Multi-source multi-hop question answering (QA) represents a challenging task in natural language processing due to the need for dynamic integration of heterogeneous knowledge sources and multi-step reasoning. Existing methods often suffer from cascading errors, insufficient handling of knowledge conflicts, and computational inefficiency. In this paper, we propose Adaptive Multi-source Knowledge-Oriented Reasoning (AMKOR), a generative framework that leverages large language models (LLMs) to dynamically fuse parametric and retrieved knowledge while exploring reasoning trajectories using probabilistic beam reasoning. AMKOR is further enhanced by a multi-granular learning strategy, optimizing both local reasoning steps and global answer accuracy. Experiments conducted on four widely-used multi-hop QA datasets, including HotpotQA and MuSiQue, demonstrate that AMKOR achieves state-of-the-art performance, significantly outperforming baseline methods on both reasoning accuracy and robustness. Additional analyses confirm its scalability, adaptability to noisy knowledge, and superior ability to handle complex multi-hop tasks. This work establishes a new benchmark for multi-source multi-hop QA by effectively combining reasoning quality and efficiency.

Completing Long-Horizon (LH) tasks in open-ended worlds is an important yet difficult problem for embodied agents. Existing approaches suffer from two key challenges: (1) they heavily rely on experiences obtained from human-created data or curricula, lacking the ability to continuously update multimodal experiences, and (2) they may encounter catastrophic forgetting issues when faced with new tasks, lacking the ability to continuously update world knowledge. To solve these challenges, this paper presents EvoAgent, an autonomous-evolving agent with a continual World Model (WM), which can autonomously complete various LH tasks across environments through self-planning, self-control, and self-reflection, without human intervention. Our proposed EvoAgent contains three modules, i.e., i) the memory-driven planner which uses an LLM along with the WM and interaction memory, to convert LH tasks into executable sub-tasks; ii) the WM-guided action controller which leverages WM to generate low-level actions and incorporates a self-verification mechanism to update multimodal experiences; iii) the experience-inspired reflector which implements a two-stage curriculum learning algorithm to select experiences for task-adaptive WM updates. Moreover, we develop a continual World Model for EvoAgent, which can continuously update the multimodal experience pool and world knowledge through closed-loop dynamics. We conducted extensive experiments on Minecraft, compared with existing methods, EvoAgent can achieve an average success rate improvement of 105% and reduce ineffective actions by more than 6x.

The paper introduces lifted and/or schematics, which are an abstract representation of the PTP search space. Analyzing the schematic lets us quantify the size of the search space. This is of independent theoretical interest (although the theory is not quite there yet to provide deep insights), and is practically useful in the context of Rao-Blackwellized importance sampling. I believe this type of work is important; we still lack understanding of lifted inference algorithms. The proposed schematics are not really simple enough to provide a whole lot of insight, but it's a move in the right direction.

There is a general concern that present developments in artificial intelligence (AI) research will lead to sentient AI systems, and these may pose an existential threat to humanity. But why cannot sentient AI systems benefit humanity instead? This paper endeavours to put this question in a tractable manner. I ask whether a putative AI system will develop an altruistic or a malicious disposition towards our society, or what would be the nature of its agency? Given that AI systems are being developed into formidable problem solvers, we can reasonably expect these systems to preferentially take on conscious aspects of human problem solving. I identify the relevant phenomenal aspects of agency in human problem solving. The functional aspects of conscious agency can be monitored using tools provided by functionalist theories of consciousness. A recent expert report (Butlin et al. 2023) has identified functionalist indicators of agency based on these theories. I show how to use the Integrated Information Theory (IIT) of consciousness, to monitor the phenomenal nature of this agency. If we are able to monitor the agency of AI systems as they develop, then we can dissuade them from becoming a menace to society while encouraging them to be an aid.

The paper contains some interesting ideas of partitioning the search space in Bayesian optimization. Experimental studies are good, although the paper could benefit from another round of editing for the camera ready. Reviewers were not overly excited about the paper, but also did not identify any fundamental flaws. As such, it is recommended for acceptance as a poster. We strongly encourage the authors to take the feedback from the reviews into account when preparing the camera-ready version.

