Language models (LMs) often struggle to pay enough attention to the input context, and generate texts that are unfaithful or contain hallucinations. To mitigate this issue, we present context-aware decoding (CAD), which follows a contrastive output distribution that amplifies the difference between the output probabilities when a model is used with and without context. Our experiments show that CAD, without additional training, significantly improves the faithfulness of different LM families, including OPT, GPT, LLaMA and FLAN-T5 for summarization tasks (e.g., 14.3% gain for LLaMA in factuality metrics). Furthermore, CAD is particularly effective in overriding a model's prior knowledge when it contradicts the provided context, leading to substantial improvements in tasks where resolving the knowledge conflict is essential.

Federated learning is a distributed learning paradigm in which a set of agents aim to collaboratively train a global statistical model. Due to privacy considerations, rather than sharing local training data, agents are incentivized to communicate either their local models or gradient information to their network. Yet, communication efficiency is a crucial factor to consider, as the network could potentially incur latency, congestion, and failures. A growing body of work, e.g., [Lian et al., 2015, McMahan et al., 2016, Li et al., 2020] has deployed a server in the network to facilitate reaching consensus among the agents, which despite its efficiency, creates a communication bottleneck at the server. On the other hand, another line of work that traces back to decentralized optimization [Nedic and Ozdaglar, 2009, Wei and Ozdaglar, 2012, Wu et al., 2017] has been investigated to design federated learning frameworks without a central server, compromising communication efficiency and convergence rates [Vanhaesebrouck et al., 2017, Liu et al., 2022a,b]. Indeed, the two categories have been enriched by the advances in iterative optimization algorithms and, in particular, stochastic gradient descent (SGD) and its variants. While learning frameworks have significantly benefited from well-crafted optimization algorithms, the converse has also been made possible due to algorithm unrolling [Chen et al., 2021b, Monga et al., 2021].

Agent-based modelling (ABMing) is a powerful and intuitive approach to modelling complex systems; however, the intractability of ABMs' likelihood functions and the non-differentiability of the mathematical operations comprising these models present a challenge to their use in the real world. These difficulties have in turn generated research on approximate Bayesian inference methods for ABMs and on constructing differentiable approximations to arbitrary ABMs, but little work has been directed towards designing approximate Bayesian inference techniques for the specific case of differentiable ABMs. In this work, we aim to address this gap and discuss how generalised variational inference procedures may be employed to provide misspecification-robust Bayesian parameter inferences for differentiable ABMs. We demonstrate with experiments on a differentiable ABM of the COVID-19 pandemic that our approach can result in accurate inferences, and discuss avenues for future work.

LLM-powered chatbots are becoming widely adopted in applications such as healthcare, personal assistants, industry hiring decisions, etc. In many of these cases, chatbots are fed sensitive, personal information in their prompts, as samples for in-context learning, retrieved records from a database, or as part of the conversation. The information provided in the prompt could directly appear in the output, which might have privacy ramifications if there is sensitive information there. As such, in this paper, we aim to understand the input copying and regurgitation capabilities of these models during inference and how they can be directly instructed to limit this copying by complying with regulations such as HIPAA and GDPR, based on their internal knowledge of them. More specifically, we find that when ChatGPT is prompted to summarize cover letters of a 100 candidates, it would retain personally identifiable information (PII) verbatim in 57.4% of cases, and we find this retention to be non-uniform between different subgroups of people, based on attributes such as gender identity. We then probe ChatGPT's perception of privacy-related policies and privatization mechanisms by directly instructing it to provide compliant outputs and observe a significant omission of PII from output.

Semantic segmentation is a critical task in computer vision that aims to identify and classify individual pixels in an image, with numerous applications for example autonomous driving and medical image analysis. However, semantic segmentation can be super challenging particularly due to the need for large amounts of annotated data. Annotating images is a time-consuming and costly process, often requiring expert knowledge and significant effort. In this paper, we propose a novel approach for semantic segmentation by eliminating the need of ground-truth segmentation maps. Instead, our approach requires only the rough information of individual semantic class proportions, shortened as semantic proportions. It greatly simplifies the data annotation process and thus will significantly reduce the annotation time and cost, making it more feasible for large-scale applications. Moreover, it opens up new possibilities for semantic segmentation tasks where obtaining the full ground-truth segmentation maps may not be feasible or practical. Extensive experimental results demonstrate that our approach can achieve comparable and sometimes even better performance against the benchmark method that relies on the ground-truth segmentation maps. Utilising semantic proportions suggested in this work offers a promising direction for future research in the field of semantic segmentation.

