Installing probabilistic world models into artificial agents opens an efficient channel for humans to communicate with and control these agents. In addition to updating agent policies, humans can modify their internal world models in order to influence their decisions. The challenge, however, is that currently existing world models are difficult for humans to adapt because they lack a natural communication interface. Aimed at addressing this shortcoming, we develop Language-Guided World Models (LWMs), which can capture environment dynamics by reading language descriptions. These models enhance agent communication efficiency, allowing humans to simultaneously alter their behavior on multiple tasks with concise language feedback. They also enable agents to self-learn from texts originally written to instruct humans. To facilitate the development of LWMs, we design a challenging benchmark based on the game of MESSENGER (Hanjie et al., 2021), requiring compositional generalization to new language descriptions and environment dynamics. Our experiments reveal that the current state-of-the-art Transformer architecture performs poorly on this benchmark, motivating us to design a more robust architecture. To showcase the practicality of our proposed LWMs, we simulate a scenario where these models augment the interpretability and safety of an agent by enabling it to generate and discuss plans with a human before execution. By effectively incorporating language feedback on the plan, the models boost the agent performance in the real environment by up to three times without collecting any interactive experiences in this environment.

Knowledge Organization (KO) and Knowledge Representation (KR) have been the two mainstream methodologies of knowledge modelling in the Information Science community and the Artificial Intelligence community, respectively. The facet-analytical tradition of KO has developed an exhaustive set of guiding canons for ensuring quality in organising and managing knowledge but has remained limited in terms of technology-driven activities to expand its scope and services beyond the bibliographic universe of knowledge. KR, on the other hand, boasts of a robust ecosystem of technologies and technology-driven service design which can be tailored to model any entity or scale to any service in the entire universe of knowledge. This paper elucidates both the facet-analytical KO and KR methodologies in detail and provides a functional mapping between them. Out of the mapping, the paper proposes an integrated KR-enriched KO methodology with all the standard components of a KO methodology plus the advanced technologies provided by the KR approach. The practical benefits of the methodological integration has been exemplified through the flagship application of the Digital University at the University of Trento, Italy.

Robust evidence suggests that humans explore their environment using a combination of topological landmarks and coarse-grained path integration. This approach relies on identifiable environmental features (topological landmarks) in tandem with estimations of distance and direction (coarse-grained path integration) to construct cognitive maps of the surroundings. This cognitive map is believed to exhibit a hierarchical structure, allowing efficient planning when solving complex navigation tasks. Inspired by human behaviour, this paper presents a scalable hierarchical active inference model for autonomous navigation, exploration, and goal-oriented behaviour. The model uses visual observation and motion perception to combine curiosity-driven exploration with goal-oriented behaviour. Motion is planned using different levels of reasoning, i.e., from context to place to motion. This allows for efficient navigation in new spaces and rapid progress toward a target. By incorporating these human navigational strategies and their hierarchical representation of the environment, this model proposes a new solution for autonomous navigation and exploration. The approach is validated through simulations in a mini-grid environment.

In decision-making problems with limited training data, policy functions approximated using deep neural networks often exhibit suboptimal performance. An alternative approach involves learning a world model from the limited data and determining actions through online search. However, the performance is adversely affected by compounding errors arising from inaccuracies in the learnt world model. While methods like TreeQN have attempted to address these inaccuracies by incorporating algorithmic structural biases into their architectures, the biases they introduce are often weak and insufficient for complex decision-making tasks. In this work, we introduce Differentiable Tree Search (DTS), a novel neural network architecture that significantly strengthens the inductive bias by embedding the algorithmic structure of a best-first online search algorithm. DTS employs a learnt world model to conduct a fully differentiable online search in latent state space. The world model is jointly optimised with the search algorithm, enabling the learning of a robust world model and mitigating the effect of model inaccuracies. We address potential Q-function discontinuities arising from naive incorporation of best-first search by adopting a stochastic tree expansion policy, formulating search tree expansion as a decision-making task, and introducing an effective variance reduction technique for the gradient computation. We evaluate DTS in an offline-RL setting with a limited training data scenario on Procgen games and grid navigation task, and demonstrate that DTS outperforms popular model-free and model-based baselines.

Planning and reasoning about actions and processes, in addition to reasoning about propositions, are important issues in recent logical and computer science studies. The widespread use of actions in everyday life such as IoT, semantic web services, etc., and the limitations and issues in the action formalisms are two factors that lead us to study how actions are represented. Since 2007, there have been some ideas to integrate Description Logic (DL) and action formalisms for representing both static and dynamic knowledge. Meanwhile, time is an important factor in dynamic situations, and actions change states over time. In this study, on the one hand, we examined related logical structures such as extensions of description logics (DLs), temporal formalisms, and action formalisms. On the other hand, we analyzed possible tools for designing and developing the Knowledge and Action Base (KAB). For representation and reasoning about actions, we embedded actions into DLs (such as Dynamic-ALC and its extensions). We propose a terminable algorithm for action projection, planning, checking the satisfiability, consistency, realizability, and executability, and also querying from KAB. Actions in this framework were modeled with SPIN and added to state space. This framework has also been implemented as a plugin for the Prot\'eg\'e ontology editor. During the last two decades, various algorithms have been presented, but due to the high computational complexity, we face many problems in implementing dynamic ontologies. In addition, an algorithm to detect the inconsistency of actions' effects was not explicitly stated. In the proposed strategy, the interactions of actions with other parts of modeled knowledge, and a method to check consistency between the effects of actions are presented. With this framework, the ramification problem can be well handled in future works.

