, without risking infection with Listeria bacteria.

Listeria accounts for approximately 32% of all deaths associated with foodborne disease outbreaks in the EU. The infection poses a particular risk to older people, those with underlying health conditions and other vulnerable groups, as well as pregnant women.

“A labelling scheme should be introduced for ready-to-eat foods that have been stabilised during production to prevent the growth of Listeria. This will enable us to reduce the number of people who fall ill with Listeria infections. The label should give consumers confidence in eating these ordinarily healthy fish products, which may however, be prone to contain Listeria,” says Martin Laage Kragh, a researcher at the DTU National Food Institute.

Products already exist that are produced in a way so that Listeria cannot grow in them. However, it is currently difficult for consumers to determine which products have been stabilised to prevent the growth of Listeria and are therefore safe to eat.

The researchers propose a voluntary labelling scheme using the word “STABILISED”, which should make it easier for consumers to choose foods with a minimal risk of Listeria infection.

“The label should only be used by producers who can demonstrate that Listeria cannot grow in their products,” says Professor Lisbeth Truelstrup Hansen from the DTU National Food Institute.

In short, stability means that the product’s preservation must be appropriate for the stated shelf life.

“Many products have a shelf life that is too long given the way they are made. But by changing the recipe – and often this requires only a small change – the products can be made safe,” says Professor Emeritus Paw Dalgaard from the DTU National Food Institute.

How to prevent Listeria from growing

A shorter shelf life limits the growth of Listeria and generally reduces the risk of Listeria infections.

In the case of smoked and marinated/gravad fish, which account for many cases of listeria infections, one can, for example, add vinegar during the salting process – in such small quantities that it cannot be tasted – thereby preventing listeria from growing in the products.

Another option is to sell frozen products. This will effectively prevent the growth of Listeria.

For other types of food, including cold cuts, heat or high-pressure treatment of products in consumer packaging may be a solution. This kills Listeria if the bacterium has entered the product during packaging.

There will be different approaches to making the products safe, depending on the type of food involved.

Traditionally, efforts to combat Listeria have focused on cleaning in the food industry, but this has proved not to be enough.

“Cleaning at the production stage is very important when it comes to preventing Listeria, but it has not prevented a rise in the number of people falling ill. We suggest maintaining the focus on cleaning, but also placing a new and equally important focus on producing food in such a way that Listeria cannot grow in ready-to-eat products,” says researcher Martin Laage Kragh.

The cause of current outbreaks should be analysed

The researchers also suggest that products which have led to recent outbreaks of Listeria infections should be analysed in a new way, so that the reason Listeria grew in the products can be quickly identified.

As things stand, the authorities often identify the source of infection in outbreaks, but this has frequently failed to prevent new outbreaks.

“We propose that investigations should also look into whether the manufacturer is using a formulation where preservation could be improved, or whether the product’s shelf life should, for example, be shortened to prevent new outbreaks,” says Paw Dalgaard.

It has in fact been shown that outbreaks of Listeria infections can occur repeatedly and be traced back to the same products – and sometimes even to the same manufacturers. It is therefore important to identify what is going wrong with those specific products during production and distribution.

Read more

The researchers’ recommendations have been published in a scientific article in the journal Current Opinion in Food Science: The solution to Listeria monocytogenes problems in the food industry is product stabilisation combined with cleaning and disinfection.

Facts

The number of Listeria infections in the EU rose from 0.40 per 100,000 inhabitants in 2010 to 0.69 per 100,000 in 2024.

The disease has a case fatality rate of 15.6% and is particularly associated with the consumption of ready-to-eat foods such as smoked or gravad fish, cold cuts and soft cheeses made from unpasteurised milk.

Older people, pregnant women and vulnerable individuals are at particular risk. The increase in infections may be due, amongst other factors, to growing demand for ready-to-eat foods with a long shelf life and to the rising number of older people and individuals with underlying health conditions.

This trend requires changes to the production and/or shelf life of several types of products. Consequently, from 1 July 2026, the EU has introduced stricter rules requiring manufacturers of non-stabilised products to demonstrate that the level of Listeria monocytogenes does not exceed 100 bacteria per gram throughout the entire shelf life of the ready-to-eat product. If they are unable to do so, the bacterium must not be detectable in a 25-gram sample.

