Prior research has enhanced the ability of Large Language Models (LLMs) to solve logic puzzles using techniques such as chain-of-thought prompting or introducing a symbolic representation. These frameworks are still usually insufficient to solve complicated logical problems, such as Zebra puzzles, due to the inherent complexity of translating natural language clues into logical statements. We introduce a multi-agent system, ZPS, that integrates LLMs with an off the shelf theorem prover. This system tackles the complex puzzle-solving task by breaking down the problem into smaller, manageable parts, generating SMT (Satisfiability Modulo Theories) code to solve them with a theorem prover, and using feedback between the agents to repeatedly improve their answers. We also introduce an automated grid puzzle grader to assess the correctness of our puzzle solutions and show that the automated grader is reliable by evaluating it in a user-study. Our approach shows improvement in all three LLMs we tested, with GPT-4 showing 166% improvement in the number of fully correct solutions.
相關內容
Time-Lock Puzzles (TLPs) have been developed to securely transmit sensitive information into the future without relying on a trusted third party. Multi-instance TLP is a scalable variant of TLP that enables a server to efficiently find solutions to different puzzles provided by a client at once. Nevertheless, existing multi-instance TLPs lack support for (verifiable) homomorphic computation. To address this limitation, we introduce the "Multi-Instance partially Homomorphic TLP" (MH-TLP), a multi-instance TLP supporting efficient verifiable homomorphic linear combinations of puzzles belonging to a client. It ensures anyone can verify the correctness of computations and solutions. Building on MH-TLP, we further propose the "Multi-instance Multi-client verifiable partially Homomorphic TLP" (MMH-TLP). It not only supports all the features of MH-TLP but also allows for verifiable homomorphic linear combinations of puzzles from different clients. Our schemes refrain from using asymmetric-key cryptography for verification and, unlike most homomorphic TLPs, do not require a trusted third party. A comprehensive cost analysis demonstrates that our schemes scale linearly with the number of clients and puzzles.
An exploit of the Sequential Importance Sampling (SIS) algorithm using Differential Algebra (DA) techniques is derived to develop an efficient particle filter. The filter creates an original kind of particles, called scout particles, that bring information from the measurement noise onto the state prior probability density function. Thanks to the creation of high-order polynomial maps and their inversions, the scouting of the measurements helps the SIS algorithm identify the region of the prior more affected by the likelihood distribution. The result of the technique is two different versions of the proposed Scout Particle Filter (SPF), which identifies and delimits the region where the true posterior probability has high density in the SIS algorithm. Four different numerical applications show the benefits of the methodology both in terms of accuracy and efficiency, where the SPF is compared to other particle filters, with a particular focus on target tracking and orbit determination problems.
Maintaining consistent uncertainty estimates in localization systems is crucial as the perceived uncertainty commonly affects high-level system components, such as control or decision processes. A method for constructing an observability-constrained magnetic field-aided inertial navigation system is proposed to address the issue of erroneous yaw observability, which leads to inconsistent estimates of yaw uncertainty. The proposed method builds upon the previously proposed observability-constrained extended Kalman filter and extends it to work with a magnetic field-based odometry-aided inertial navigation system. The proposed method is evaluated using simulation and real-world data, showing that (i) the system observability properties are preserved, (ii) the estimation accuracy increases, and (iii) the perceived uncertainty calculated by the EKF is more consistent with the true uncertainty of the filter estimates.
While there is a widespread belief that artificial general intelligence (AGI) -- or even superhuman AI -- is imminent, complex problems in expert domains are far from being solved. We argue that such problems require human-AI cooperation and that the current state of the art in generative AI is unable to play the role of a reliable partner due to a multitude of shortcomings, including inability to keep track of a complex solution artifact (e.g., a software program), limited support for versatile human preference expression and lack of adapting to human preference in an interactive setting. To address these challenges, we propose HAI-Co2, a novel human-AI co-construction framework. We formalize HAI-Co2 and discuss the difficult open research problems that it faces. Finally, we present a case study of HAI-Co2 and demonstrate its efficacy compared to monolithic generative AI models.
Most Bundle Adjustment (BA) solvers like the Levenberg-Marquardt algorithm require a good initialization. Instead, initialization-free BA remains a largely uncharted territory. The under-explored Variable Projection algorithm (VarPro) exhibits a wide convergence basin even without initialization. Coupled with object space error formulation, recent works have shown its ability to solve small-scale initialization-free bundle adjustment problem. To make such initialization-free BA approaches scalable, we introduce Power Variable Projection (PoVar), extending a recent inverse expansion method based on power series. Importantly, we link the power series expansion to Riemannian manifold optimization. This projective framework is crucial to solve large-scale bundle adjustment problems without initialization. Using the real-world BAL dataset, we experimentally demonstrate that our solver achieves state-of-the-art results in terms of speed and accuracy. To our knowledge, this work is the first to address the scalability of BA without initialization opening new venues for initialization-free structure-from-motion.
