As the latest advancements in natural language processing, large language models (LLMs) have achieved human-level language understanding and generation abilities in many real-world tasks, and even have been regarded as a potential path to the artificial general intelligence. To better facilitate research on LLMs, many open-source LLMs, such as Llama 2 and Falcon, have recently been proposed and gained comparable performances to proprietary models. However, these models are primarily designed for English scenarios and exhibit poor performances in Chinese contexts. In this technical report, we propose YAYI 2, including both base and chat models, with 30 billion parameters. YAYI 2 is pre-trained from scratch on a multilingual corpus which contains 2.65 trillion tokens filtered by our pre-training data processing pipeline. The base model is aligned with human values through supervised fine-tuning with millions of instructions and reinforcement learning from human feedback. Extensive experiments on multiple benchmarks, such as MMLU and CMMLU, consistently demonstrate that the proposed YAYI 2 outperforms other similar sized open-source models.

Lagrangian relaxation is a versatile mathematical technique employed to relax constraints in an optimization problem, enabling the generation of dual bounds to prove the optimality of feasible solutions and the design of efficient propagators in constraint programming (such as the weighted circuit constraint). However, the conventional process of deriving Lagrangian multipliers (e.g., using subgradient methods) is often computationally intensive, limiting its practicality for large-scale or time-sensitive problems. To address this challenge, we propose an innovative unsupervised learning approach that harnesses the capabilities of graph neural networks to exploit the problem structure, aiming to generate accurate Lagrangian multipliers efficiently. We apply this technique to the well-known Held-Karp Lagrangian relaxation for the travelling salesman problem. The core idea is to predict accurate Lagrangian multipliers and to employ them as a warm start for generating Held-Karp relaxation bounds. These bounds are subsequently utilized to enhance the filtering process carried out by branch-and-bound algorithms. In contrast to much of the existing literature, which primarily focuses on finding feasible solutions, our approach operates on the dual side, demonstrating that learning can also accelerate the proof of optimality. We conduct experiments across various distributions of the metric travelling salesman problem, considering instances with up to 200 cities. The results illustrate that our approach can improve the filtering level of the weighted circuit global constraint, reduce the optimality gap by a factor two for unsolved instances up to a timeout, and reduce the execution time for solved instances by 10%.

EODECA (Engineered Ordinary Differential Equations as Classification Algorithm) is a novel approach at the intersection of machine learning and dynamical systems theory, presenting a unique framework for classification tasks [1]. This method stands out with its dynamical system structure, utilizing ordinary differential equations (ODEs) to efficiently handle complex classification challenges. The paper delves into EODECA's dynamical properties, emphasizing its resilience against random perturbations and robust performance across various classification scenarios. Notably, EODECA's design incorporates the ability to embed stable attractors in the phase space, enhancing reliability and allowing for reversible dynamics. In this paper, we carry out a comprehensive analysis by expanding on the work [1], and employing a Euler discretization scheme. In particular, we evaluate EODECA's performance across five distinct classification problems, examining its adaptability and efficiency. Significantly, we demonstrate EODECA's effectiveness on the MNIST and Fashion MNIST datasets, achieving impressive accuracies of $98.06\%$ and $88.21\%$, respectively. These results are comparable to those of a multi-layer perceptron (MLP), underscoring EODECA's potential in complex data processing tasks. We further explore the model's learning journey, assessing its evolution in both pre and post training environments and highlighting its ability to navigate towards stable attractors. The study also investigates the invertibility of EODECA, shedding light on its decision-making processes and internal workings. This paper presents a significant step towards a more transparent and robust machine learning paradigm, bridging the gap between machine learning algorithms and dynamical systems methodologies.

