Mathematical reasoning continues to be a critical challenge in large language model (LLM) development with significant interest. However, most of the cutting-edge progress in mathematical reasoning with LLMs has become closed-source due to lack of access to training data. This lack of data access limits researchers from understanding the impact of different choices for synthesizing and utilizing the data. With the goal of creating a high-quality finetuning (SFT) dataset for math reasoning, we conduct careful ablation experiments on data synthesis using the recently released Llama3.1 family of models. Our experiments show that: (a) solution format matters, with excessively verbose solutions proving detrimental to SFT performance, (b) data generated by a strong teacher outperforms on-policy data generated by a weak student model, (c) SFT is robust to low-quality solutions, allowing for imprecise data filtering, and (d) question diversity is crucial for achieving data scaling gains. Based on these insights, we create the OpenMathInstruct-2 dataset which consists of 14M question-solution pairs (≈ 600K unique questions), making it nearly eight times larger than the previous largest open-source math reasoning dataset. Finetuning the Llama-3.1-8B-Base using OpenMathInstruct-2 outperforms Llama3.1-8B-Instruct on MATH by an absolute 15.9% (51.9% → 67.8%). Finally, to accelerate the open-source efforts, we release the code, the finetuned models, and the OpenMathInstruct-2 dataset under a commercially permissive license.

Object detection (OD), a crucial vision task, remains challenged by the lack of large training datasets with precise object localization labels. In this work, we propose ALWOD, a new framework that addresses this problem by fusing active learning (AL) with weakly and semi-supervised object detection paradigms. Because the performance of AL critically depends on the model initialization, we propose a new auxiliary image generator strategy that utilizes an extremely small labeled set, coupled with a large weakly tagged set of images, as a warm-start for AL. We then propose a new AL acquisition function, another critical factor in AL success, that leverages the student-teacher OD pair disagreement and uncertainty to effectively propose the most informative images to annotate. Finally, to complete the AL loop, we introduce a new labeling task delegated to human annotators, based on selection and correction of model-proposed detections, which is both rapid and effective in labeling the informative images. We demonstrate, across several challenging benchmarks, that ALWOD significantly narrows the gap between the ODs trained on few partially labeled but strategically selected image instances and those that rely on the fullylabeled data. Our code is publicly available on https://github.com/seqam-lab/ALWOD.

In this paper we introduce the Embedded Vision Engine (EVE), a novel vision accelerator designed to complement digital signal processors on low- and mid-level vision algorithms. We illustrate EVE's performance on three important mid-level vision functions, the Hough transform for circles, integral image and calculating the Rotation invariant Binary Robust Independent Elementary Feature (RBRIEF) descriptor. EVE can execute these functions 4 times faster than the state-of-the-art digital signal processor. With other vision functions accelerated 3–12X, the acceleration of these popular vision functions contribute to the application level speed up of 5x on common automotive vision applications. This demonstrates that EVE, while working at powers comparable to the power of a DSP core, can deliver significant boost in performance.

We present an automotive-grade, real-time, vision-based driver state monitor. Upon detecting and tracking the driver's facial features, the system analyzes eye-closures and head pose to infer his/her fatigue or distraction. This information is used to warn the driver and to modulate the actions of other safety systems. The purpose of this monitor is to increase road safety by preventing drivers from falling asleep or from being overly distracted, and to improve the effectiveness of other safety systems.

In this correspondence, we present a fast thresholded linear convolution representation of morphological operations. The thresholded linear convolution representation of dilation and erosion is first proposed. A comparison of the efficiency of the direct implementation of morphological operations and the thresholded linear convolution representation of morphological operations is subsequently conducted. Mathematical morphology has emerged as a powerful new tool for image processing.