Efficient deployment of small language models (SLMs) is essential for numerous real-world applications with stringent latency constraints.While previous work on SLM design has primarily focused on reducing the number of parameters to achieve parameter-optimal SLMs, parameter efficiency does not necessarily translate into proportional real-device speed-ups. This work aims to identify the key determinants of SLMs' real-device latency and offer generalizable principles and methodologies for SLM design and training when real-device latency is the primary consideration. Specifically, we identify two central architectural factors: depth–width ratios and operator choices. The former is crucial for small-batch-size latency, while the latter affects both latency and large-batch-size throughput. In light of this, we first study latency-optimal depth–width ratios, with the key finding that although deep–thin models generally achieve better accuracy under the same parameter budget, they may not lie on the accuracy–latency trade-off frontier. Next, we explore emerging efficient attention alternatives to evaluate their potential as candidate building operators. Using the identified promising operators, we construct an evolutionary search framework to automatically discover latency-optimal combinations of these operators within hybrid SLMs, thereby advancing the accuracy–latency frontier. In addition to architectural improvements, we further enhance SLM training using a weight normalization technique that enables more effective weight updates and improves final convergence. This technique can serve as a generalizable component for future SLMs. Combining these methods, we introduce a new family of hybrid SLMs, called Nemotron-Flash, which significantly advances the accuracy–efficiency frontier of state-of-the-art SLMs, e.g., achieving over +5.5\% average accuracy, 1.3$\times$/1.9$\times$ lower latency, and 18.7$\times$/45.6$\times$ higher throughput compared to Qwen3-1.7B/0.6B, respectively.

Continuous deep learning architectures enable learning of flexible probabilistic models for predictive modeling as neural ordinary differential equations (ODEs), and for generative modeling as continuous normalizing flows. In this work, we design a framework to decipher the internal dynamics of these continuous depth models by pruning their network architectures. Our empirical results suggest that pruning improves generalization for neural ODEs in generative modeling. We empirically show that the improvement is because pruning helps avoid mode-collapse and flatten the loss surface. Moreover, pruning finds efficient neural ODE representations with up to 98% less parameters compared to the original network, without loss of accuracy. We hope our results will invigorate further research into the performance-size trade-offs of modern continuous-depth models.

We present a novel global compression framework for deep neural networks that automatically analyzes each layer to identify the optimal per-layer compression ratio, while simultaneously achieving the desired overall compression. Our algorithm hinges on the idea of compressing each convolutional (or fully-connected) layer by slicing its channels into multiple groups and decomposing each group via low-rank decomposition. At the core of our algorithm is the derivation of layer-wise error bounds from the Eckart–Young–Mirsky theorem. We then leverage these bounds to frame the compression problem as an optimization problem where we wish to minimize the maximum compression error across layers and propose an efficient algorithm towards a solution. Our experiments indicate that our method outperforms existing low-rank compression approaches across a wide range of networks and data sets. We believe that our results open up new avenues for future research into the global performance-size trade-offs of modern neural networks.

Recent advances in autonomous driving have raised the problem of safety to the forefront and incentivized research into establishing safety guarantees. In this paper, we propose a safety verification framework as a safety standard for driving controllers with full or shared autonomy based on compositional and contract-based principles. Our framework enables us to synthesize safety guarantees over entire road networks by first building a library of locally verified models, and then composing local models together to verify the entire network. Composition is achieved using assume-guarantee contracts that are synthesized concurrently during verification. Thus, we can reuse local models within and across networks, add additional models to cover local road geometries without re-verifying the entire library, and perform all computations in a parallel and distributed way, which enables computational tractability. Furthermore, we employ controller contracts such that any controller satisfying them can be certified safe. We demonstrate the practical effectiveness of our framework by certifying controllers over parts of the Manhattan road network.