The paper received mixed ratings: two reviewers recommend acceptance, and two reviewers consider the paper is marginally below the threshold. All reviewers agree that the paper provides useful insights, e.g., the observation that resolution and depth are more important than width for tiny networks. The main concerns raised by the reviewers were (i) novelty is not highly significant/the method is too heuristic (ii) issues with experiments and lack of analysis on other tasks, such as object detection. The rebuttal helped clarify several other questions raised by the reviewers, and included new experiments on COCO object detection using Faster-RCNN. All reviewers actively participated in the discussion phase.

We introduce a novel molecular representation through Algebraic Data Types (ADTs) - composite data structures formed through the combination of simpler types that obey algebraic laws. By explicitly considering how the datatype of a representation constrains the operations which may be performed, we ensure meaningful inference can be performed over generative models (programs with sample} and score operations). This stands in contrast to string-based representations where string-type operations may only indirectly correspond to chemical and physical molecular properties, and at worst produce nonsensical output. The ADT presented implements the Dietz representation for molecular constitution via multigraphs and bonding systems, and uses atomic coordinate data to represent 3D information and stereochemical features. This creates a general digital molecular representation which surpasses the limitations of the string-based representations and the 2D-graph based models on which they are based. In addition, we present novel support for quantum information through representation of shells, subshells, and orbitals, greatly expanding the representational scope beyond current approaches, for instance in Molecular Orbital theory. The framework's capabilities are demonstrated through key applications: Bayesian probabilistic programming is demonstrated through integration with LazyPPL, a lazy probabilistic programming library; molecules are made instances of a group under rotation, necessary for geometric learning techniques which exploit the invariance of molecular properties under different representations; and the framework's flexibility is demonstrated through an extension to model chemical reactions. After critiquing previous representations, we provide an open-source solution in Haskell - a type-safe, purely functional programming language.

The twenty-first century is a data-driven era where human activities and behavior, physical phenomena, scientific discoveries, technology advancements, and almost everything that happens in the world resulting in massive generation, collection, and utilization of data. Connectivity in data is a crucial property. A straightforward example is the World Wide Web, where every webpage is connected to other web pages through hyperlinks, providing a form of directed connectivity. Combinatorial data refers to combinations of data items based on certain connectivity rules. Other forms of combinatorial data include social networks, meshes, community clusters, set systems, and molecules. This Ph.D. dissertation focuses on learning and computing with combinatorial data. We study and examine topological and connectivity features within and across connected data to improve the performance of learning and achieve high algorithmic efficiency.

We introduce in-context denoising, a task that refines the connection between attention-based architectures and dense associative memory (DAM) networks, also known as modern Hopfield networks. Using a Bayesian framework, we show theoretically and empirically that certain restricted denoising problems can be solved optimally even by a single-layer transformer. We demonstrate that a trained attention layer processes each denoising prompt by performing a single gradient descent update on a context-aware DAM energy landscape, where context tokens serve as associative memories and the query token acts as an initial state. This one-step update yields better solutions than exact retrieval of either a context token or a spurious local minimum, providing a concrete example of DAM networks extending beyond the standard retrieval paradigm. Overall, this work solidifies the link between associative memory and attention mechanisms first identified by Ramsauer et al., and demonstrates the relevance of associative memory models in the study of in-context learning.

Training large language models (LLMs) with high-quality Chain-of-Thought (CoT) annotations has become a widely adopted strategy due to its significant enhancement of reasoning capabilities. To fully comprehend this approach, two questions naturally arise: (Q1) What advantages does training with CoT offer compared to training without CoT? (Q2) If there are advantages, what are the underlying mechanisms of explicit CoT training? Analyzing the advantages and mechanisms of CoT training is challenging due to the many factors involved. To address this, we conduct a detailed analysis using clear and controllable data distributions and, for the first time, reveal that CoT training offers the following advantages: (1) Training with CoT markedly improves reasoning generalization, extending it from in-distribution (ID) to both ID and out-of-distribution (OOD) scenarios, while also speeding up convergence; (2) Even when training with CoT includes a certain range of erroneous reasoning steps, it still enables the model to learn reasoning patterns, leading to systematic generalization. We further explore the underlying mechanisms from a circuit perspective: (1) The data distribution (e.g., ratio $\lambda$ and pattern) plays a crucial role in influencing the model's systematic generalization; (2) CoT training (with two-hop facts) internalizes reasoning into a two-stage generalizing circuit, where the number of stages corresponds to the explicit reasoning steps during training. Our findings elucidate the mechanisms underlying explicit CoT training and offer critical insights into tuning strategies for LLMs to achieve robust generalization.