It is likely that AI systems driven by pre-trained language models (PLMs) will increasingly be used to assist humans in high-stakes interactions with other agents, such as negotiation or conflict resolution. Consistent with the goals of Cooperative AI \citep{dafoe_open_2020}, we wish to understand and shape the multi-agent behaviors of PLMs in a pro-social manner. An important first step is the evaluation of model behaviour across diverse cooperation problems. Since desired behaviour in an interaction depends upon precise game-theoretic structure, we focus on generating scenarios with particular structures with both crowdworkers and a language model. Our work proceeds as follows. First, we discuss key methodological issues in the generation of scenarios corresponding to particular game-theoretic structures. Second, we employ both crowdworkers and a language model to generate such scenarios. We find that the quality of generations tends to be mediocre in both cases. We additionally get both crowdworkers and a language model to judge whether given scenarios align with their intended game-theoretic structure, finding mixed results depending on the game. Third, we provide a dataset of scenario based on our data generated. We provide both quantitative and qualitative evaluations of UnifiedQA and GPT-3 on this dataset. We find that instruct-tuned models tend to act in a way that could be perceived as cooperative when scaled up, while other models seemed to have flat scaling trends.

A study was conducted to prove AI software could be used to translate and generate illustrations without any human intervention. This was done with the purpose of showing and distributing it to the external customer, Pratham Books. The project aligns with the company's vision by leveraging the generalisation and scalability of Machine Learning algorithms, offering significant cost efficiency increases to a wide range of literary audiences in varied geographical locations. A comparative study methodology was utilised to determine the best performant method out of the 3 devised, Prompt Augmentation using Keywords, CLIP Embedding Mask, and Cross Attention Control with Editorial Prompts. A thorough evaluation process was completed using both quantitative and qualitative measures. Each method had its own strengths and weaknesses, but through the evaluation, method 1 was found to have the best yielding results. Promising future advancements may be made to further increase image quality by incorporating Large Language Models and personalised stylistic models. The presented approach can also be adapted to Video and 3D sculpture generation for novel illustrations in digital webbooks.

Emotional Reaction Intensity(ERI) estimation is an important task in multimodal scenarios, and has fundamental applications in medicine, safe driving and other fields. In this paper, we propose a solution to the ERI challenge of the fifth Affective Behavior Analysis in-the-wild(ABAW), a dual-branch based multi-output regression model. The spatial attention is used to better extract visual features, and the Mel-Frequency Cepstral Coefficients technology extracts acoustic features, and a method named modality dropout is added to fusion multimodal features. Our method achieves excellent results on the official validation set.

Runtime-tunable context-dependent network compression would make mobile deep learning adaptable to often varying resource availability, input "difficulty", or user needs. The existing compression techniques significantly reduce the memory, processing, and energy tax of deep learning, yet, the resulting models tend to be permanently impaired, sacrificing the inference power for reduced resource usage. The existing tunable compression approaches, on the other hand, require expensive re-training, seldom provide mobile-ready implementations, and do not support arbitrary strategies for adapting the compression. In this paper we present Mobiprox, a framework enabling flexible-accuracy on-device deep learning. Mobiprox implements tunable approximations of tensor operations and enables runtime adaptation of individual network layers. A profiler and a tuner included with Mobiprox identify the most promising neural network approximation configurations leading to the desired inference quality with the minimal use of resources. Furthermore, we develop control strategies that depending on contextual factors, such as the input data difficulty, dynamically adjust the approximation level of a model. We implement Mobiprox in Android OS and through experiments in diverse mobile domains, including human activity recognition and spoken keyword detection, demonstrate that it can save up to 15% system-wide energy with a minimal impact on the inference accuracy.

Grammatical inference consists in learning a formal grammar as a finite state machine or as a set of rewrite rules. In this paper, we are concerned with inferring Nondeterministic Finite Automata (NFA) that must accept some words, and reject some other words from a given sample. This problem can naturally be modeled in SAT. The standard model being enormous, some models based on prefixes, suffixes, and hybrids were designed to generate smaller SAT instances. There is a very simple and obvious property that says: if there is an NFA of size k for a given sample, there is also an NFA of size k+1. We first strengthen this property by adding some characteristics to the NFA of size k+1. Hence, we can use this property to tighten the bounds of the size of the minimal NFA for a given sample. We then propose simplified and refined models for NFA of size k+1 that are smaller than the initial models for NFA of size k. We also propose a reduction algorithm to build an NFA of size k from a specific NFA of size k+1. Finally, we validate our proposition with some experimentation that shows the efficiency of our approach.

