In the era of large language models, Mixture-of-Experts (MoE) is a promising architecture for managing computational costs when scaling up model parameters. However, conventional MoE architectures like GShard, which activate the top-K out of N experts, face challenges in ensuring expert specialization, i.e. each expert acquires non-overlapping and focused knowledge. In response, we propose the DeepSeekMoE architecture towards ultimate expert specialization. It involves two principal strategies: (1) finely segmenting the experts into mN ones and activating mK from them, allowing for a more flexible combination of activated experts; (2) isolating K_s experts as shared ones, aiming at capturing common knowledge and mitigating redundancy in routed experts. Starting from a modest scale with 2B parameters, we demonstrate that DeepSeekMoE 2B achieves comparable performance with GShard 2.9B, which has 1.5 × expert parameters and computation. In addition, DeepSeekMoE 2B nearly approaches the performance of its dense counterpart with the same number of total parameters, which sets the upper bound of MoE models. Subsequently, we scale up DeepSeekMoE to 16B parameters and show that it achieves comparable performance with DeepSeek 7B and LLaMA2 7B, with only about 40% of computations.

In machine translation evaluation, a good candidate translation can be regarded as a paraphrase of the reference. We notice that some words are always copied during paraphrasing, which we call copy knowledge. Considering the stability of such knowledge, a good candidate translation should contain all these words appeared in the reference sentence. Therefore, in this participation of the WMT’2018 metrics shared task we introduce a simple statistical method for copy knowledge extraction, and incorporate it into Meteor metric, resulting in a new machine translation metric Meteor++. Our experiments show that Meteor++ can nicely integrate copy knowledge and improve the performance significantly on WMT17 and WMT15 evaluation sets.

Ontologies are powerful to support semantic based applications and intelligent systems. While ontology learning are challenging due to its bottleneck in handcrafting structured knowledge sources and training data. To address this difficulty, many researchers turn to ontology enrichment and population using external knowledge sources such as DBpedia. In this paper, we propose a method using DBpedia in a different manner. We utilize relation instances in DBpedia to supervise the ontology learning procedure from unstructured text, rather than populate the ontology structure as a post-processing step. We construct three language resources in areas of computer science: enriched Wikipedia concept tree, domain ontology, and gold standard from NSFC taxonomy. Experiment shows that the result of ontology learning from corpus of computer science can be improved via the relation instances extracted from DBpedia in the same field. Furthermore, making distinction between the relation instances and applying a proper weighting scheme in the learning procedure lead to even better result.