Chain of Thought (CoT) is significant in improving the reasoning abilities of large language models (LLMs). However, the correlation between the effectiveness of CoT and the length of reasoning steps in prompts remains largely unknown. To shed light on this, we have conducted several empirical experiments to explore the relations. Specifically, we design experiments that expand and compress the rationale reasoning steps within CoT demonstrations, while keeping all other factors constant. We have the following key findings. First, the results indicate that lengthening the reasoning steps in prompts, even without adding new information into the prompt, considerably enhances LLMs' reasoning abilities across multiple datasets. Alternatively, shortening the reasoning steps, even while preserving the key information, significantly diminishes the reasoning abilities of models. This finding highlights the importance of the number of steps in CoT prompts and provides practical guidance to make better use of LLMs' potential in complex problem-solving scenarios. Second, we also investigated the relationship between the performance of CoT and the rationales used in demonstrations. Surprisingly, the result shows that even incorrect rationales can yield favorable outcomes if they maintain the requisite length of inference. Third, we observed that the advantages of increasing reasoning steps are task-dependent: simpler tasks require fewer steps, whereas complex tasks gain significantly from longer inference sequences.

Large Language Models (LLMs) and Large Multimodal Models (LMMs) exhibit impressive problem-solving skills in many tasks and domains, but their ability in mathematical reasoning in visual contexts has not been systematically studied. To bridge this gap, we present MathVista, a benchmark designed to combine challenges from diverse mathematical and visual tasks. It consists of 6,141 examples, derived from 28 existing multimodal datasets involving mathematics and 3 newly created datasets (i.e., IQTest, FunctionQA, and PaperQA). Completing these tasks requires fine-grained, deep visual understanding and compositional reasoning, which all state-of-the-art foundation models find challenging. With MathVista, we have conducted a comprehensive, quantitative evaluation of 12 prominent foundation models. The best-performing GPT-4V model achieves an overall accuracy of 49.9%, substantially outperforming Bard, the second-best performer, by 15.1%. Our in-depth analysis reveals that the superiority of GPT-4V is mainly attributed to its enhanced visual perception and mathematical reasoning. However, GPT-4V still falls short of human performance by 10.4%, as it often struggles to understand complex figures and perform rigorous reasoning. This significant gap underscores the critical role that MathVista will play in the development of general-purpose AI agents capable of tackling mathematically intensive and visually rich real-world tasks. We further explore the new ability of self-verification, the application of self-consistency, and the interactive chatbot capabilities of GPT-4V, highlighting its promising potential for future research. The project is available at https://mathvista.github.io/.

The generalization of decision-making agents encompasses two fundamental elements: learning from past experiences and reasoning in novel contexts. However, the predominant emphasis in most interactive environments is on learning, often at the expense of complexity in reasoning. In this paper, we introduce CivRealm, an environment inspired by the Civilization game. Civilization's profound alignment with human history and society necessitates sophisticated learning, while its ever-changing situations demand strong reasoning to generalize. Particularly, CivRealm sets up an imperfect-information general-sum game with a changing number of players; it presents a plethora of complex features, challenging the agent to deal with open-ended stochastic environments that require diplomacy and negotiation skills. Within CivRealm, we provide interfaces for two typical agent types: tensor-based agents that focus on learning, and language-based agents that emphasize reasoning. To catalyze further research, we present initial results for both paradigms. The canonical RL-based agents exhibit reasonable performance in mini-games, whereas both RL- and LLM-based agents struggle to make substantial progress in the full game. Overall, CivRealm stands as a unique learning and reasoning challenge for decision-making agents. The code is available at https://github.com/bigai-ai/civrealm.

The branch-and-bound algorithm based on decision diagrams introduced by Bergman et al. in 2016 is a framework for solving discrete optimization problems with a dynamic programming formulation. It works by compiling a series of bounded-width decision diagrams that can provide lower and upper bounds for any given subproblem. Eventually, every part of the search space will be either explored or pruned by the algorithm, thus proving optimality. This paper presents new ingredients to speed up the search by exploiting the structure of dynamic programming models. The key idea is to prevent the repeated expansion of nodes corresponding to the same dynamic programming states by querying expansion thresholds cached throughout the search. These thresholds are based on dominance relations between partial solutions previously found and on the pruning inequalities of the filtering techniques introduced by Gillard et al. in 2021. Computational experiments show that the pruning brought by this caching mechanism allows significantly reducing the number of nodes expanded by the algorithm. This results in more benchmark instances of difficult optimization problems being solved in less time while using narrower decision diagrams.

We show that it is possible to understand and identify a decision maker's subjective causal judgements by observing her preferences over interventions. Following Pearl [2000], we represent causality using causal models (also called structural equations models), where the world is described by a collection of variables, related by equations. We show that if a preference relation over interventions satisfies certain axioms (related to standard axioms regarding counterfactuals), then we can define (i) a causal model, (ii) a probability capturing the decision-maker's uncertainty regarding the external factors in the world and (iii) a utility on outcomes such that each intervention is associated with an expected utility and such that intervention $A$ is preferred to $B$ iff the expected utility of $A$ is greater than that of $B$. In addition, we characterize when the causal model is unique. Thus, our results allow a modeler to test the hypothesis that a decision maker's preferences are consistent with some causal model and to identify causal judgements from observed behavior.