搜索结果：Hows

Distributed knowledge is a key concept in the standard epistemic logic of knowledge-that. In this paper, we propose a corresponding notion of distributed knowledge-how and study its logic. Our framework generalizes two existing traditions in the logic of know-how: the individual-based multi-step framework and the coalition-based single-step framework. In particular, we assume a group can accomplish more than what its individuals can jointly do. The distributed knowledge-how is based on the distributed knowledge-that of a group whose multi-step strategies derive from distributed actions that subgroups can collectively perform. As the main result, we obtain a sound and strongly complete proof system for our logic of distributed knowledge-how, which closely resembles the logic of distributed knowledge-that in both the axioms and the proof method of completeness.

How proteins fold remains a central unsolved problem in biology. While the idea of a folding code embedded in the amino acid sequence was introduced more than 6 decades ago, this code remains undefined. While we now have powerful predictive tools to predict the final native structure of proteins, we still lack a predictive framework for how sequences dictate folding pathways. Two main conceptual models dominate as explanations of folding mechanism: the funnel model, in which folding proceeds through many alternative routes on a rugged, hyperdimensional energy landscape; and the foldon model, which proposes a hierarchical sequence of discrete intermediates. Recent advances on two fronts are now enabling folding studies in unprecedented ways. Powerful experimental approaches; in particular, single-molecule force spectroscopy and hydrogen (deuterium exchange assays) allow time-resolved tracking of the folding process at high resolution. At the same time, computational breakthroughs culminating in algorithms such as AlphaFold have revolutionized static structure prediction, opening opportunities to extend machine learning toward dynamics. Together, these developments mark a turning poi

From the formation of ice in small clusters of water molecules to the mass raids of army ant colonies, the emergent behavior of collectives depends critically on their size. At the same time, common wisdom holds that such behaviors are robust to the loss of individuals. This tension points to the need for a more systematic study of how number influences collective behavior. We initiate this study by focusing on collective behaviors that change abruptly at certain critical numbers of individuals. We show that a subtle modification of standard bifurcation analysis identifies such critical numbers, including those associated with discreteness- and noise-induced transitions. By treating them as instances of the same phenomenon, we show that critical numbers across physical scales and scientific domains commonly arise from competing feedbacks that scale differently with number. We then use this idea to find overlooked critical numbers in past studies of collective behavior and explore the implications for their conclusions. In particular, we highlight how deterministic approximations of stochastic models can fail near critical numbers. We close by distinguishing these qualitative change

Large language models learn and continually learn through the accumulation of gradient-based updates, but how individual pieces of new information affect existing knowledge, leading to both beneficial generalization and problematic hallucination, remains poorly understood. We demonstrate that when learning new information, LLMs exhibit a "priming" effect: learning a new fact can cause the model to inappropriately apply that knowledge in unrelated contexts. To systematically study this phenomenon, we introduce "Outlandish," a carefully curated dataset of 1320 diverse text samples designed to probe how new knowledge permeates through an LLM's existing knowledge base. Using this dataset, we show that the degree of priming after learning new information can be predicted by measuring the token probability of key words before learning. This relationship holds robustly across different model architectures (PALM-2, Gemma, Llama), sizes, and training stages. Finally, we develop two novel techniques to modulate how new knowledge affects existing model behavior: (1) a ``stepping-stone'' text augmentation strategy and (2) an ``ignore-k'' update pruning method. These approaches reduce undesirab

An agent facing a planning problem can use answers to how-to questions to reduce uncertainty and fill knowledge gaps, helping it solve both current and future tasks. However, their open ended nature, where valid answers to "How do I X?" range from executable actions to high-level descriptions of X's sub-goals, makes them challenging for AI agents to ask, and for AI experts to answer, in ways that support efficient planning. We introduce $How^{2}$, a memory agent framework that enables agents to ask how-to questions, store the answers, and reuse them for lifelong learning in interactive environments. We evaluate our approach in Plancraft, a Minecraft crafting environment, where agents must complete an assembly task by manipulating inventory items. Using teacher models that answer at varying levels of abstraction, from executable action sequences to high-level subgoal descriptions, we show that lifelong learning agents benefit most from answers that are abstracted and decoupled from the current state. $How^{2}$ offers a way for LLM-based agents to improve their planning capabilities over time by asking questions in interactive environments.

One of the most puzzling and unsolved challenges in molecular biology is understanding how proteins fold. Despite having advanced predictive tools that can accurately estimate the native structures of proteins, we still lack a comprehensive model that explains how amino acid sequences dictate folding pathways and trajectories. This manuscript takes a fresh approach to this problem by resorting to the principle of least action. This approach enables us to explore an intriguing question: how does a protein achieve its native state at a constant folding rate and within a time frame that is biologically plausible? A response to this inquiry will help us understand why proteins must fold along specific pathways and identify the boundary conditions that restrict their availability. It will also clarify why different folding pathways could be characterized by a common effective folding trajectory. Finally, it will provide a clear explanation for Levinthal's paradox. Our results are expected to pave the way for a more profound understanding of how proteins fold, shedding light on how the amino acid sequence and its surrounding environment encode the protein's folding pathways and, conseque

The logic of goal-directed knowing-how extends the standard epistemic logic with an operator of knowing-how. The knowing-how operator is interpreted as that there exists a strategy such that the agent knows that the strategy can make sure that p. This paper presents a tableau procedure for the multi-agent version of the logic of strategically knowing-how and shows the soundness and completeness of this tableau procedure. This paper also shows that the satisfiability problem of the logic can be decided in PSPACE.

