MIT researchers have developed a new explainable AI method that improves the accuracy and transparency of computer vision models, helping users trust AI predictions in healthcare and autonomous driving.

Researchers at MIT have developed a new approach to make computer vision models more transparent, offering a potential boost to trust and accountability in safety-critical applications such as medical diagnosis and autonomous driving.

In a media statement, the researchers said the method improves on a widely used explainability technique known as concept bottleneck modeling, which enables AI systems to show the human-understandable concepts behind a prediction. The new approach is designed to produce clearer explanations while also improving prediction accuracy.

Why explainable AI matters

In areas such as health care, users often need more than just a model’s output. They want to understand why a system arrived at a particular conclusion before deciding whether to rely on it. Concept bottleneck models attempt to address that need by forcing an AI system to make predictions through a set of intermediate concepts that humans can interpret.

For example, when analysing a medical image for melanoma, a clinician might define concepts such as “clustered brown dots” or “variegated pigmentation.” The model would first identify those concepts and then use them to arrive at its final prediction.

But the researchers said pre-defined concepts can sometimes be too broad, irrelevant or incomplete for a specific task, limiting both the quality of explanations and the model’s performance. To overcome that, the MIT team developed a method that extracts concepts the model has already learned during training and then compels it to use those concepts when making decisions.

The approach relies on two specialised machine-learning models. One extracts the most relevant internal features learned by the target model, while the other translates them into plain-language concepts that humans can understand. This makes it possible to convert a pretrained computer vision model into one capable of explaining its reasoning through interpretable concepts.

“In a sense, we want to be able to read the minds of these computer vision models. A concept bottleneck model is one way for users to tell what the model is thinking and why it made a certain prediction. Because our method uses better concepts, it can lead to higher accuracy and ultimately improve the accountability of black-box AI models,” Antonio De Santis, lead author of the study, said in a media statement.

De Santis is a graduate student at Polytechnic University of Milan and carried out the research while serving as a visiting graduate student at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL). The paper was co-authored by Schrasing Tong, Marco Brambilla of Polytechnic University of Milan, and Lalana Kagal of CSAIL. The research will be presented at the International Conference on Learning Representations.

Concept bottleneck models have gained attention as a way to improve AI explainability by introducing an intermediate reasoning step between an input image and the final output. In one example, a bird-classification model might identify concepts such as “yellow legs” and “blue wings” before predicting a barn swallow.

However, the researchers noted that these concepts are often generated in advance by humans or large language models, which may not always match the needs of the task. Even when a model is given a fixed concept set, it can still rely on hidden information not visible to users, a challenge known as information leakage.

“These models are trained to maximize performance, so the model might secretly use concepts we are unaware of,” De Santis said in a media statement.

The team’s solution was to tap into the knowledge the model had already acquired from large volumes of training data. Using a sparse autoencoder, the method isolates the most relevant learned features and reconstructs them into a small number of concepts. A multimodal large language model then describes each concept in simple language and labels the training images by marking which concepts are present or absent.

The annotated dataset is then used to train a concept bottleneck module, which is inserted into the target model. This forces the model to make predictions using only the extracted concepts.

The researchers said one of the biggest challenges was ensuring that the automatically identified concepts were both accurate and understandable to humans. To reduce the risk of hidden reasoning, the model is limited to just five concepts for each prediction, encouraging it to focus only on the most relevant information and making the explanation easier to follow.

When tested against state-of-the-art concept bottleneck models on tasks including bird species classification and skin lesion identification, the new method delivered the highest accuracy while also producing more precise explanations, according to the researchers. It also generated concepts that were more relevant to the images in the dataset.

Still, the team acknowledged that the broader challenge of balancing accuracy and interpretability remains unresolved.

“We’ve shown that extracting concepts from the original model can outperform other CBMs, but there is still a tradeoff between interpretability and accuracy that needs to be addressed. Black-box models that are not interpretable still outperform ours,” De Santis said in a media statement.

Looking ahead, the researchers plan to explore ways to further reduce information leakage, possibly by adding additional concept bottleneck modules. They also aim to scale up the method by using a larger multimodal language model to annotate a larger training dataset, which could improve performance further.

This latest work adds to growing efforts to make AI systems not only more powerful, but also more understandable in domains where trust can be as important as accuracy.

Stretchable material heat conduction has taken a major leap forward as engineers at MIT have developed a polymer that can rapidly and reversibly switch how it conducts heat simply by being stretched. The discovery opens new possibilities for adaptive cooling technologies in clothing, electronics, and building infrastructure.

Engineers at the Massachusetts Institute of Technology have developed a new polymer material that can rapidly and reversibly switch how it conducts heat—simply by being stretched.

The research shows that a commonly used soft polymer, known as an olefin block copolymer (OBC), can more than double its thermal conductivity when stretched, shifting from heat-handling behaviour similar to plastic to levels closer to marble. When the material relaxes back to its original form, its heat-conducting ability drops again, returning to its plastic-like state.

The transition happens extremely fast—within just 0.22 seconds—making it the fastest thermal switching ever observed in a material, according to the researchers.

The findings open up possibilities for adaptive materials that respond to temperature changes in real time, with potential applications ranging from cooling fabrics and wearable technology to electronics, buildings, and infrastructure.

