Large Language Models (LLMs) are increasingly costly to fine-tune due to their size, with embedding layers alone accounting for up to 20% of model parameters. While Parameter-Efficient Fine-Tuning (PEFT) methods exist, they largely overlook the embedding layer. In this paper, we introduce TinyTE, a novel PEFT approach that steers model behavior via minimal translational transformations in the embedding space. TinyTE modifies input embeddings without altering hidden layers, achieving competitive performance while requiring approximately 0.0001% of the parameters needed for full fine-tuning. Experiments across architectures provide a new lens for understanding the relationship between input representations and model behavior—revealing them to be more flexible at their foundation than previously thought.

An ideal length-extrapolatable Transformer language model can handle sequences longer than the training length without any fine-tuning. Such long-context utilization capability relies heavily on a flexible positional embedding design. Upon investigating the flexibility of existing large pre-trained Transformer language models, we find that the T5 family deserves a closer look, as its positional embeddings capture rich and flexible attention patterns. However, T5 suffers from the dispersed attention issue: the longer the input sequence, the flatter the attention distribution. To alleviate the issue, we propose two attention alignment strategies via temperature scaling. Our findings show improvement on the long-context utilization capability of T5 on language modeling, retrieval, multi-document question answering, and code completion tasks without any fine-tuning. This suggests that a flexible positional embedding design and attention alignment can go a long way toward Transformer length extrapolation. The code is released at: https://github.com/chijames/T5-Attention-Alignment

In recent studies, linear recurrent neural networks (LRNNs) have achieved Transformer-level performance in natural language and long-range modeling, while offering rapid parallel training and constant inference cost. With the resurgence of interest in LRNNs, we study whether they can learn the hidden rules in training sequences, such as the grammatical structures of regular language. We theoretically analyze some existing LRNNs and discover their limitations in modeling regular language. Motivated by this analysis, we propose a new LRNN equipped with a block-diagonal and input-dependent transition matrix. Experiments suggest that the proposed model is the only LRNN capable of performing length extrapolation on regular language tasks such as Sum, Even Pair, and Modular Arithmetic. The code is released at https://github.com/tinghanf/RegluarLRNN.

Length extrapolation permits training a transformer language model on short sequences that preserves perplexities when tested on substantially longer sequences.A relative positional embedding design, ALiBi, has had the widest usage to date. We dissect ALiBi via the lens of receptive field analysis empowered by a novel cumulative normalized gradient tool. The concept of receptive field further allows us to modify the vanilla Sinusoidal positional embedding to create Sandwich, the first parameter-free relative positional embedding design that truly length information uses longer than the training sequence. Sandwich shares with KERPLE and T5 the same logarithmic decaying temporal bias pattern with learnable relative positional embeddings; these elucidate future extrapolatable positional embedding design.

The use of positional embeddings in transformer language models is widely accepted. However, recent research has called into question the necessity of such embeddings. We further extend this inquiry by demonstrating that a randomly initialized and frozen transformer language model, devoid of positional embeddings, inherently encodes strong positional information through the shrinkage of self-attention variance. To quantify this variance, we derive the underlying distribution of each step within a transformer layer. Through empirical validation using a fully pretrained model, we show that the variance shrinkage effect still persists after extensive gradient updates. Our findings serve to justify the decision to discard positional embeddings and thus facilitate more efficient pretraining of transformer language models.

The advent and fast development of neural networks have revolutionized the research on dialogue systems and subsequently have triggered various challenges regarding their automatic evaluation. Automatic evaluation of open-domain dialogue systems as an open challenge has been the center of the attention of many researchers. Despite the consistent efforts to improve automatic metrics’ correlations with human evaluation, there have been very few attempts to assess their robustness over multiple domains and dimensions. Also, their focus is mainly on the English language. All of these challenges prompt the development of automatic evaluation metrics that are reliable in various domains, dimensions, and languages. This track in the 11th Dialogue System Technology Challenge (DSTC11) is part of the ongoing effort to promote robust and multilingual automatic evaluation metrics. This article describes the datasets and baselines provided to participants and discusses the submission and result details of the two proposed subtasks.

Unlike recurrent models, conventional wisdom has it that Transformers cannot perfectly model regular languages. Inspired by the notion of working memory, we propose a new Transformer variant named RegularGPT. With its novel combination of Weight-Sharing, Adaptive-Depth, and Sliding-Dilated-Attention, RegularGPT constructs working memory along the depth dimension, thereby enabling efficient and successful modeling of regular languages such as PARITY. We further test RegularGPT on the task of natural language length extrapolation and surprisingly find that it rediscovers the local windowed attention effect deemed necessary in prior work for length extrapolation.

