Large Language Models (LLMs) are often misleadingly recognized as having a personality or a set of values. We argue that an LLM can be seen as a superposition of perspectives with different values and personality traits. LLMs exhibit context-dependent values and personality traits that change based on the induced perspective (as opposed to humans, who tend to have more coherent values and personality traits across contexts). We introduce the concept of perspective controllability, which refers to a model's affordance to adopt various perspectives with differing values and personality traits. In our experiments, we use questionnaires from psychology (PVQ, VSM, IPIP) to study how exhibited values and personality traits change based on different perspectives. Through qualitative experiments, we show that LLMs express different values when those are (implicitly or explicitly) implied in the prompt, and that LLMs express different values even when those are not obviously implied (demonstrating their context-dependent nature). We then conduct quantitative experiments to study the controllability of different models (GPT-4, GPT-3.5, OpenAssistant, StableVicuna, StableLM), the effectiveness of various methods for inducing perspectives, and the smoothness of the models' drivability. We conclude by examining the broader implications of our work and outline a variety of associated scientific questions. The project website is available at //sites.google.com/view/llm-superpositions .
相關內容
Stance detection aims to identify the attitude expressed in a document towards a given target. Techniques such as Chain-of-Thought (CoT) prompting have advanced this task, enhancing a model's reasoning capabilities through the derivation of intermediate rationales. However, CoT relies primarily on a model's pre-trained internal knowledge during reasoning, thereby neglecting the valuable external information that is previously unknown to the model. This omission, especially within the unsupervised reasoning process, can affect the model's overall performance. Moreover, while CoT enhances Large Language Models (LLMs), smaller LMs, though efficient operationally, face challenges in delivering nuanced reasoning. In response to these identified gaps, we introduce the Ladder-of-Thought (LoT) for the stance detection task. Constructed through a dual-phase Progressive Optimization Framework, LoT directs the small LMs to assimilate high-quality external knowledge, refining the intermediate rationales produced. These bolstered rationales subsequently serve as the foundation for more precise predictions - akin to how a ladder facilitates reaching elevated goals. LoT achieves a balance between efficiency and performance. Our empirical evaluations underscore LoT's efficacy, marking a 16% improvement over GPT-3.5 and a 10% enhancement compared to GPT-3.5 with CoT on stance detection task.
Molecular dynamics simulations have emerged as a fundamental instrument for studying biomolecules. At the same time, it is desirable to perform simulations of a collection of particles under various conditions in which the molecules can fluctuate. In this paper, we explore and adapt the soft prompt-based learning method to molecular dynamics tasks. Our model can remarkably generalize to unseen and out-of-distribution scenarios with limited training data. While our work focuses on temperature as a test case, the versatility of our approach allows for efficient simulation through any continuous dynamic conditions, such as pressure and volumes. Our framework has two stages: 1) Pre-trains with data mixing technique, augments molecular structure data and temperature prompts, then applies a curriculum learning method by increasing the ratio of them smoothly. 2) Meta-learning-based fine-tuning framework improves sample-efficiency of fine-tuning process and gives the soft prompt-tuning better initialization points. Comprehensive experiments reveal that our framework excels in accuracy for in-domain data and demonstrates strong generalization capabilities for unseen and out-of-distribution samples.
Temporal Knowledge Graph (TKG) is an extension of traditional Knowledge Graph (KG) that incorporates the dimension of time. Reasoning on TKGs is a crucial task that aims to predict future facts based on historical occurrences. The key challenge lies in uncovering structural dependencies within historical subgraphs and temporal patterns. Most existing approaches model TKGs relying on entity modeling, as nodes in the graph play a crucial role in knowledge representation. However, the real-world scenario often involves an extensive number of entities, with new entities emerging over time. This makes it challenging for entity-dependent methods to cope with extensive volumes of entities, and effectively handling newly emerging entities also becomes a significant challenge. Therefore, we propose Temporal Inductive Path Neural Network (TiPNN), which models historical information in an entity-independent perspective. Specifically, TiPNN adopts a unified graph, namely history temporal graph, to comprehensively capture and encapsulate information from history. Subsequently, we utilize the defined query-aware temporal paths to model historical path information related to queries on history temporal graph for the reasoning. Extensive experiments illustrate that the proposed model not only attains significant performance enhancements but also handles inductive settings, while additionally facilitating the provision of reasoning evidence through history temporal graphs.