The research team initially began studying the material while searching for more sustainable alternatives to spandex, a petroleum-based elastic fabric that is difficult to recycle. During mechanical testing, the researchers noticed unexpected changes in how the polymer handled heat as it was stretched and released.

A new direction for adaptive materials

“We need materials that are inexpensive, widely available, and able to adapt quickly to changing environmental temperatures,” said Svetlana Boriskina, principal research scientist in MIT’s Department of Mechanical Engineering, in a media statement. She explained that the discovery of rapid thermal switching in this polymer creates new opportunities to design materials that actively manage heat rather than passively resisting it.

The research team initially began studying the material while searching for more sustainable alternatives to spandex, a petroleum-based elastic fabric that is difficult to recycle. During mechanical testing, the researchers noticed unexpected changes in how the polymer handled heat as it was stretched and released.

“What caught our attention was that the material’s thermal conductivity increased when stretched and decreased again when relaxed, even after thousands of cycles,” said Duo Xu, a co-author of the study, in a media statement. He added that the effect was fully reversible and occurred while the material remained largely amorphous, which contradicted existing assumptions in polymer science.

The discovery demonstrates how stretchable material heat conduction can be actively controlled in real time, allowing materials to respond dynamically to temperature changes.

How stretching unlocks heat flow

At the microscopic level, most polymers consist of tangled chains of carbon atoms that block heat flow. The MIT team found that stretching the olefin block copolymer temporarily straightens these tangled chains and aligns small crystalline regions, creating clearer pathways for heat to travel through the material.

“This gives the material the ability to toggle its heat conduction thousands of times without degrading

Unlike earlier work on polyethylene—where similar alignment permanently increased thermal conductivity—the new material does not crystallise under strain. Instead, its internal structure switches back and forth between straightened and tangled states, allowing repeated and reversible thermal switching.

“This gives the material the ability to toggle its heat conduction thousands of times without degrading,” Xu said.

From smart clothing to cooler electronics

The researchers say the material could be engineered into fibres for clothing that normally retain heat but instantly dissipate excess warmth when stretched. Similar concepts could be applied to electronics, laptops, and buildings, where materials could respond dynamically to overheating without external cooling systems.

“The difference in heat dissipation is similar to the tactile difference between touching plastic and touching marble,” Boriskina said in a media statement, highlighting how noticeable the effect can be.

The team is now working on optimising the polymer’s internal structure and exploring related materials that could produce even larger thermal shifts.

“If we can further enhance this effect, the industrial and societal impact could be substantial,” Boriskina said.

Researchers say advances in stretchable material heat conduction could significantly influence future designs of smart textiles, electronics cooling, and energy-efficient buildings.

The study has been published in the journal Advanced Materials. The authors include researchers from MIT and the Southern University of Science and Technology in China.

Researchers say advances in stretchable material heat conduction could significantly influence future designs of smart textiles, electronics cooling, and energy-efficient buildings.

Physicists working at the CERN(The European Organization for Nuclear Research) have reported the first direct experimental evidence that quark–gluon plasma—the primordial matter that filled the universe moments after the Big Bang—behaves like a true liquid.

Using heavy-ion collisions at the Large Hadron Collider, researchers recreated the extreme conditions of the early universe and observed that quarks moving through this plasma generate wake-like patterns, similar to ripples trailing a duck across water.

The study, led by physicists from the Massachusetts Institute of Technology, shows that the quark–gluon plasma responds collectively, flowing and splashing rather than scattering randomly.

“It has been a long debate in our field, on whether the plasma should respond to a quark,” said Yen-Jie Lee in a media statement. “Now we see the plasma is incredibly dense, such that it is able to slow down a quark, and produces splashes and swirls like a liquid. So quark-gluon plasma really is a primordial soup.”

Quark–gluon plasma is believed to be the first liquid to have existed in the universe and the hottest ever observed, reaching temperatures of several trillion degrees Celsius. It is also considered a near-perfect liquid, flowing with almost no resistance.

To isolate the wake produced by a single quark, the team developed a new experimental technique. Instead of tracking pairs of quarks and antiquarks—whose effects can overlap—they identified rare collision events that produced a single quark traveling in the opposite direction of a Z boson. Because a Z boson interacts weakly with its surroundings, it acts as a clean marker, allowing scientists to attribute any observed plasma ripples solely to the quark.

“We have figured out a new technique that allows us to see the effects of a single quark in the QGP, through a different pair of particles,” Lee said.

Analysing data from around 13 billion heavy-ion collisions, the researchers identified roughly 2,000 Z-boson events. In these cases, they consistently observed fluid-like swirls in the plasma opposite to the Z boson’s direction—clear signatures of quark-induced wakes.

The results align with theoretical predictions made by MIT physicist Krishna Rajagopal, whose hybrid model suggested that quarks should drag plasma along as they move through it.

“This is something that many of us have argued must be there for a good many years, and that many experiments have looked for,” Rajagopal said.

“We’ve gained the first direct evidence that the quark indeed drags more plasma with it as it travels,” Lee added. “This will enable us to study the properties and behavior of this exotic fluid in unprecedented detail.”

The research was carried out by members of the CMS Collaboration using the Compact Muon Solenoid detector at CERN. The open-access study has been published in the journal Physics Letters B.