In this work, we examine the problems associated with neural dialog models under the common theme of compositionality. Specifically, we investigate three manifestations of compositionality: (1) Productivity, (2) Substitutivity, and (3) Systematicity. These manifestations shed light on the generalization, syntactic robustness, and semantic capabilities of neural dialog models. We design probing experiments by perturbing the training data to study the above phenomenon. We make informative observations based on automated metrics and hope that this work increases research interest in understanding the capacity of these models.

Dialogue discourse parsing aims to uncover the internal structure of a multi-participant conversation by finding all the discourse links and corresponding relations. Previous work either treats this task as a series of independent multiple-choice problems, in which the link existence and relations are decoded separately, or the encoding is restricted to only local interaction, ignoring the holistic structural information. In contrast, we propose a principled method that improves upon previous work from two perspectives: encoding and decoding. From the encoding side, we perform structured encoding on the adjacency matrix followed by the matrix-tree learning algorithm, where all discourse links and relations in the dialogue are jointly optimized based on latent tree-level distribution. From the decoding side, we perform structured inference using the modified Chiu-Liu-Edmonds algorithm, which explicitly generates the labeled multi-root non-projective spanning tree that best captures the discourse structure. In addition, unlike in previous work, we do not rely on hand-crafted features; this improves the model’s robustness. Experiments show that our method achieves new state-of-the-art, surpassing the previous model by 2.3 on STAC and 1.5 on Molweni (F1 scores).

Dialogue disentanglement aims to group utterances in a long and multi-participant dialogue into threads. This is useful for discourse analysis and downstream applications such as dialogue response selection, where it can be the first step to construct a clean context/response set. Unfortunately, labeling all reply-to links takes quadratic effort w.r.t the number of utterances: an annotator must check all preceding utterances to identify the one to which the current utterance is a reply. In this paper, we are the first to propose a zero-shot dialogue disentanglement solution. Firstly, we train a model on a multi-participant response selection dataset harvested from the web which is not annotated; we then apply the trained model to perform zero-shot dialogue disentanglement. Without any labeled data, our model can achieve a cluster F1 score of 25. We also fine-tune the model using various amounts of labeled data. Experiments show that with only 10% of the data, we achieve nearly the same performance of using the full dataset.

Natural Language Image Editing (NLIE) aims to use natural language instructions to edit images. Since novices are inexperienced with image editing techniques, their instructions are often ambiguous and contain high-level abstractions which require complex editing steps. Motivated by this inexperience aspect, we aim to smooth the learning curve by teaching the novices to edit images using low-level command terminologies. Towards this end, we develop a task-oriented dialogue system to investigate low-level instructions for NLIE. Our system grounds language on the level of edit operations, and suggests options for users to choose from. Though compelled to express in low-level terms, user evaluation shows that 25% of users found our system easy-to-use, resonating with our motivation. Analysis shows that users generally adapt to utilizing the proposed low-level language interface. We also identified object segmentation as the key factor to user satisfaction. Our work demonstrates advantages of low-level, direct language-action mapping approach that can be applied to other problem domains beyond image editing such as audio editing or industrial design.

Users will interact with an individual app on smart devices (e.g., phone, TV, car) to fulfill a specific goal (e.g. find a photographer), but users may also pursue more complex tasks that will span multiple domains and apps (e.g. plan a wedding ceremony). Planning and executing such multi-app tasks are typically managed by users, considering the required global context awareness. To investigate how users arrange domains/apps to fulfill complex tasks in their daily life, we conducted a user study on 14 participants to collect such data from their Android smart phones. This document 1) summarizes the techniques used in the data collection and 2) provides a brief statistical description of the data. This data guilds the future direction for researchers in the fields of conversational agent and personal assistant, etc. This data is available at http://AppDialogue.com.

The Translating Telephone has been a major goal of speech translation for many years. Previous approaches have attempted to work from limited-domain, fully-automatic translation towards broad-coverage, fully-automatic translation. We are approaching the problem from a different direction: starting with a broad-coverage but not fully-automatic system, and working towards full automation. We believe that working in this direction will provide us with better feedback, by observing users and collecting language data under realistic conditions, and thus may allow more rapid progress towards the same ultimate goal. Our initial approach relies on the wide-spread availability of Internet connections and web browsers to provide a user interface. We describe our initial work, which is an extension of the Diplomat wearable speech translator.