Training Large Language Models (LLMs) to follow user instructions has been shown to supply the LLM with ample capacity to converse fluently while being aligned with humans. Yet, it is not completely clear how an LLM can lead a plan-grounded conversation in mixed-initiative settings where instructions flow in both directions of the conversation, i.e. both the LLM and the user provide instructions to one another. In this paper, we tackle a dual goal mixed-initiative conversational setting where the LLM not only grounds the conversation on an arbitrary plan but also seeks to satisfy both a procedural plan and user instructions. The LLM is then responsible for guiding the user through the plan and, at the same time, adapting to new circumstances, answering questions, and activating safety guardrails when needed. We propose a novel LLM that grounds the dialogue on a procedural plan, can take the dialogue initiative, and enforces guardrails on the system's behavior, while also improving the LLM's responses to unexpected user behavior. Experiments in controlled settings and with real users show that the best-performing model, which we call PlanLLM, achieves a 2.1x improvement over a strong baseline. Moreover, experiments also show good generalization to unseen domains.

LLM agents are typically prompted to produce actions by generating JSON or text in a pre-defined format, which is usually limited by constrained action space (e.g., the scope of pre-defined tools) and restricted flexibility (e.g., inability to compose multiple tools). This work proposes to use executable Python code to consolidate LLM agents' actions into a unified action space (CodeAct). Integrated with a Python interpreter, CodeAct can execute code actions and dynamically revise prior actions or emit new actions upon new observations through multi-turn interactions. Our extensive analysis of 17 LLMs on API-Bank and a newly curated benchmark shows that CodeAct outperforms widely used alternatives (up to 20% higher success rate). The encouraging performance of CodeAct motivates us to build an open-source LLM agent that interacts with environments by executing interpretable code and collaborates with users using natural language. To this end, we collect an instruction-tuning dataset CodeActInstruct that consists of 7k multi-turn interactions using CodeAct. We show that it can be used with existing data to improve models in agent-oriented tasks without compromising their general capability. CodeActAgent, finetuned from Llama2 and Mistral, is integrated with Python interpreter and uniquely tailored to perform sophisticated tasks (e.g., model training) using existing libraries and autonomously self-debug.

Text-to-image generative AI models such as Stable Diffusion are used daily by millions worldwide. However, many have raised concerns regarding how these models amplify racial and gender stereotypes. To study this phenomenon, we develop a classifier to predict the race, gender, and age group of any given face image, and show that it achieves state-of-the-art performance. Using this classifier, we quantify biases in Stable Diffusion across six races, two genders, five age groups, 32 professions, and eight attributes. We then propose novel debiasing solutions that outperform state-of-the-art alternatives. Additionally, we examine the degree to which Stable Diffusion depicts individuals of the same race as being similar to one another. This analysis reveals a high degree of stereotyping, e.g., depicting most middle eastern males as being dark-skinned, bearded, and wearing a traditional headdress. We address these limitations by proposing yet another novel solution that increases facial diversity across genders and racial groups. Our solutions are open-sourced and made publicly available.

Knowledge Graphs popularity has been rapidly growing in last years. All that knowledge is available for people to query it through the many online databases on the internet. Though, it would be a great achievement if non-programmer users could access whatever information they want to know. There has been a lot of effort oriented to solve this task using natural language processing tools and creativity encouragement by way of many challenges. Our approach focuses on assuming a correct entity linking on the natural language questions and training a GPT model to create SPARQL queries from them. We managed to isolate which property of the task can be the most difficult to solve at few or zero-shot and we proposed pre-training on all entities (under CWA) to improve the performance. We obtained a 62.703% accuracy of exact SPARQL matches on testing at 3-shots, a F1 of 0.809 on the entity linking challenge and a F1 of 0.009 on the question answering challenge.

Predictive multiplicity refers to the phenomenon in which classification tasks may admit multiple competing models that achieve almost-equally-optimal performance, yet generate conflicting outputs for individual samples. This presents significant concerns, as it can potentially result in systemic exclusion, inexplicable discrimination, and unfairness in practical applications. Measuring and mitigating predictive multiplicity, however, is computationally challenging due to the need to explore all such almost-equally-optimal models, known as the Rashomon set, in potentially huge hypothesis spaces. To address this challenge, we propose a novel framework that utilizes dropout techniques for exploring models in the Rashomon set. We provide rigorous theoretical derivations to connect the dropout parameters to properties of the Rashomon set, and empirically evaluate our framework through extensive experimentation. Numerical results show that our technique consistently outperforms baselines in terms of the effectiveness of predictive multiplicity metric estimation, with runtime speedup up to $20\times \sim 5000\times$. With efficient Rashomon set exploration and metric estimation, mitigation of predictive multiplicity is then achieved through dropout ensemble and model selection.