Abstract--When handling streaming graphs, existing graph representation learning models encounter a catastrophic forgetting problem, where previously learned knowledge of these models is easily overwritten when learning with newly incoming graphs. In response, Continual Graph Learning (CGL) emerges as a novel paradigm enabling graph representation learning from static to streaming graphs. Our prior work, Condense and T rain (CaT) [1] is a replay-based CGL framework with a balanced continual learning procedure, which designs a small yet effective memory bank for replaying data by condensing incoming graphs. Although the CaT alleviates the catastrophic forgetting problem, there exist three issues: (1) The graph condensation algorithm derived in CaT only focuses on labelled nodes while neglecting abundant information carried by unlabelled nodes; (2) The continual training scheme of the CaT overemphasises on the previously learned knowledge, limiting the model capacity to learn from newly added memories; (3) Both the condensation process and replaying process of the CaT are time-consuming. In this paper, we propose a PsUdo-label guided Memory b Ank (PUMA) CGL framework, extending from the CaT to enhance its efficiency and effectiveness by overcoming the above-mentioned weaknesses and limits. T o fully exploit the information in a graph, PUMA expands the coverage of nodes during graph condensation with both labelled and unlabelled nodes. Furthermore, a training-from-scratch strategy is proposed to upgrade the previous continual learning scheme for a balanced training between the historical and the new graphs. Besides, PUMA uses a one-time prorogation and wide graph encoders to accelerate the graph condensation and the graph encoding process in the training stage to improve the efficiency of the whole framework. Extensive experiments on four datasets demonstrate the state-of-the-art performance and efficiency over existing methods. Generally, graphs are treated as static data in traditional graph representation learning, where a model is fixed once it has been trained. However, for many scenarios in the real world (e.g., social networks, cation networks and knowledge graphs), graph data are changing and evolving in a streaming manner [5], [6], [7]. Recently, the Continual graph learning (CGL) has emerged to handle the streaming graph by adapting the static graph neural networks (GNNs). In CGL, the most significant challenge is how to address a catastrophic forgetting problem, where a model easily forgets the previously learned knowledge while overempha-sising on the incoming graphs [1], [8], [9], [10].

In this survey, we examine algorithms for conducting credit assignment in artificial neural networks that are inspired or motivated by neurobiology. These processes are unified under one possible taxonomy, which is constructed based on how a learning algorithm answers a central question underpinning the mechanisms of synaptic plasticity in complex adaptive neuronal systems: where do the signals that drive the learning in individual elements of a network come from and how are they produced? In this unified treatment, we organize the ever-growing set of brain-inspired learning schemes into six general families and consider these in the context of backpropagation of errors and its known criticisms. The results of this review are meant to encourage future developments in neuro-mimetic systems and their constituent learning processes, wherein lies an important opportunity to build a strong bridge between machine learning, computational neuroscience, and cognitive science.

Abstract--Kinesthetic Teaching is a popular approach to collecting expert robotic demonstrations of contact-rich tasks for imitation learning (IL), but it typically only measures motion, ignoring the force placed on the environment by the robot. Furthermore, contact-rich tasks require accurate sensing of both reaching and touching, which can be difficult to provide with conventional sensing modalities. We address these challenges with a See-Through-your-Skin (STS) visuotactile sensor, using the sensor both (i) as a measurement tool to improve kinesthetic teaching, and (ii) as a policy input in contact-rich door manipulation tasks. An STS sensor can be switched between visual and tactile modes by leveraging a semi-transparent surface and controllable lighting, allowing for both pre-contact visual sensing and during-contact tactile sensing with a single sensor. First, we propose tactile force matching, a methodology that enables a robot to match forces read during kinesthetic teaching using tactile signals. Second, we develop a policy that controls STS mode switching, allowing a policy to learn the appropriate moment to switch an STS from its visual to its tactile mode. Finally, we study multiple observation configurations to compare and contrast the value of visual and tactile data from an STS with visual data Figure 1: Our STS sensor before and during contact with a cabinet knob from a wrist-mounted eye-in-hand camera. In visual mode, the camera sees through episodes from real-world manipulation experiments, we find that the gel and allows finding and reaching the knob, while tactile mode the inclusion of force matching raises average policy success rates provides contact-based feedback, via gel deformation and resultant by 62.5%, STS mode switching by 30.3%, and STS data as a dot displacement, upon initial contact and during opening. This dot policy input by 42.5%. Our results highlight the utility of seethrough displacement can also be used to measure a signal linearly related to tactile sensing for IL, both for data collection to allow force. Red circles highlight knob in sensor view.