Place recognition using SOund Navigation and Ranging (SONAR) images is an important task for simultaneous localization and mapping(SLAM) in underwater environments. This paper proposes a robust and efficient imaging SONAR based place recognition, SONAR context, and loop closure method. Unlike previous methods, our approach encodes geometric information based on the characteristics of raw SONAR measurements without prior knowledge or training. We also design a hierarchical searching procedure for fast retrieval of candidate SONAR frames and apply adaptive shifting and padding to achieve robust matching on rotation and translation changes. In addition, we can derive the initial pose through adaptive shifting and apply it to the iterative closest point (ICP) based loop closure factor. We evaluate the performance of SONAR context in the various underwater sequences such as simulated open water, real water tank, and real underwater environments. The proposed approach shows the robustness and improvements of place recognition on various datasets and evaluation metrics. Supplementary materials are available at https://github.com/sparolab/sonar_context.git.

Most judicial decisions involve the interpretation of legal texts; as such, judicial opinion requires the use of language as a medium to comment on or draw attention to other language. Language used this way is called metalanguage. We develop an annotation schema for categorizing types of legal metalanguage and apply our schema to a set of U.S. Supreme Court opinions, yielding a corpus totaling 59k tokens. We remark on several patterns observed in the kinds of metalanguage used by the justices.

The transformer model is known to be computationally demanding, and prohibitively costly for long sequences, as the self-attention module uses a quadratic time and space complexity with respect to sequence length. Many researchers have focused on designing new forms of self-attention or introducing new parameters to overcome this limitation, however a large portion of them prohibits the model to inherit weights from large pretrained models. In this work, the transformer's inefficiency has been taken care of from another perspective. We propose Fourier Transformer, a simple yet effective approach by progressively removing redundancies in hidden sequence using the ready-made Fast Fourier Transform (FFT) operator to perform Discrete Cosine Transformation (DCT). Fourier Transformer is able to significantly reduce computational costs while retain the ability to inherit from various large pretrained models. Experiments show that our model achieves state-of-the-art performances among all transformer-based models on the long-range modeling benchmark LRA with significant improvement in both speed and space. For generative seq-to-seq tasks including CNN/DailyMail and ELI5, by inheriting the BART weights our model outperforms the standard BART and other efficient models. \footnote{Our code is publicly available at \url{https://github.com/LUMIA-Group/FourierTransformer}}

Large language models (LLMs) encode a large amount of world knowledge. However, as such knowledge is frozen at the time of model training, the models become static and limited by the training data at that time. In order to further improve the capacity of LLMs for knowledge-intensive tasks, we consider augmenting LLMs with the large-scale web using search engine. Unlike previous augmentation sources (e.g., Wikipedia data dump), the web provides broader, more comprehensive and constantly updated information. In this paper, we present a web-augmented LLM UNIWEB, which is trained over 16 knowledge-intensive tasks in a unified text-to-text format. Instead of simply using the retrieved contents from web, our approach has made two major improvements. Firstly, we propose an adaptive search engine assisted learning method that can self-evaluate the confidence level of LLM's predictions, and adaptively determine when to refer to the web for more data, which can avoid useless or noisy augmentation from web. Secondly, we design a pretraining task, i.e., continual knowledge learning, based on salient spans prediction, to reduce the discrepancy between the encoded and retrieved knowledge. Experiments on a wide range of knowledge-intensive tasks show that our model significantly outperforms previous retrieval-augmented methods.

Parameter-efficient tuning aims to mitigate the large memory requirements of adapting pretrained language models for downstream tasks. For example, one popular method, prefix-tuning, prepends trainable tokens to sequences while freezing the rest of the model's parameters. Although such models attain comparable performance with fine-tuning when applied to sequences with short to moderate lengths, we show their inferior performance when modelling long sequences. To bridge this gap, we propose prefix-propagation, a simple but effective approach that conditions prefixes on previous hidden states. We empirically demonstrate that prefix-propagation outperforms prefix-tuning across long-document tasks, while using 50% fewer parameters. To further investigate the proposed architecture, we also show its advantage in calibration, and perform additional study on its relationship with kernel attention. To the best of our knowledge, this work is the first to focus on parameter-efficient learning for long-sequence language tasks.