Recent advancements in Large Language Models (LLMs) have set themselves apart with their exceptional performance in complex language modelling tasks. However, these models are also known for their significant computational and storage requirements, primarily due to the quadratic computation complexity of softmax attention. To mitigate this issue, linear attention has been designed to reduce the quadratic space-time complexity that is inherent in standard transformers. In this work, we embarked on a comprehensive exploration of three key components that substantially impact the performance of the Gated Linear Attention module: feature maps, normalization, and the gating mechanism. We developed a feature mapping function to address some crucial issues that previous suggestions overlooked. Then we offered further rationale for the integration of normalization layers to stabilize the training process. Moreover, we explored the saturation phenomenon of the gating mechanism and augmented it with a refining module. We conducted extensive experiments and showed our architecture outperforms previous Gated Linear Attention mechanisms in extensive tasks including training from scratch and post-linearization with continual pre-training.

This paper addresses the challenge of train-short-test-long (TSTL) scenarios in Large Language Models (LLMs) equipped with Rotary Position Embedding (RoPE), where models pre-trained on shorter sequences face difficulty with out-of-distribution (OOD) token positions in longer sequences. We introduce Resonance RoPE, a novel approach designed to narrow the generalization gap in TSTL scenarios by refining the interpolation of RoPE features for OOD positions, significantly improving the model performance without additional online computational costs. Furthermore, we present PosGen, a new synthetic benchmark specifically designed for fine-grained behavior analysis in TSTL scenarios, aiming to isolate the constantly increasing difficulty of token generation on long contexts from the challenges of recognizing new token positions. Our experiments on synthetic tasks show that after applying Resonance RoPE, Transformers recognize OOD position better and more robustly. Our extensive LLM experiments also show superior performance after applying Resonance RoPE to the current state-of-the-art RoPE scaling method, YaRN, on both upstream language modeling tasks and a variety of downstream long-text applications.

We present a simple on the fly method for faster inference of large language models. Unlike other (self-)speculative decoding techniques, our method does not require fine-tuning or black-box optimization to generate a fixed draft model, relying instead on simple rules to generate varying draft models adapted to the input context. We show empirically that our light-weight algorithm is competitive with the current SOTA for self-speculative decoding, while being a truly plug-and-play method.

Knowledge Distillation (KD) is a prominent neural model compression technique that heavily relies on teacher network predictions to guide the training of a student model. Considering the ever-growing size of pre-trained language models (PLMs), KD is often adopted in many NLP tasks involving PLMs. However, it is evident that in KD, deploying the teacher network during training adds to the memory and computational requirements of training. In the computer vision literature, the necessity of the teacher network is put under scrutiny by showing that KD is a label regularization technique that can be replaced with lighter teacher-free variants such as the label-smoothing technique. However, to the best of our knowledge, this issue is not investigated in NLP. Therefore, this work concerns studying different label regularization techniques and whether we actually need them to improve the fine-tuning of smaller PLM networks on downstream tasks. In this regard, we did a comprehensive set of experiments on different PLMs such as BERT, RoBERTa, and GPT with more than 600 distinct trials and ran each configuration five times. This investigation led to a surprising observation that KD and other label regularization techniques do not play any meaningful role over regular fine-tuning when the student model is pre-trained. We further explore this phenomenon in different settings of NLP and computer vision tasks and demonstrate that pre-training itself acts as a kind of regularization, and additional label regularization is unnecessary.

Transformer-based models have achieved state-of-the-art performance on numerous NLP applications. However, long documents which are prevalent in real-world scenarios cannot be efficiently processed by transformers with the vanilla self-attention module due to their quadratic computation complexity and limited length extrapolation ability. Instead of tackling the computation difficulty for self-attention with sparse or hierarchical structures, in this paper, we investigate the use of State-Space Models (SSMs) for long document classification tasks. We conducted extensive experiments on six long document classification datasets, including binary, multi-class, and multi-label classification, comparing SSMs (with and without pre-training) to self-attention-based models. We also introduce the SSM-pooler model and demonstrate that it achieves comparable performance while being on average 36% more efficient. Additionally our method exhibits higher robustness to the input noise even in the extreme scenario of 40%.

Regularization techniques are crucial to improving the generalization performance and training efficiency of deep neural networks. Many deep learning algorithms rely on weight decay, dropout, batch/layer normalization to converge faster and generalize. Label Smoothing (LS) is another simple, versatile and efficient regularization which can be applied to various supervised classification tasks. Conventional LS, however, regardless of the training instance assumes that each non-target class is equally likely. In this work, we present a general framework for training with label regularization, which includes conventional LS but can also model instance-specific variants. Based on this formulation, we propose an efficient way of learning LAbel regularization by devising a Bi-level Optimization (LABO) problem. We derive a deterministic and interpretable solution of the inner loop as the optimal label smoothing without the need to store the parameters or the output of a trained model. Finally, we conduct extensive experiments and demonstrate our LABO consistently yields improvement over conventional label regularization on various fields, including seven machine translation and three image classification tasks across various neural network architectures while maintaining training efficiency.

Knowledge Distillation (KD) is a commonly used technique for improving the generalization of compact Pre-trained Language Models (PLMs) on downstream tasks. However, such methods impose the additional burden of training a separate teacher model for every new dataset.Alternatively, one may directly work on the improvement of the optimization procedure of the compact model towards better generalization. Recent works observe that the flatness of the local minimum correlates well with better generalization.In this work, we adapt Stochastic Weight Averaging (SWA), a method encouraging convergence to a flatter minimum, to fine-tuning PLMs. We conduct extensive experiments on various NLP tasks (text classification, question answering, and generation) and different model architectures and demonstrate that our adaptation improves the generalization without extra computation cost. Moreover, we observe that this simple optimization technique is able to outperform the state-of-the-art KD methods for compact models.

Knowledge Distillation (KD) is extensively used in Natural Language Processing to compress the pre-training and task-specific fine-tuning phases of large neural language models. A student model is trained to minimize a convex combination of the prediction loss over the labels and another over the teacher output. However, most existing works either fix the interpolating weight between the two losses apriori or vary the weight using heuristics. In this work, we propose a novel sample-wise loss weighting method, RW-KD. A meta-learner, simultaneously trained with the student, adaptively re-weights the two losses for each sample. We demonstrate, on 7 datasets of the GLUE benchmark, that RW-KD outperforms other loss re-weighting methods for KD.

Multi-task learning (MTL) has been studied recently for sequence labeling. Typically, auxiliary tasks are selected specifically in order to improve the performance of a target task. Jointly learning multiple tasks in a way that benefit all of them simultaneously can increase the utility of MTL. In order to do so, we propose a new LSTM cell which contains both shared parameters that can learn from all tasks, and task-specific parameters that can learn task-specific information. We name it a Shared-Cell Long-Short Term Memory SC-LSTM. Experimental results on three sequence labeling benchmarks (named-entity recognition, text chunking, and part-of-speech tagging) demonstrate the effectiveness of our SC-LSTM cell.