Automatic Heuristic Design (AHD) is an active research area due to its utility in solving complex search and NP-hard combinatorial optimization problems in the real world. The recent advancements in Large Language Models (LLMs) introduce new possibilities by coupling LLMs with evolutionary computation to automatically generate heuristics, known as LLM-based Evolutionary Program Search (LLM-EPS). While previous LLM-EPS studies obtained great performance on various tasks, there is still a gap in understanding the properties of heuristic search spaces and achieving a balance between exploration and exploitation, which is a critical factor in large heuristic search spaces. In this study, we address this gap by proposing two diversity measurement metrics and perform an analysis on previous LLM-EPS approaches, including FunSearch, EoH, and ReEvo. Results on black-box AHD problems reveal that while EoH demonstrates higher diversity than FunSearch and ReEvo, its objective score is unstable. Conversely, ReEvo's reflection mechanism yields good objective scores but fails to optimize diversity effectively. With this finding in mind, we introduce HSEvo, an adaptive LLM-EPS framework that maintains a balance between diversity and convergence with a harmony search algorithm. Through experimentation, we find that HSEvo achieved high diversity indices and good objective scores while remaining cost-effective. These results underscore the importance of balancing exploration and exploitation and understanding heuristic search spaces in designing frameworks in LLM-EPS.

Neural Architecture Search (NAS) is a promising approach to automate the design of neural network architectures, which can find architectures that perform better than manually designed ones. Hardware-aware NAS is a real-world application of NAS where the architectures found also need to satisfy certain requirements for the deployment of specific devices. Despite the practical importance, hardware-aware NAS still receives a lack of attention from the community. Existing research mostly focuses on the search space with a limited number of architectures, reducing the search process to finding the optimal hyperparameters. In addition, the performance evaluation of found networks is resources-intensive, which can severely hinder reproducibility. In this work, we propose a genetic algorithm approach to the hardware-aware NAS problem, incorporating a latency filtering selection to guarantee the latency validity of candidate solutions. We also introduce an extended search space that can cover various existing architectures from previous research. To speed up the search process, we also present a method to estimate the latency of candidate networks and a training-free performance estimation method to quickly evaluate candidate networks. Our experiments demonstrate that our method achieves competitive performance with state-of-the-art networks while maintaining lower latency with less computation requirements for searching.

Real-world continuous control problems often require optimizing for multiple conflicting objectives. Various works in multi-objective reinforcement learning have been conducted to tackle such issues and obtained impressive performance. At the same time, evolutionary algorithms (EAs), which are extensively used in multi-objective optimization, have recently been demonstrated their competitiveness to RL algorithms, including multi-objective control for environments with discrete action spaces. However, using EAs for multi-objective continuous robot control is still an under-explored topic. For the single-objective setting, the Proximal Distilled Evolutionary Reinforcement Learning (PDERL) framework succeeds in combining the robustness of EA and the efficiency of RL methods. In this work, we bring the strengths of PDERL to the multiobjective realm to create the novel multi-objective PDERL framework called MOPDERL that consists of a warm-up stage and an evolution stage. In particular, MOPDERL collaboratively optimizes policies for each separate objective in the warm-up stage, and then exchanges that knowledge for further policy improvement during the multi-objective evolution stage. We benchmark MOPDERL on six MuJoCo robot locomotion environments, which have been modified for the multi-objective context. The results show that MOPDERL produces better-quality Pareto fronts and higher metric scores than the state-of-the-art PGMORL algorithm across five out of six environments.

Neural Architecture Search (NAS) has received much attraction from the research community in recent years. However, due to the massive amount of computational resources required, it is still infeasible to employ NAS into research and production in small labs and companies. Recent methods in NAS that aim to speed up the evaluation process are often limit themselves in other aspects, such as the search space that NAS operates on. In this work, we propose TF-GeneNAS, an evolution-based training-free NAS approach with a dynamic search space and search strategy based on Gene Expression Programming. We conduct experiments on three tasks in both Computer Vision and Natural Language Processing domains to demonstrates the effectiveness of our method. With only 3 CPU days of searching needed, TF-GeneNAS can find network architectures with better performance than previous evolution-based methods, which can require days of GPU resources, thus significantly lower the cost of searching. We also perform further studies to show the impact of our training-free estimation strategy on the NAS process. We hope that our promising results can encourage further research into more efficient evolution-based NAS methods.

In this paper, we introduce a high-quality and large-scale benchmark dataset for English-Vietnamese speech translation with 508 audio hours, consisting of 331K triplets of (sentencelengthed audio, English source transcript sentence, Vietnamese target subtitle sentence).We also conduct empirical experiments using strong baselines and find that the traditional "Cascaded" approach still outperforms the modern "End-to-End" approach.To the best of our knowledge, this is the first largescale English-Vietnamese speech translation study.We hope both our publicly available dataset and study can serve as a starting point for future research and applications on English-Vietnamese speech translation.

We introduce a high-quality and large-scale Vietnamese-English parallel dataset of 3.02M sentence pairs, which is 2.9M pairs larger than the benchmark Vietnamese-English machine translation corpus IWSLT15.We conduct experiments comparing strong neural baselines and well-known automatic translation engines on our dataset and find that in both automatic and human evaluations: the best performance is obtained by fine-tuning the pretrained sequence-to-sequence denoising autoencoder mBART.To our best knowledge, this is the first large-scale Vietnamese-English machine translation study.We hope our publicly available dataset and study can serve as a starting point for future research and applications on Vietnamese-English machine translation.We release our dataset at: https:// github.com/VinAIResearch/PhoMT.