Reliability Check: Technion Researchers Develop a New Method to Detect AI Errors and Hallucinations

As large language models (LLMs) become increasingly integrated into everyday applications—from translation and content generation to coding and scientific research—ensuring their reliability has emerged as a critical challenge. Researchers at the Technion – Israel Institute of Technology have developed an innovative approach for identifying errors, hallucinations, and other undesirable behaviors in AI systems without requiring a full understanding of how the models work internally.

The research was led by Dr. Haggai Maron of the Andrew and Erna Viterbi Faculty of Electrical and Computer Engineering, together with Ph.D. student Guy Bar-Shalom, postdoctoral researcher Dr. Fabrizio Frasca, Prof. Ran El-Yaniv, and Dr. Yftah Ziser of the University of Groningen and NVIDIA. Their findings were presented in three papers accepted to leading machine learning conferences: NeurIPS 2025, AAAI 2026, and ICLR 2026.

Rather than attempting to fully interpret the complex internal mechanisms of large language models, the researchers developed a practical and computationally efficient framework that analyzes the models’ internal computations. By training lightweight machine-learning systems on these internal signals, the method can uncover hidden information that reveals when a model is likely to make mistakes, generate inaccurate content, ignore instructions, or behave unexpectedly.

The researchers demonstrate that it is possible to monitor, diagnose, and predict risks in AI-generated outputs externally and at low cost, enabling users to supervise and control model behavior without access to the model’s training process or a complete understanding of its internal workings.

The work addresses one of the most pressing questions in artificial intelligence: how to determine when an AI system is producing unreliable information. The new methods could support the development of warning mechanisms, quality-assurance tools, and safety standards for AI systems deployed in high-stakes fields such as medicine, education, scientific research, and regulation.

The studies are part of a broader research program in Dr. Maron’s laboratory exploring how information embedded in trained models—including their weights and training signals—can be used to improve the safety, reliability, and interpretability of artificial intelligence systems.

Researchers at the Technion – Israel Institute of Technology have developed a new technology called Time to Move (TTM) that enables users to create realistic AI-generated videos simply by controlling motion with mouse movements. The method provides precise control over how objects and characters move over time, addressing one of the major challenges in AI video generation.

Unlike existing approaches, TTM does not require retraining video models or access to extensive computational resources, making advanced video creation more accessible to a wider range of users. The technology can be integrated into existing video-generation systems as a plug-in and operates without additional computational cost.

Developed by Dr. Or Litany, Prof. Ron Kimmel, and students Asaf Singer, Noam Rotstein, and Amir Mann from the Technion’s Henry and Marilyn Taub Faculty of Computer Science, the work was presented at the International Conference on Learning Representations (ICLR) 2026. The key innovation, known as dual-clock denoising, balances adherence to user instructions with natural-looking motion. Experimental results show that TTM matches or outperforms training-based methods in motion accuracy and realism while offering additional capabilities, including object appearance editing and the addition of new objects to a scene.

The research represents a significant step toward intuitive, controllable, and widely accessible tools for AI-powered video generation.

Zhejiang University researchers introduce a universal 'best-matched absorption' criterion to select optimal near-infrared (NIR) imaging windows for fluorescence wide-field microscopy. This work solves the critical gap of scattered light background in high numerical aperture (NA) systems, clarifies how system NA, imaging depth and probe brightness shift the optimal window, and offers tailored strategies for complex in vivo imaging scenarios, greatly improving image contrast and penetration for biomedical research and clinical applications.

A review accepted by PhotoniX Life examines how deep learning is transforming cell segmentation in live-cell microscopy. The article traces the field from thresholding and watershed algorithms to U-Net, Mask R-CNN, StarDist, Cellpose, Mesmer, transformer-based models, and emerging generalist segmentation frameworks. It highlights why segmentation is central to quantitative cell biology, where current models succeed, and what remains difficult, including cross-modality generalization, 3D volumetric data, annotation scarcity, user-friendly deployment, and organelle-level analysis.

The study reconceptualizes classroom assessment as a driver of teacher reflection and professional growth, introducing a structured reflection method that makes teachers' thinking visible and open to change.

MIT researchers created a technique that captures chemical arrangements across materials to improve predictions of how metal alloys and other complex materials will behave.