Bridging the gap between diffusion models and human preferences is crucial for their integration into practical generative workflows. While optimizing downstream reward models has emerged as a promising alignment strategy, concerns arise regarding the risk of excessive optimization with learned reward models, which potentially compromises ground-truth performance. In this work, we confront the reward overoptimization problem in diffusion model alignment through the lenses of both inductive and primacy biases. We first identify the divergence of current methods from the temporal inductive bias inherent in the multi-step denoising process of diffusion models as a potential source of overoptimization. Then, we surprisingly discover that dormant neurons in our critic model act as a regularization against overoptimization, while active neurons reflect primacy bias in this setting. Motivated by these observations, we propose Temporal Diffusion Policy Optimization with critic active neuron Reset (TDPO-R), a policy gradient algorithm that exploits the temporal inductive bias of intermediate timesteps, along with a novel reset strategy that targets active neurons to counteract the primacy bias. Empirical results demonstrate the superior efficacy of our algorithms in mitigating reward overoptimization.

This paper presents three established theories of human decision-making and describes how they can be integrated to provide a model of purposive human action. Taking seriously the idea of language as action the model is then applied to the conversational user interfaces. Theory based AI research has had a hard time recently and the aim here is to revitalise interest in understanding what LLMs are actually doing other than running poorly understood machine learning routines over all the data the relevant Big Tech company can hoover up. When a raspberry pi computer for under 50USD is up to 400 times faster than the first commercial Cray super computer~\cite{crayVpi}, Big Tech can get really close to having an infinite number of monkeys typing at random and producing text, some of which will make sense. By understanding where ChatGPT's apparent intelligence comes from, perhaps we can perform the magic with fewer resources and at the same time gain some understanding about our relationship with our world.

Multi-step forecasting (MSF) in time-series, the ability to make predictions multiple time steps into the future, is fundamental to almost all temporal domains. To make such forecasts, one must assume the recursive complexity of the temporal dynamics. Such assumptions are referred to as the forecasting strategy used to train a predictive model. Previous work shows that it is not clear which forecasting strategy is optimal a priori to evaluating on unseen data. Furthermore, current approaches to MSF use a single (fixed) forecasting strategy. In this paper, we characterise the instance-level variance of optimal forecasting strategies and propose Dynamic Strategies (DyStrat) for MSF. We experiment using 10 datasets from different scales, domains, and lengths of multi-step horizons. When using a random-forest-based classifier, DyStrat outperforms the best fixed strategy, which is not knowable a priori, 94% of the time, with an average reduction in mean-squared error of 11%. Our approach typically triples the top-1 accuracy compared to current approaches. Notably, we show DyStrat generalises well for any MSF task.

One-shot imitation is to learn a new task from a single demonstration, yet it is a challenging problem to adopt it for complex tasks with the high domain diversity inherent in a non-stationary environment. To tackle the problem, we explore the compositionality of complex tasks, and present a novel skill-based imitation learning framework enabling one-shot imitation and zero-shot adaptation; from a single demonstration for a complex unseen task, a semantic skill sequence is inferred and then each skill in the sequence is converted into an action sequence optimized for environmental hidden dynamics that can vary over time. Specifically, we leverage a vision-language model to learn a semantic skill set from offline video datasets, where each skill is represented on the vision-language embedding space, and adapt meta-learning with dynamics inference to enable zero-shot skill adaptation. We evaluate our framework with various one-shot imitation scenarios for extended multi-stage Meta-world tasks, showing its superiority in learning complex tasks, generalizing to dynamics changes, and extending to different demonstration conditions and modalities, compared to other baselines.

Large Multimodal Models (LMMs) excel in natural language and visual understanding but are challenged by exacting tasks such as Knowledge-based Visual Question Answering (KB-VQA) which involve the retrieval of relevant information from document collections to use in shaping answers to questions. We present an extensive training and evaluation framework, M2KR, for KB-VQA. M2KR contains a collection of vision and language tasks which we have incorporated into a single suite of benchmark tasks for training and evaluating general-purpose multi-modal retrievers. We use M2KR to develop PreFLMR, a pre-trained version of the recently developed Fine-grained Late-interaction Multi-modal Retriever (FLMR) approach to KB-VQA, and we report new state-of-the-art results across a range of tasks. We also present investigations into the scaling behaviors of PreFLMR intended to be useful in future developments in general-purpose multi-modal retrievers.