The current trend in developing machine learning models for reading comprehension and logical reasoning tasks is focused on improving the models' abilities to understand and utilize logical rules. This work focuses on providing a novel loss function and accompanying model architecture that has more interpretable components than some other models by representing a common strategy employed by humans when given reading comprehension and logical reasoning tasks. Our strategy involves emphasizing relative accuracy over absolute accuracy and can theoretically produce the correct answer with incomplete knowledge. We examine the effectiveness of this strategy to solve reading comprehension and logical reasoning questions. The models were evaluated on the Re-Clor dataset, a challenging reading comprehension and logical reasoning benchmark. We propose the polytuplet loss function, which forces prioritization of learning the relative correctness of answer choices over learning the true accuracy of each choice. Our results indicate that models employing polytuplet loss outperform existing baseline models, though further research is required to quantify the benefits it may present.

Meta-Learning aims to speed up the learning process on new tasks by acquiring useful inductive biases from datasets of related learning tasks. While, in practice, the number of related tasks available is often small, most of the existing approaches assume an abundance of tasks; making them unrealistic and prone to overfitting. A central question in the meta-learning literature is how to regularize to ensure generalization to unseen tasks. In this work, we provide a theoretical analysis using the PAC-Bayesian theory and present a generalization bound for meta-learning, which was first derived by Rothfuss et al. (2021a). Crucially, the bound allows us to derive the closed form of the optimal hyper-posterior, referred to as PACOH, which leads to the best performance guarantees. We provide a theoretical analysis and empirical case study under which conditions and to what extent these guarantees for meta-learning improve upon PAC-Bayesian per-task learning bounds. The closed-form PACOH inspires a practical meta-learning approach that avoids the reliance on bi-level optimization, giving rise to a stochastic optimization problem that is amenable to standard variational methods that scale well. Our experiments show that, when instantiating the PACOH with Gaussian processes and Bayesian Neural Networks models, the resulting methods are more scalable, and yield state-of-the-art performance, both in terms of predictive accuracy and the quality of uncertainty estimates.

Ten years ago, the idea of tracking your footsteps or your heartbeat was weird. Those dedicated to the pursuit of quantified self knowledge proselytized in TED Talks, while journalists attended conferences and reported on the strange new trend. Today, over 40% of households in the U.S. own a wearable device, according to statistics service Statista. It is not uncommon to hear retirees comparing or boasting about their step count for the day. The quantified self is ascendant.

We analyze a general problem in a crowd-sourced setting where one user asks a question (also called item) and other users return answers (also called labels) for this question. Different from existing crowd sourcing work which focuses on finding the most appropriate label for the question (the "truth"), our problem is to determine a ranking of the users based on their ability to answer questions. We call this problem "ability discovery" to emphasize the connection to and duality with the more well-studied problem of "truth discovery". To model items and their labels in a principled way, we draw upon Item Response Theory (IRT) which is the widely accepted theory behind standardized tests such as SAT and GRE. We start from an idealized setting where the relative performance of users is consistent across items and better users choose better fitting labels for each item. We posit that a principled algorithmic solution to our more general problem should solve this ideal setting correctly and observe that the response matrices in this setting obey the Consecutive Ones Property (C1P). While C1P is well understood algorithmically with various discrete algorithms, we devise a novel variant of the HITS algorithm which we call "HITSNDIFFS" (or HND), and prove that it can recover the ideal C1P-permutation in case it exists. Unlike fast combinatorial algorithms for finding the consecutive ones permutation (if it exists), HND also returns an ordering when such a permutation does not exist. Thus it provides a principled heuristic for our problem that is guaranteed to return the correct answer in the ideal setting. Our experiments show that HND produces user rankings with robustly high accuracy compared to state-of-the-art truth discovery methods. We also show that our novel variant of HITS scales better in the number of users than ABH, the only prior spectral C1P reconstruction algorithm.