---

Journal

Current Opinion in Food Science

DOI

10.1016/j.cofs.2026.101414 

Article Title

The solution to Listeria monocytogenes problems in the food industry is product stabilization in combination with cleaning and disinfection

 

Graph learning and LLMs advance social network alignment research

Higher Education Press

Facebook

WeChat
WhatsAppEmail

 

image: 

GRL framework for solving SNA problems

view more 

Credit: HIGHER EDUCATION PRESS

With the growing use of multiple social platforms, aligning user identities across networks, known as Social Network Alignment (SNA), has become crucial for personalized services, fraud detection, and online safety. However, existing SNA methods face significant challenges when dealing with sparse connections, heterogeneous structures, and dynamic changes in real-world social networks.

To address these problems, a research team led by Professor Dan Feng from Huazhong University of Science and Technology published a new study on 15 June 2026 in Frontiers of Computer Science, co-published by Higher Education Press and Springer Nature.

The team conducted a systematic review of SNA methods based on Graph Representation Learning (GRL), with a special focus on emerging approaches that integrate Large Language Models (LLMs) to enhance semantic reasoning and alignment accuracy.

In the research, they present a unified perspective on SNA methods, covering both static and dynamic networks as well as homogeneous and heterogeneous structures. They propose a comprehensive taxonomy of existing approaches, ranging from early matrix factorization techniques and shallow random-walk-based models to state-of-the-art deep graph neural networks. Uniquely, they emphasize the emerging role of LLMs, such as Qwen, Llama2 and ERNIE, in enhancing alignment precision through semantic reasoning, which is an aspect largely overlooked in prior research. Additionally, they provide a benchmarking analysis across more than ten real-world datasets, offering a thorough comparison of the effectiveness of various SNA methods under diverse conditions.

Future work can focus on developing more interpretable, scalable, and privacy-preserving SNA methods. In particular, there is strong potential in combining GRL with lightweight or distilled LLMs to reduce computational costs while maintaining high alignment accuracy.

---

Journal

Frontiers of Computer Science

DOI

10.1007/s11704-025-40985-2 

Method of Research

Experimental study

Subject of Research

Not applicable

Article Title

A survey of social network alignment methods based on graph representation learning

 

Environment is a nexus: Generalization process for domain generalization

Higher Education Press

Facebook

XLinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

An illustration of the inference procedure in the Generalization Process, consisting of environment-level and task-level inference.

 

view more 

Credit: HIGHER EDUCATION PRESS

In recent years, domain generalization (DG) has garnered significant attention for its goal of learning models from multiple source domains to mitigate domain shift and enable generalization to unseen target domains. Most existing methods e.g., domain-invariant representation, focus on learning a universal sample-to-label mapping (function) across domains, overlooking semantically intrinsic domain-specific information and failing to maintain invariance across diverse unseen target domains.
To solve the problems, a research team led by Songcan CHEN published their new research on 15 June 2026 in Frontiers of Computer Science co-published by Higher Education Press and Springer Nature.
The team provides a novel standpoint for DG, i.e., domains are connected by a meta-distribution, namely environment distribution, which can sample various functions. In this way, they establish a meta-function function based on Gaussian Process, mapping from the environment to functions, to induce specific functions for unseen domains from observed function set. Compared with existing methods, the proposed method demonstrates superiority both theoretically and empirically.
In this research, they critically examine the limitations of existing domain-invariant representation (DIR) methods, highlighting their dependence on observed domains and their difficulty in maintaining invariance across diverse unseen target domains. To address these shortcomings, they propose a novel perspective: instead of learning a single universal function, they advocate for learning a function over functions.
Within this framework, each domain is treated as a meta-sample drawn from the environment distribution, and each domain-specific function is regarded as a sample from this meta-distribution. This abstraction enables the learning of a meta-function that can generate domain-specific functions capable of adapting to previously unseen domains.
To realize this idea, the team introduces GPDG, a Gaussian Process-based learning paradigm. GPDG captures both intra-domain information and inter-domain correlations using domain distributions and a kernel-based architecture. A Dirichlet Mixup-based domain augmentation strategy is also employed to enhance diversity and smoothness in the functional space. Extensive experiments are constructed to demonstrate the effectiveness of proposed GPDG.
Future work can focus on exploring simpler strategies for modeling the meta-function as alternatives to the relatively complex Gaussian Process.