Understanding player shooting profiles is an essential part of basketball analysis: knowing where certain opposing players like to shoot from can help coaches neutralize offensive gameplans from their opponents; understanding where their players are most comfortable can lead them to developing more effective offensive strategies. An automatic tool that can provide these performance profiles in a timely manner can become invaluable for coaches to maximize both the effectiveness of their game plan as well as the time dedicated to practice and other related activities. Additionally, basketball is dictated by many variables, such as playstyle and game dynamics, that can change the flow of the game and, by extension, player performance profiles. It is crucial that the performance profiles can reflect the diverse playstyles, as well as the fast-changing dynamics of the game. We present a tool that can visualize player performance profiles in a timely manner while taking into account factors such as play-style and game dynamics. Our approach generates interpretable heatmaps that allow us to identify and analyze how non-spatial factors, such as game dynamics or playstyle, affect player performance profiles.

Recent breakthroughs in self supervised training have led to a new class of pretrained vision language models. While there have been investigations of bias in multimodal models, they have mostly focused on gender and racial bias, giving much less attention to other relevant groups, such as minorities with regard to religion, nationality, sexual orientation, or disabilities. This is mainly due to lack of suitable benchmarks for such groups. We seek to address this gap by providing a visual and textual bias benchmark called MMBias, consisting of around 3,800 images and phrases covering 14 population subgroups. We utilize this dataset to assess bias in several prominent self supervised multimodal models, including CLIP, ALBEF, and ViLT. Our results show that these models demonstrate meaningful bias favoring certain groups. Finally, we introduce a debiasing method designed specifically for such large pre-trained models that can be applied as a post-processing step to mitigate bias, while preserving the remaining accuracy of the model.

Coreference resolution models are often evaluated on multiple datasets. Datasets vary, however, in how coreference is realized -- i.e., how the theoretical concept of coreference is operationalized in the dataset -- due to factors such as the choice of corpora and annotation guidelines. We investigate the extent to which errors of current coreference resolution models are associated with existing differences in operationalization across datasets (OntoNotes, PreCo, and Winogrande). Specifically, we distinguish between and break down model performance into categories corresponding to several types of coreference, including coreferring generic mentions, compound modifiers, and copula predicates, among others. This break down helps us investigate how state-of-the-art models might vary in their ability to generalize across different coreference types. In our experiments, for example, models trained on OntoNotes perform poorly on generic mentions and copula predicates in PreCo. Our findings help calibrate expectations of current coreference resolution models; and, future work can explicitly account for those types of coreference that are empirically associated with poor generalization when developing models.

Gradient-based learning in multi-layer neural networks displays a number of striking features. In particular, the decrease rate of empirical risk is non-monotone even after averaging over large batches. Long plateaus in which one observes barely any progress alternate with intervals of rapid decrease. These successive phases of learning often take place on very different time scales. Finally, models learnt in an early phase are typically `simpler' or `easier to learn' although in a way that is difficult to formalize. Although theoretical explanations of these phenomena have been put forward, each of them captures at best certain specific regimes. In this paper, we study the gradient flow dynamics of a wide two-layer neural network in high-dimension, when data are distributed according to a single-index model (i.e., the target function depends on a one-dimensional projection of the covariates). Based on a mixture of new rigorous results, non-rigorous mathematical derivations, and numerical simulations, we propose a scenario for the learning dynamics in this setting. In particular, the proposed evolution exhibits separation of timescales and intermittency. These behaviors arise naturally because the population gradient flow can be recast as a singularly perturbed dynamical system.

This paper tackles the problem of recovering a low-rank signal tensor with possibly correlated components from a random noisy tensor, or so-called spiked tensor model. When the underlying components are orthogonal, they can be recovered efficiently using tensor deflation which consists of successive rank-one approximations, while non-orthogonal components may alter the tensor deflation mechanism, thereby preventing efficient recovery. Relying on recently developed random tensor tools, this paper deals precisely with the non-orthogonal case by deriving an asymptotic analysis of a parameterized deflation procedure performed on an order-three and rank-two spiked tensor. Based on this analysis, an efficient tensor deflation algorithm is proposed by optimizing the parameter introduced in the deflation mechanism, which in turn is proven to be optimal by construction for the studied tensor model. The same ideas could be extended to more general low-rank tensor models, e.g., higher ranks and orders, leading to more efficient tensor methods with a broader impact on machine learning and beyond.