The recent success of specialized Large Language Models (LLMs) in domains such as mathematical reasoning and coding has led to growing interest in methods for merging these expert LLMs into a unified Mixture-of-Experts (MoE) model, with the goal of enhancing performance in each domain while retaining effectiveness on general tasks. However, effective merging of expert models remains an open challenge, especially for models with highly divergent weight parameters or different architectures. State-of-the-art MoE merging methods only work with homogeneous model architectures and rely on simple unweighted averaging to merge expert layers, which does not address parameter interference and requires extensive fine-tuning of the merged MoE to restore performance. To address these limitations, this paper introduces new MoE merging techniques, including strategies to mitigate parameter interference, routing heuristics to reduce the need for MoE fine-tuning, and a novel method for merging experts with different architectures. Extensive experiments across multiple domains demonstrate the effectiveness of our proposed methods, reducing fine-tuning costs, improving performance over state-of-the-art methods, and expanding the applicability of MoE merging.

Task-oriented semantic parsing has drawn a lot of interest from the NLP community, and especially the voice assistant industry as it enables representing the meaning of user requests with arbitrarily nested semantics, including multiple intents and compound entities. SOTA models are large seq2seq transformers and require hundreds of thousands of annotated examples to be trained. However annotating such data to bootstrap new domains or languages is expensive and error-prone, especially for requests made of nested semantics. In addition large models easily break the tight latency constraints imposed in a user-facing production environment. As part of this work we explore leveraging external knowledge to improve model accuracy in low-resource and low-compute settings. We demonstrate that using knowledge-enhanced encoders inside seq2seq models does not result in performance gains by itself, but jointly learning to uncover entities in addition to the parse generation is a simple yet effective way of improving performance across the board. We show this is especially true in the low-compute scarce-data setting and for entity-rich domains, with relative gains up to 74.48% on the TOPv2 dataset.

This paper presents a production Semi-Supervised Learning (SSL) pipeline based on the student-teacher framework, which leverages millions of unlabeled examples to improve Natural Language Understanding (NLU) tasks. We investigate two questions related to the use of unlabeled data in production SSL context: 1) how to select samples from a huge unlabeled data pool that are beneficial for SSL training, and 2) how does the selected data affect the performance of different state-of-the-art SSL techniques. We compare four widely used SSL techniques, Pseudo-label (PL), Knowledge Distillation (KD), Virtual Adversarial Training (VAT) and Cross-View Training (CVT) in conjunction with two data selection methods including committee-based selection and submodular optimization based selection. We further examine the benefits and drawbacks of these techniques when applied to intent classification (IC) and named entity recognition (NER) tasks, and provide guidelines specifying when each of these methods might be beneficial to improve large scale NLU systems.

Virtual adversarial training (VAT) is a powerful technique to improve model robustness in both supervised and semi-supervised settings. It is effective and can be easily adopted on lots of image classification and text classification tasks. However, its benefits to sequence labeling tasks such as named entity recognition (NER) have not been shown as significant, mostly, because the previous approach can not combine VAT with the conditional random field (CRF). CRF can significantly boost accuracy for sequence models by putting constraints on label transitions, which makes it an essential component in most state-of-the-art sequence labeling model architectures. In this paper, we propose SeqVAT, a method which naturally applies VAT to sequence labeling models with CRF. Empirical studies show that SeqVAT not only significantly improves the sequence labeling performance over baselines under supervised settings, but also outperforms state-of-the-art approaches under semi-supervised settings.

Adversarial training (AT) has shown strong regularization effects on deep learning algorithms by introducing small input perturbations to improve model robustness. In language tasks, adversarial training brings word-level robustness by adding input noise, which is beneficial for text classification. However, it lacks sufficient contextual information enhancement and thus is less useful for sequence labelling tasks such as chunking and named entity recognition (NER). To address this limitation, we propose masked adversarial training (MAT) to improve robustness from contextual information in sequence labelling. MAT masks or replaces some words in the sentence when computing adversarial loss from perturbed inputs and consequently enhances model robustness using more context-level information. In our experiments, our method shows significant improvements on accuracy and robustness of sequence labelling. By further incorporating with ELMo embeddings, our model achieves better or comparable results to state-of-the-art on CoNLL 2000 and 2003 benchmarks using much less parameters.