 

---

Journal

Frontiers of Computer Science

DOI

10.1007/s11704-025-41278-4 

Method of Research

Experimental study

Subject of Research

Not applicable

Article Title

Environment is a nexus: generalization process for domain generalization ​​​​​​​

 

Key findings illustrating dark knowledge to facilitate powerful distillation

Higher Education Press

Facebook

XLinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

The observation of variance of outputs on non-target classes, which can illustrate dark knowledge.

view more 

Credit: HIGHER EDUCATION PRESS

 

As large models advance, there’s growing demand to use knowledge distillation to produce smaller, more portable models (student) that match higher-performing larger models (teacher). However, when a teacher’s capacity far exceeds the student’s, distillation often degrades, which is known as capacity mismatch. This mismatch caps the student model’s performance and has become a bottleneck in large-model distillation. Existing mitigation techniques remain ad hoc, and no study has yet systematically explained its root causes or proposed targeted methods to enable larger teachers to yield better distillation results.

 

To solve the problems, a research team led by De-Chuan Zhan published their new research on 15 June 2026 in Frontiers of Computer Science co-published by Higher Education Press and Springer Nature.

 

The team has identified two key characteristics in knowledge distillation. First, as the teacher model grows, the variance of its probability outputs on non-target classes rises then falls. This variance—reflecting sample–class relationships—is essential to transfer and correlates with distillation performance; when the teacher becomes too large, this variance shrinks and distillation degrades. Second, despite changes in teacher capacity, the ranking of class output magnitudes remains unchanged, showing that simple, order-preserving temperature scaling can adjust class variance without disrupting the teacher’s inherent knowledge. Their insights deepen the understanding of dark knowledge, reveal the origin of capacity mismatch, and guide the design of more effective distillation methods. Based on these insights, they propose Instance-Specific Asymmetric Temperature Scaling (ISATS) method. For each example, ISATS applies different distillation temperatures to the correct class versus all incorrect classes—and chooses the incorrect-class temperature to maximize their output variance—thereby enriching dark knowledge in two ways. Experiments on numerous datasets show that their method outperforms previous capacity-mismatch mitigation techniques and ensures larger teacher models teach more effectively.

---

Journal

Frontiers of Computer Science

DOI

10.1007/s11704-025-41434-w 

Method of Research

Experimental study

Subject of Research

Not applicable

Article Title

Exploring dark knowledge under various teacher capacities and addressing capacity mismatch

 

Were Clovis foragers in Late Pleistocene North America big-game hunters, or just big-game scavengers?

Kent State University

Facebook

XLinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

Clovis fluted points from Blackwater Draw, New Mexico.

view more 

Credit: Journal of Archaeological Science: Reports

There are currently 15 well-documented Late Pleistocene localities in North America in which Clovis points are found associated with proboscidean remains (of mammoth, mastodon, and gomphothere). Archaeologists routinely assume these localities represent evidence that Clovis people hunted these multi-tone animals, and in turn invoke that evidence to claim humans had a role in the extinction of these large mammals. Yet, archaeologists have not thoroughly tested their assumption, nor fully considered the possibility that Clovis foragers were facultative scavengers, which might as readily account for the association of artifacts with proboscidean remains at some or even all these localities. A significant obstacle to differentiating hunting from scavenging archaeologically is the challenge of equifinality.

Five researchers from Kent State University, Southern Methodist University (SMU), the Smithsonian, the University of Michigan, and the University of Utah explored whether Clovis foragers hunted, scavenged, or did both, and whether it was possible to tell the difference archaeologically. The new research is now published in the Journal of Archaeological Science: Reports.

To set a broad foundation for considering the question, they began their study showing the near ubiquity of scavenging among non-human animals. They document the fact that the paleoanthropological, ethnohistoric, and ethnographic records showed that scavenging was also quite common among human groups past and present. Following that, they consider the many opportunities scavengers have – and Clovis foragers might have had – in exploiting proboscidean carrion. Lastly, they assess the proposed archaeological evidence for Clovis hunting and scavenging and show that while Clovis foragers likely practiced both. “Researchers cannot currently distinguish the two archaeologically and thus cannot reliably show how many Clovis proboscidean sites represent hunting versus scavenging events,” said Kent State’s Metin I. Eren.

