A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Qi Sun, Edoardo Cetin, Yujin Tang
Transformer-Squared is a self-adaptive LLM framework that adjusts only the singular components of weight matrices in real-time, using a two-pass dispatch and reinforcement-learned expert vectors to outperform LoRA with fewer parameters.
Lukas Radl, Michael Steiner, Mathias Parger, et al.
StopThePop introduces a hierarchical rasterization approach for 3D Gaussian Splatting that eliminates popping artifacts and view inconsistencies with minimal overhead, enabling faster rendering and reduced memory.
Yipeng Hu, Marc Modat, Eli Gibson, et al.
This paper introduces a weakly-supervised CNN approach for multimodal image registration that learns voxel-level transformations from anatomical labels, enabling real-time, fully-automated inference without labels.
Unknown
This paper proposes a UAV-deployed deep learning network for real-time multi-class damage detection, enhanced by model quantization techniques.
Hamid Mirshekali, Mohammad Reza Shadi, Fatemehsadat Ghanadi Ladani, et al.
This review synthesizes LLM applications across energy systems, consolidates evaluation and fine-tuning methods, and analyzes deployment constraints for real-time and edge adoption.
Unknown
This paper introduces a training framework for real-time hallucination detection in LLMs using internal states, along with the HELM benchmark for evaluation.
Eduardo Romera, Jose M. Álvarez, Luis M. Bergasa, et al.
ERFNet introduces a novel efficient residual factorized convolutional layer for real-time semantic segmentation, achieving high accuracy at over 83 FPS on Titan X and 7 FPS on Jetson TX1.
Sungjae Park, Shubham Tulsiani
πR² makes large pretrained action-chunking flow policies reactive and real-time via dual-channel conditioning and a latency-adaptive flow schedule.
Shuai Liu, Dongye Liu, Gautam Srivastava, et al.
Survey of correlation filter algorithms for real-time object tracking, covering background, key technologies, datasets, and reliability-based methods.
Martin Schoeberl, Sahar Abbaspour, Benny Åkesson, et al.
T-CREST proposes a time-predictable multi-core architecture optimized for worst-case execution time (WCET) rather than average-case performance, enabling static WCET analysis for real-time systems.
Clément Farabet, Camille Couprie, Laurent Najman, et al.
Proposes a multiscale convolutional network trained from raw pixels for dense scene labeling, achieving record accuracy and real-time performance.
Unknown
Faster R-CNN introduces a Region Proposal Network (RPN) that shares convolutional features with the detection network, enabling near real-time object detection.