Speculative decoding has emerged as a prominent alternative to autoregressive decoding for expediting inference in large language models (LLMs). However, prevailing assumptions often focus solely on latency reduction, neglecting the computational expenses. In this paper, we present Speculate Less, validate More (SLiM), a speculative decoding enhancement to reduce the speculation set while validating more effective tokens. SLiM is designed to mitigate LLMs’ computation costs associated with the token verification by introducing hypothesis reduction based on a fast posterior estimation. It consistently surpasses counterparts lacking cost reduction across a spectrum from CPU to GPU. Our evaluation with diverse conversational datasets shows that SLiM can achieve a substantial 70% reduction in FLOPs while generating more effective predictions on top of prior arts.

Singular Value Decomposition (SVD) or its weighted variants has significantly progressed in compressing language models. Previous works assume the same importance for all operations and assign the same number of ranks for different layers in a language model. However, such a uniform rank selection is sub-optimal since different operations (layers) have non-uniform demand in capacity. In other words, a desired SVD strategy should allocate more ranks for important operations and vice versa. However, a globally-optimized selection of ranks for neural networks is still an open problem, and this is a non-trivial challenge since the selection is discrete. In this work, we propose a novel binary masking mechanism for optimizing the number of ranks in a differentiable framework. Our strategy uses a novel regularization to enable the masking to comply with the SVD property where the ranks have sorted singular values. The experiments examined both types of language models, encoder-only and decoder-only models, including large language models like LLaMA. Our compressed model achieves much better accuracy than previous SVD and their SOTA variants. More interestingly, our method retains significantly better accuracy with zero or limited fine-tuning, proving the substantial advantage of adaptive rank selection.

Traditional language models operate autoregressively, i.e., they predict one token at a time. Rapid explosion in model sizes has resulted in high inference times. In this work, we propose DynaMo, a suite of multi-token prediction language models that reduce net inference times. Our models *dynamically* predict multiple tokens based on their confidence in the predicted joint probability distribution. We propose a lightweighttechnique to train these models, leveraging the weights of traditional autoregressive counterparts. Moreover, we propose novel ways to enhance the estimated joint probability to improve text generation quality, namely co-occurrence weighted masking and adaptive thresholding. We also propose systematic qualitative and quantitative methods to rigorously test the quality of generated text for non-autoregressive generation. One of the models in our suite, DynaMo-7.3B-T3, achieves same-quality generated text as the baseline (Pythia-6.9B) while achieving 2.57× speed-up with only 5.87% and 2.67% parameter and training time overheads, respectively.

Matrix decomposition methods, such as Singular Value Decomposition (SVD) and its importance-weighted variants, have been widely used for compressing Transformer-based language models. While importance-weighted decomposition methods alleviate the strong assumption of equal importance for each parameter in SVD, they still rely on two fundamental assumptions: 1) unchanged importance distribution during further fine-tuning, 2) equal importance across weight matrices in different layers. Furthermore, these methods necessitate a well-trained task-specific model as the starting point and require additional fine-tuning after compression. In this work, we proposed RankDyna, a matrix decomposition method that enables dynamic rank resource allocation among matrices across different layers during the training process. Starting from a general pre-trained model, RankDyna accomplishes the dual goals of compression and adaptation to the downstream task, all within a single round of fine-tuning. The extensive evaluations demonstrate that RankDyna can outperform current SOTA methods under various parameter budget levels, and the advantage of RankDyna is further enhanced with higher compression rates.

Singular value decomposition (SVD) is one of the most popular compression methods that approximate a target matrix with smaller matrices. However, standard SVD treats the parameters within the matrix with equal importance, which is a simple but unrealistic assumption. The parameters of a trained neural network model may affect the task performance unevenly, which suggests non-equal importance among the parameters. Compared to SVD, the decomposition method aware of parameter importance is the more practical choice in real cases. Unlike standard SVD, weighed value decomposition is a non-convex optimization problem that lacks a closed-form solution. We systematically investigated multiple optimization strategies to tackle the problem and examined our method by compressing Transformer-based language models.Further, we designed a metric to predict when the SVD may introduce a significant performance drop, for which our method can be a rescue strategy.The extensive evaluations demonstrate that our method can perform better than current SOTA methods in compressing Transformer-based language models.

Intent classification is a major task in spoken language understanding (SLU). Since most models are built with pre-collected in-domain (IND) training utterances, their ability to detect unsupported out-of-domain (OOD) utterances has a critical effect in practical use. Recent works have shown that using extra data and labels can improve the OOD detection performance, yet it could be costly to collect such data. This paper proposes to train a model with only IND data while supporting both IND intent classification and OOD detection. Our method designs a novel domain-regularized module (DRM) to reduce the overconfident phenomenon of a vanilla classifier, achieving a better generalization in both cases. Besides, DRM can be used as a drop-in replacement for the last layer in any neural network-based intent classifier, providing a low-cost strategy for a significant improvement. The evaluation on four datasets shows that our method built on BERT and RoBERTa models achieves state-of-the-art performance against existing approaches and the strong baselines we created for the comparisons.

Domain classification is the fundamental task in natural language understanding (NLU), which often requires fast accommodation to new emerging domains. This constraint makes it impossible to retrain all previous domains, even if they are accessible to the new model. Most existing continual learning approaches suffer from low accuracy and performance fluctuation, especially when the distributions of old and new data are significantly different. In fact, the key real-world problem is not the absence of old data, but the inefficiency to retrain the model with the whole old dataset. Is it potential to utilize some old data to yield high accuracy and maintain stable performance, while at the same time, without introducing extra hyperparameters? In this paper, we proposed a hyperparameter-free continual learning model for text data that can stably produce high performance under various environments. Specifically, we utilize Fisher information to select exemplars that can “record” key information of the original model. Also, a novel scheme called dynamical weight consolidation is proposed to enable hyperparameter-free learning during the retrain process. Extensive experiments demonstrate baselines provide fluctuated performance which makes them useless in practice. On the contrary, our proposed model significantly and consistently outperforms the best state-of-the-art method by up to 20% in average accuracy, and each of its component contributes effectively to overall performance.