Computing planar orthogonal drawings with the minimum number of bends is one of the most relevant topics in Graph Drawing. The problem is known to be NP-hard, even when we want to test the existence of a rectilinear planar drawing, i.e., an orthogonal drawing without bends (Garg and Tamassia, 2001). From the parameterized complexity perspective, the problem is fixed-parameter tractable when parameterized by the sum of three parameters: the number of bends, the number of vertices of degree at most two, and the treewidth of the input graph (Di Giacomo et al., 2022). We improve this last result by showing that the problem remains fixed-parameter tractable when parameterized only by the number of vertices of degree at most two plus the number of bends. As a consequence, rectilinear planarity testing lies in \FPT~parameterized by the number of vertices of degree at most two.
Modern depth sensors can generate a huge number of 3D points in few seconds to be latter processed by Localization and Mapping algorithms. Ideally, these algorithms should handle efficiently large sizes of Point Clouds under the assumption that using more points implies more information available. The Eigen Factors (EF) is a new algorithm that solves SLAM by using planes as the main geometric primitive. To do so, EF exhaustively calculates the error of all points at complexity $O(1)$, thanks to the {\em Summation matrix} $S$ of homogeneous points. The solution of EF is highly efficient: i) the state variables are only the sensor poses -- trajectory, while the plane parameters are estimated previously in closed from and ii) EF alternating optimization uses a Newton-Raphson method by a direct analytical calculation of the gradient and the Hessian, which turns out to be a block diagonal matrix. Since we require to differentiate over eigenvalues and matrix elements, we have developed an intuitive methodology to calculate partial derivatives in the manifold of rigid body transformations $SE(3)$, which could be applied to unrelated problems that require analytical derivatives of certain complexity. We evaluate EF and other state-of-the-art plane SLAM back-end algorithms in a synthetic environment. The evaluation is extended to ICL dataset (RGBD) and LiDAR KITTI dataset. Code is publicly available at //github.com/prime-slam/EF-plane-SLAM.
The Distributional Random Forest (DRF) is a recently introduced Random Forest algorithm to estimate multivariate conditional distributions. Due to its general estimation procedure, it can be employed to estimate a wide range of targets such as conditional average treatment effects, conditional quantiles, and conditional correlations. However, only results about the consistency and convergence rate of the DRF prediction are available so far. We characterize the asymptotic distribution of DRF and develop a bootstrap approximation of it. This allows us to derive inferential tools for quantifying standard errors and the construction of confidence regions that have asymptotic coverage guarantees. In simulation studies, we empirically validate the developed theory for inference of low-dimensional targets and for testing distributional differences between two populations.
In this note we highlight some connections of UMAP to the basic principles of Information Geometry. Originally, UMAP was derived from Category Theory observations. However, we posit that it also has a natural geometric interpretation.
Longest Increasing Subsequence (LIS) is a fundamental problem in combinatorics and computer science. Previously, there have been numerous works on both upper bounds and lower bounds of the time complexity of computing and approximating LIS, yet only a few on the equally important space complexity. In this paper, we further study the space complexity of computing and approximating LIS in various models. Specifically, we prove non-trivial space lower bounds in the following two models: (1) the adaptive query-once model or read-once branching programs, and (2) the streaming model where the order of streaming is different from the natural order. As far as we know, there are no previous works on the space complexity of LIS in these models. Besides the bounds, our work also leaves many intriguing open problems.
Large Language Models (LLMs) have shown excellent generalization capabilities that have led to the development of numerous models. These models propose various new architectures, tweaking existing architectures with refined training strategies, increasing context length, using high-quality training data, and increasing training time to outperform baselines. Analyzing new developments is crucial for identifying changes that enhance training stability and improve generalization in LLMs. This survey paper comprehensively analyses the LLMs architectures and their categorization, training strategies, training datasets, and performance evaluations and discusses future research directions. Moreover, the paper also discusses the basic building blocks and concepts behind LLMs, followed by a complete overview of LLMs, including their important features and functions. Finally, the paper summarizes significant findings from LLM research and consolidates essential architectural and training strategies for developing advanced LLMs. Given the continuous advancements in LLMs, we intend to regularly update this paper by incorporating new sections and featuring the latest LLM models.
Graph Neural Networks (GNNs) have been studied from the lens of expressive power and generalization. However, their optimization properties are less well understood. We take the first step towards analyzing GNN training by studying the gradient dynamics of GNNs. First, we analyze linearized GNNs and prove that despite the non-convexity of training, convergence to a global minimum at a linear rate is guaranteed under mild assumptions that we validate on real-world graphs. Second, we study what may affect the GNNs' training speed. Our results show that the training of GNNs is implicitly accelerated by skip connections, more depth, and/or a good label distribution. Empirical results confirm that our theoretical results for linearized GNNs align with the training behavior of nonlinear GNNs. Our results provide the first theoretical support for the success of GNNs with skip connections in terms of optimization, and suggest that deep GNNs with skip connections would be promising in practice.
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