Discrimination mitigation with machine learning (ML) models could be complicated because multiple factors may interweave with each other including hierarchically and historically. Yet few existing fairness measures are able to capture the discrimination level within ML models in the face of multiple sensitive attributes. To bridge this gap, we propose a fairness measure based on distances between sets from a manifold perspective, named as 'harmonic fairness measure via manifolds (HFM)' with two optional versions, which can deal with a fine-grained discrimination evaluation for several sensitive attributes of multiple values. To accelerate the computation of distances of sets, we further propose two approximation algorithms named 'Approximation of distance between sets for one sensitive attribute with multiple values (ApproxDist)' and 'Approximation of extended distance between sets for several sensitive attributes with multiple values (ExtendDist)' to respectively resolve bias evaluation of one single sensitive attribute with multiple values and that of several sensitive attributes with multiple values. Moreover, we provide an algorithmic effectiveness analysis for ApproxDist under certain assumptions to explain how well it could work. The empirical results demonstrate that our proposed fairness measure HFM is valid and approximation algorithms (i.e., ApproxDist and ExtendDist) are effective and efficient.
To address the data scarcity associated with 3D assets, 2D-lifting techniques such as Score Distillation Sampling (SDS) have become a widely adopted practice in text-to-3D generation pipelines. However, the diffusion models used in these techniques are prone to viewpoint bias and thus lead to geometric inconsistencies such as the Janus problem. To counter this, we introduce MT3D, a text-to-3D generative model that leverages a high-fidelity 3D object to overcome viewpoint bias and explicitly infuse geometric understanding into the generation pipeline. Firstly, we employ depth maps derived from a high-quality 3D model as control signals to guarantee that the generated 2D images preserve the fundamental shape and structure, thereby reducing the inherent viewpoint bias. Next, we utilize deep geometric moments to ensure geometric consistency in the 3D representation explicitly. By incorporating geometric details from a 3D asset, MT3D enables the creation of diverse and geometrically consistent objects, thereby improving the quality and usability of our 3D representations.
Diacritization plays a pivotal role in improving readability and disambiguating the meaning of Arabic texts. Efforts have so far focused on marking every eligible character (Full Diacritization). Comparatively overlooked, Partial Diacritzation (PD) is the selection of a subset of characters to be marked to aid comprehension where needed. Research has indicated that excessive diacritic marks can hinder skilled readers -- reducing reading speed and accuracy. We conduct a behavioral experiment and show that partially marked text is often easier to read than fully marked text, and sometimes easier than plain text. In this light, we introduce Context-Contrastive Partial Diacritization (CCPD) -- a novel approach to PD which integrates seamlessly with existing Arabic diacritization systems. CCPD processes each word twice, once with context and once without, and diacritizes only the characters with disparities between the two inferences. Further, we introduce novel indicators for measuring partial diacritization quality, essential for establishing this as a machine learning task. Lastly, we introduce TD2, a Transformer-variant of an established model which offers a markedly different performance profile on our proposed indicators compared to all other known systems.
Variants of the GSEMO algorithm using multi-objective formulations have been successfully analyzed and applied to optimize chance-constrained submodular functions. However, due to the effect of the increasing population size of the GSEMO algorithm considered in these studies from the algorithms, the approach becomes ineffective if the number of trade-offs obtained grows quickly during the optimization run. In this paper, we apply the sliding-selection approach introduced in [21] to the optimization of chance-constrained monotone submodular functions. We theoretically analyze the resulting SW-GSEMO algorithm which successfully limits the population size as a key factor that impacts the runtime and show that this allows it to obtain better runtime guarantees than the best ones currently known for the GSEMO. In our experimental study, we compare the performance of the SW-GSEMO to the GSEMO and NSGA-II on the maximum coverage problem under the chance constraint and show that the SW-GSEMO outperforms the other two approaches in most cases. In order to get additional insights into the optimization behavior of SW-GSEMO, we visualize the selection behavior of SW-GSEMO during its optimization process and show it beats other algorithms to obtain the highest quality of solution in variable instances.
Multi-Label Text Classification (MLTC) is a practical yet challenging task that involves assigning multiple non-exclusive labels to each document. Previous studies primarily focus on capturing label correlations to assist label prediction by introducing special labeling schemes, designing specific model structures, or adding auxiliary tasks. Recently, the $k$ Nearest Neighbor ($k$NN) framework has shown promise by retrieving labeled samples as references to mine label co-occurrence information in the embedding space. However, two critical biases, namely embedding alignment bias and confidence estimation bias, are often overlooked, adversely affecting prediction performance. In this paper, we introduce a DEbiased Nearest Neighbors (DENN) framework for MLTC, specifically designed to mitigate these biases. To address embedding alignment bias, we propose a debiased contrastive learning strategy, enhancing neighbor consistency on label co-occurrence. For confidence estimation bias, we present a debiased confidence estimation strategy, improving the adaptive combination of predictions from $k$NN and inductive binary classifications. Extensive experiments conducted on four public benchmark datasets (i.e., AAPD, RCV1-V2, Amazon-531, and EUR-LEX57K) showcase the effectiveness of our proposed method. Besides, our method does not introduce any extra parameters.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191