The knowledge concept recommendation in Massive Open Online Courses (MOOCs) is a significant issue that has garnered widespread attention. Existing methods primarily rely on the explicit relations between users and knowledge concepts on the MOOC platforms for recommendation. However, there are numerous implicit relations (e.g., shared interests or same knowledge levels between users) generated within the users' learning activities on the MOOC platforms. Existing methods fail to consider these implicit relations, and these relations themselves are difficult to learn and represent, causing poor performance in knowledge concept recommendation and an inability to meet users' personalized needs. To address this issue, we propose a novel framework based on contrastive learning, which can represent and balance the explicit and implicit relations for knowledge concept recommendation in MOOCs (CL-KCRec). Specifically, we first construct a MOOCs heterogeneous information network (HIN) by modeling the data from the MOOC platforms. Then, we utilize a relation-updated graph convolutional network and stacked multi-channel graph neural network to represent the explicit and implicit relations in the HIN, respectively. Considering that the quantity of explicit relations is relatively fewer compared to implicit relations in MOOCs, we propose a contrastive learning with prototypical graph to enhance the representations of both relations to capture their fruitful inherent relational knowledge, which can guide the propagation of students' preferences within the HIN. Based on these enhanced representations, to ensure the balanced contribution of both towards the final recommendation, we propose a dual-head attention mechanism for balanced fusion. Experimental results demonstrate that CL-KCRec outperforms several state-of-the-art baselines on real-world datasets in terms of HR, NDCG and MRR.

Language Models as a Service (LMaaS) offers convenient access for developers and researchers to perform inference using pre-trained language models. Nonetheless, the input data and the inference results containing private information are exposed as plaintext during the service call, leading to privacy issues. Recent studies have started tackling the privacy issue by transforming input data into privacy-preserving representation from the user-end with the techniques such as noise addition and content perturbation, while the exploration of inference result protection, namely decision privacy, is still a blank page. In order to maintain the black-box manner of LMaaS, conducting data privacy protection, especially for the decision, is a challenging task because the process has to be seamless to the models and accompanied by limited communication and computation overhead. We thus propose Instance-Obfuscated Inference (IOI) method, which focuses on addressing the decision privacy issue of natural language understanding tasks in their complete life-cycle. Besides, we conduct comprehensive experiments to evaluate the performance as well as the privacy-protection strength of the proposed method on various benchmarking tasks.

Uncertainty estimation for unlabeled data is crucial to active learning. With a deep neural network employed as the backbone model, the data selection process is highly challenging due to the potential over-confidence of the model inference. Existing methods resort to special learning fashions (e.g. adversarial) or auxiliary models to address this challenge. This tends to result in complex and inefficient pipelines, which would render the methods impractical. In this work, we propose a novel algorithm that leverages noise stability to estimate data uncertainty. The key idea is to measure the output derivation from the original observation when the model parameters are randomly perturbed by noise. We provide theoretical analyses by leveraging the small Gaussian noise theory and demonstrate that our method favors a subset with large and diverse gradients. Our method is generally applicable in various tasks, including computer vision, natural language processing, and structural data analysis. It achieves competitive performance compared against state-of-the-art active learning baselines.

Research into large-scale crop monitoring has flourished due to increased accessibility to satellite imagery. This review delves into previously unexplored and under-explored areas in sugarcane health monitoring and disease/pest detection using satellite-based spectroscopy and Machine Learning (ML). It discusses key considerations in system development, including relevant satellites, vegetation indices, ML methods, factors influencing sugarcane reflectance, optimal growth conditions, common diseases, and traditional detection methods. Many studies highlight how factors like crop age, soil type, viewing angle, water content, recent weather patterns, and sugarcane variety can impact spectral reflectance, affecting the accuracy of health assessments via spectroscopy. However, these variables have not been fully considered in the literature. In addition, the current literature lacks comprehensive comparisons between ML techniques and vegetation indices. We address these gaps in this review. We discuss that, while current findings suggest the potential for an ML-driven satellite spectroscopy system for monitoring sugarcane health, further research is essential. This paper offers a comprehensive analysis of previous research to aid in unlocking this potential and advancing the development of an effective sugarcane health monitoring system using satellite technology.

TurtleRabbit is a new RoboCup SSL team from Western Sydney University. This team description paper presents our approach in navigating some of the challenges in developing a new SSL team from scratch. SSL is dominated by teams with extensive experience and customised equipment that has been developed over many years. Here, we outline our approach in overcoming some of the complexities associated with replicating advanced open-sourced designs and managing the high costs of custom components. Opting for simplicity and cost-effectiveness, our strategy primarily employs off-the-shelf electronics components and ``hobby'' brushless direct current (BLDC) motors, complemented by 3D printing and CNC milling. This approach helped us to streamline the development process and, with our open-sourced hardware design, hopefully will also lower the bar for other teams to enter RoboCup SSL in the future. The paper details the specific hardware choices, their approximate costs, the integration of electronics and mechanics, and the initial steps taken in software development, for our entry into SSL that aims to be simple yet competitive.