Given conflicting probability estimates for a set of events, how can we quantify how much they conflict? How can we find a single probability distribution that best encapsulates the given estimates? One approach is to minimize a loss function such as binary KL-divergence that quantifies the dissimilarity between the given estimates and the candidate probability distribution. Given a set of events, we characterize the facets of the polytope of coherent probability estimates about those events. We explore two applications of these ideas: eliciting the beliefs of large language models, and merging expert forecasts into a single coherent forecast.

Activation patching is a popular mechanistic interpretability technique, but has many subtleties regarding how it is applied and how one may interpret the results. We provide a summary of advice and best practices, based on our experience using this technique in practice. We include an overview of the different ways to apply activation patching and a discussion on how to interpret the results. We focus on what evidence patching experiments provide about circuits, and on the choice of metric and associated pitfalls.

Big data has ushered in a new wave of predictive power using machine learning models. In this work, we assess what {\it big} means in the context of typical materials-science machine-learning problems. This concerns not only data volume, but also data quality and veracity as much as infrastructure issues. With selected examples, we ask (i) how models generalize to similar datasets, (ii) how high-quality datasets can be gathered from heterogenous sources, (iii) how the feature set and complexity of a model can affect expressivity, and (iv) what infrastructure requirements are needed to create larger datasets and train models on them. In sum, we find that big data present unique challenges along very different aspects that should serve to motivate further work.

We establish a duality between a pair of mirabolic quantum groups, i.e., the mirabolic counterpart of quantum Howe duality.

The arrows of a category are elements of particular sets, the hom-sets. These sets are functorial, and their functoriality specifies how to compose the arrows with other arrows of the same category. In particular, it allows to form diagrams, making many abstract concepts graphically visible. Presheaves and set-valued functors, in general, are not representable, and so their elements are not arrows in the usual sense. They can however still be seen as "arrow-like structures", which can be post-composed but not pre-composed (for the case of set functors), or pre-composed but not post-composed (for the case of presheaves). Therefore, we can still represent their structure graphically. In this exposition we show how to draw and interpret these generalized diagrams, and how to use them to prove theorems. We will then study in detail, and represent graphically, a few concepts of category theory which are often considered hard to visualize: representability, weighted limits and colimits, and Cauchy completion (for unenriched categories). We also sketch how to interpret the more general case of profunctors, and the Day convolution of presheaves.

This guide introduces Large Language Models (LLM) as a highly versatile text analysis method within the social sciences. As LLMs are easy-to-use, cheap, fast, and applicable on a broad range of text analysis tasks, ranging from text annotation and classification to sentiment analysis and critical discourse analysis, many scholars believe that LLMs will transform how we do text analysis. This how-to guide is aimed at students and researchers with limited programming experience, and offers a simple introduction to how LLMs can be used for text analysis in your own research project, as well as advice on best practices. We will go through each of the steps of analyzing textual data with LLMs using Python: installing the software, setting up the API, loading the data, developing an analysis prompt, analyzing the text, and validating the results. As an illustrative example, we will use the challenging task of identifying populism in political texts, and show how LLMs move beyond the existing state-of-the-art.

How much can you say about the gradient of a neural network without computing a loss or knowing the label? This may sound like a strange question: surely the answer is "very little." However, in this paper, we show that gradients are more structured than previously thought. Gradients lie in a predictable low-dimensional subspace which depends on the network architecture and incoming features. Exploiting this structure can significantly improve gradient-free optimization schemes based on directional derivatives, which have struggled to scale beyond small networks trained on toy datasets. We study how to narrow the gap in optimization performance between methods that calculate exact gradients and those that use directional derivatives. Furthermore, we highlight new challenges in overcoming the large gap between optimizing with exact gradients and guessing the gradients.

We investigate the complexity of the satisfiability problem for a modal logic expressing `knowing how' assertions, related to an agent's abilities to achieve a certain goal. We take one of the most standard semantics for this kind of logics based on linear plans. Our main result is a proof that checking satisfiability of a `knowing how' formula can be done in $Σ_2^P$. The algorithm we present relies on eliminating nested modalities in a formula, and then performing multiple calls to a satisfiability checking oracle for propositional logic.

In this paper, we present an alternative interpretation of propositional inquisitive logic as an epistemic logic of knowing how. In our setting, an inquisitive logic formula $α$ being supported by a state is formalized as "knowing how to resolve $α$" (more colloquially, "knowing how $α$ is true") holds on the S5 epistemic model corresponding to the state. Based on this epistemic interpretation, we use a dynamic epistemic logic with both know-how and know-that operators to capture the epistemic information behind the innocent-looking connectives in inquisitive logic. We show that the set of valid know-how formulas corresponds precisely to the inquisitive logic. The main result is a complete axiomatization with intuitive axioms using the full dynamic epistemic language. Moreover, we show that the know-how operator and the dynamic operator can both be eliminated without changing the expressivity over models, which is consistent with the modal translation of inquisitive logic existing in the literature. We hope our framework can give an intuitive alternative interpretation of various concepts and technical results in inquisitive logic, and also provide a powerful and flexible tool to d

Taking the quantization of electromagnetism as the paradigm, we show how this procedure cannot work for Einstein gravity. However, it does work for conformal gravity, a fourth-order derivative, renormalizable theory of gravity that Bender and Mannheim have shown to be ghost free. We show that in any gravity theory gravity cannot be quantized canonically. Rather, because of an interplay between the zero-point energy of gravity and that of its matter source, gravity is quantized purely by its coupling to a quantized matter source, with gravity being intrinsically quantum mechanical. Treating the zero-point energy this way provides a solution to the cosmological constant problem. With the gravitational zero-point energy issue having been ignored in standard Einstein gravity, it is not possible to solve the cosmological constant problem in standard gravity, since without the zero-point contribution gravity does not know where the zero of energy is.