One well motivated explanation method for classifiers leverages counterfactuals which are hypothetical events identical to real observations in all aspects except for one categorical feature. Constructing such counterfactual poses specific challenges for texts, however, as some attribute values may not necessarily align with plausible real-world events. In this paper we propose a simple method for generating counterfactuals by intervening in the space of text representations which bypasses this limitation. We argue that our interventions are minimally disruptive and that they are theoretically sound as they align with counterfactuals as defined in Pearl's causal inference framework. To validate our method, we first conduct experiments on a synthetic dataset of counterfactuals, allowing for a direct comparison between classifier predictions based on ground truth counterfactuals (obtained through explicit text interventions) and our counterfactuals, derived through interventions in the representation space. Second, we study a real world scenario where our counterfactuals can be leveraged both for explaining a classifier and for bias mitigation.

This study explores the usefulness of machine learning classifiers for modeling freight mode choice. We investigate eight commonly used machine learning classifiers, namely Naive Bayes, Support Vector Machine, Artificial Neural Network, K-Nearest Neighbors, Classification and Regression Tree, Random Forest, Boosting and Bagging, along with the classical Multinomial Logit model. US 2012 Commodity Flow Survey data are used as the primary data source; we augment it with spatial attributes from secondary data sources. The performance of the classifiers is compared based on prediction accuracy results. The current research also examines the role of sample size and training-testing data split ratios on the predictive ability of the various approaches. In addition, the importance of variables is estimated to determine how the variables influence freight mode choice. The results show that the tree-based ensemble classifiers perform the best. Specifically, Random Forest produces the most accurate predictions, closely followed by Boosting and Bagging. With regard to variable importance, shipment characteristics, such as shipment distance, industry classification of the shipper and shipment size, are the most significant factors for freight mode choice decisions.

We research into the clinical, biochemical and neuroimaging factors associated with the outcome of stroke patients to generate a predictive model using machine learning techniques for prediction of mortality and morbidity 3 months after admission. The dataset consisted of patients with ischemic stroke (IS) and non-traumatic intracerebral hemorrhage (ICH) admitted to Stroke Unit of a European Tertiary Hospital prospectively registered. We identified the main variables for machine learning Random Forest (RF), generating a predictive model that can estimate patient mortality/morbidity. In conclusion, machine learning algorithms RF can be effectively used in stroke patients for long-term outcome prediction of mortality and morbidity.

The identification of political actors who put forward claims in public debate is a crucial step in the construction of discourse networks, which are helpful to analyze societal debates. Actor identification is, however, rather challenging: Often, the locally mentioned speaker of a claim is only a pronoun ("He proposed that [claim]"), so recovering the canonical actor name requires discourse understanding. We compare a traditional pipeline of dedicated NLP components (similar to those applied to the related task of coreference) with a LLM, which appears a good match for this generation task. Evaluating on a corpus of German actors in newspaper reports, we find surprisingly that the LLM performs worse. Further analysis reveals that the LLM is very good at identifying the right reference, but struggles to generate the correct canonical form. This points to an underlying issue in LLMs with controlling generated output. Indeed, a hybrid model combining the LLM with a classifier to normalize its output substantially outperforms both initial models.

We introduce Instruction Document Visual Question Answering (iDocVQA) dataset and Large Language Document (LLaDoc) model, for training Language-Vision (LV) models for document analysis and predictions on document images, respectively. Usually, deep neural networks for the DocVQA task are trained on datasets lacking instructions. We show that using instruction-following datasets improves performance. We compare performance across document-related datasets using the recent state-of-the-art (SotA) Large Language and Vision Assistant (LLaVA)1.5 as the base model. We also evaluate the performance of the derived models for object hallucination using the Polling-based Object Probing Evaluation (POPE) dataset. The results show that instruction-tuning performance ranges from 11X to 32X of zero-shot performance and from 0.1% to 4.2% over non-instruction (traditional task) finetuning. Despite the gains, these still fall short of human performance (94.36%), implying there's much room for improvement.