“While Clovis foragers likely hunted mammoths, it would be odd indeed if Clovis foragers – alone among ethnohistoric and ethnographic human groups and nearly all omnivores and carnivores – did not scavenge,” said SMU’s David J. Meltzer. Meltzer added that “scavenging also could possibly explain the high δ15N values recently reported for the Anzick child, which could readily result from his mother eating maggots and not mammoth meat.” The researchers concluded that given the present state of knowledge, the Clovis archaeological record cannot be used to argue that Clovis groups routinely hunted proboscideans, or that there are sufficient “kill sites” to support a human role in proboscidean extinctions.

The Smithsonian’s Briana Pobiner, the University of Utah’s James O’Connell, and the University of Michigan’s John D. Speth were the other contributing authors to the study.

“If we cannot definitively conclude that proboscidean killing took place at any single Clovis site because there is archaeological equifinality with scavenging, then proboscidean overkilling is not supported either,” Eren said. “Despite some archaeologists’ and other scientists’ long-standing beliefs, there is just no definitive scientific evidence for a human role in the North American Late Pleistocene extinction of proboscideans.”

---

Journal

Journal of Archaeological Science Reports

DOI

10.1016/j.jasrep.2026.105896 

Method of Research

Literature review

Subject of Research

People

Article Publication Date

1-Jul-2026

 

China's planted forests are more vulnerable to flash droughts, study finds

SciOpen

LinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

The analysis identified and characterized flash and slow droughts in natural and planted forests. A Bi-GRU network simulated baseline monthly drought days (MPDD), while a KNN algorithm matched forest stands by environmental covariates to ensure comparable climates. Using these matched pairs, counterfactual scenarios isolated how forest type affects drought regulation, holding climate constant. Finally, SHAP analysis identified key drought risk drivers for each forest type.

view more 

Credit: Kai Duan, School of Civil Engineering, Sun Yat-Sen University, Guangzhou, China

China has planted more than 77 million hectares (around 770,000 km²) of forests over the past several decades as part of the world's largest reforestation program. The effort has helped restore degraded landscapes, store carbon, and expand forest cover. But new research suggests that planted forests respond to drought differently from natural forests. The research finding was published in Forest Ecosystems.

Using satellite observations spanning 1990 to 2024, researchers from Sun Yat-Sen University investigated how planted and natural forests across China respond to two distinct types of drought: rapid-onset "flash droughts" and longer-lasting "slow droughts."

The results reveal a clear trade-off. Planted forests experienced more frequent flash droughts than natural forests, averaging 7.05 events every five years compared with 5.76 in natural forests. In contrast, natural forests experienced more slow droughts, averaging 3.77 events compared with 3.25 in planted forests.

Even after accounting for differences in climate, the researchers found that forest characteristics themselves influenced drought risk. Compared with similar natural forests, planted forests showed a 7% higher risk of flash drought but a 12% lower risk of slow drought.

The contrasting responses appear to be linked to differences in forest structure. Natural forests typically contain a greater variety of tree species, trees of different ages, and more complex canopy layers. Older trees often have deeper root systems that can access water stored deeper in the soil, helping forests withstand sudden moisture shortages.

Planted forests, by contrast, are often younger and less diverse, with simpler structures and shallower roots. These characteristics make them more vulnerable when moisture levels decline rapidly during heatwaves or other extreme weather events. At the same time, their lower overall biomass and water demand may help reduce vulnerability to prolonged drought.

The differences were most pronounced in China's mid-temperate semi-humid regions and on the Tibetan Plateau, where large areas have been converted to planted forests.

Increasing forest area alone may not be enough to ensure long-term resilience. Instead, the study highlights the importance of forest composition and structure. Management strategies that promote mixed-species forests, greater structural diversity, and more natural forest characteristics could help reduce vulnerability to rapid drought events.

Moreover, unrestricted expansion of plantations in water-limited regions should also be cautioned against. Although planted forests generally experienced fewer slow droughts, their recovery was more protracted and challenging.

DOI Link:

https://doi.org/10.1016/j.fecs.2026.100462

---

Journal

Forest Ecosystems

DOI

10.1016/j.fecs.2026.100462 

Article Title

Understanding flash drought variations in China's planted and natural forests through machine learning counterfactual analysis