Language model pretraining with next token prediction has proved effective for scaling compute but is limited to the amount of available training data. Scaling reinforcement learning (RL) unlocks a new axis for the continued improvement of artificial intelligence, with the promise that large language models (LLMs) can scale their training data by learning to explore with rewards. However, prior published work has not produced competitive results. In light of this, we report on the training practice of Kimi k1.5, our latest multi-modal LLM trained with RL, including its RL training techniques, multi-modal data recipes, and infrastructure optimization. Long context scaling and improved policy optimization methods are key ingredients of our approach, which establishes a simplistic, effective RL framework without relying on more complex techniques such as Monte Carlo tree search, value functions, and process reward models. Notably, our system achieves state-of-the-art reasoning performance across multiple benchmarks and modalities -- e.g., 77.5 on AIME, 96.2 on MATH 500, 94-th percentile on Codeforces, 74.9 on MathVista -- matching OpenAI's o1. Moreover, we present effective long2short methods that use long-CoT techniques to improve short-CoT models, yielding state-of-the-art short-CoT reasoning results -- e.g., 60.8 on AIME, 94.6 on MATH500, 47.3 on LiveCodeBench -- outperforming existing short-CoT models such as GPT-4o and Claude Sonnet 3.5 by a large margin (up to +550%).

Unsupervised out-of-distribution (OOD) detection aims to identify out-of-domain data by learning only from unlabeled In-Distribution (ID) training samples, which is crucial for developing a safe real-world machine learning system. Current reconstruction-based method provides a good alternative approach, by measuring the reconstruction error between the input and its corresponding generative counterpart in the pixel/feature space. However, such generative methods face the key dilemma, i.e., improving the reconstruction power of the generative model, while keeping compact representation of the ID data. To address this issue, we propose the diffusion-based layer-wise semantic reconstruction approach for unsupervised OOD detection. The innovation of our approach is that we leverage the diffusion model's intrinsic data reconstruction ability to distinguish ID samples from OOD samples in the latent feature space. Moreover, to set up a comprehensive and discriminative feature representation, we devise a multi-layer semantic feature extraction strategy. Through distorting the extracted features with Gaussian noises and applying the diffusion model for feature reconstruction, the separation of ID and OOD samples is implemented according to the reconstruction errors. Extensive experimental results on multiple benchmarks built upon various datasets demonstrate that our method achieves state-of-the-art performance in terms of detection accuracy and speed. Code is available at https://github.com/xbyym/DLSR.

In multivariate spline regression, the number and locations of knots influence the performance and interpretability significantly. However, due to non-differentiability and varying dimensions, there is no desirable frequentist method to make inference on knots. In this article, we propose a fully Bayesian approach for knot inference in multivariate spline regression. The existing Bayesian method often uses BIC to calculate the posterior, but BIC is too liberal and it will heavily overestimate the knot number when the candidate model space is large. We specify a new prior on the knot number to take into account the complexity of the model space and derive an analytic formula in the normal model. In the non-normal cases, we utilize the extended Bayesian information criterion to approximate the posterior density. The samples are simulated in the space with differing dimensions via reversible jump Markov chain Monte Carlo. We apply the proposed method in knot inference and manifold denoising. Experiments demonstrate the splendid capability of the algorithm, especially in function fitting with jumping discontinuity.

This paper offers a detailed investigation of switchback designs in A/B testing, which alternate between baseline and new policies over time. Our aim is to thoroughly evaluate the effects of these designs on the accuracy of their resulting average treatment effect (ATE) estimators. We propose a novel "weak signal analysis" framework, which substantially simplifies the calculations of the mean squared errors (MSEs) of these ATEs in Markov decision process environments. Our findings suggest that (i) when the majority of reward errors are positively correlated, the switchback design is more efficient than the alternating-day design which switches policies in a daily basis. Additionally, increasing the frequency of policy switches tends to reduce the MSE of the ATE estimator. (ii) When the errors are uncorrelated, however, all these designs become asymptotically equivalent. (iii) In cases where the majority of errors are negative correlated, the alternating-day design becomes the optimal choice. These insights are crucial, offering guidelines for practitioners on designing experiments in A/B testing. Our analysis accommodates a variety of policy value estimators, including model-based estimators, least squares temporal difference learning estimators, and double reinforcement learning estimators, thereby offering a comprehensive understanding of optimal design strategies for policy evaluation in reinforcement learning.

The aim of this paper is to establish a causal link between the policies implemented by technology companies and the outcomes they yield within intricate temporal and/or spatial dependent experiments. We propose a novel temporal/spatio-temporal Varying Coefficient Decision Process (VCDP) model, capable of effectively capturing the evolving treatment effects in situations characterized by temporal and/or spatial dependence. Our methodology encompasses the decomposition of the Average Treatment Effect (ATE) into the Direct Effect (DE) and the Indirect Effect (IE). We subsequently devise comprehensive procedures for estimating and making inferences about both DE and IE. Additionally, we provide a rigorous analysis of the statistical properties of these procedures, such as asymptotic power. To substantiate the effectiveness of our approach, we carry out extensive simulations and real data analyses.

Traffic flow prediction (TFP) is a fundamental problem of the Intelligent Transportation System (ITS), as it models the latent spatial-temporal dependency of traffic flow for potential congestion prediction. Recent graph-based models with multiple kinds of attention mechanisms have achieved promising performance. However, existing methods for traffic flow prediction tend to inherit the bias pattern from the dataset and lack interpretability. To this end, we propose a Counterfactual Graph Transformer (CGT) model with an instance-level explainer (e.g., finding the important subgraphs) specifically designed for TFP. We design a perturbation mask generator over input sensor features at the time dimension and the graph structure on the graph transformer module to obtain spatial and temporal counterfactual explanations. By searching the optimal perturbation masks on the input data feature and graph structures, we can obtain the concise and dominant data or graph edge links for the subsequent TFP task. After re-training the utilized graph transformer model after counterfactual perturbation, we can obtain improved and interpretable traffic flow prediction. Extensive results on three real-world public datasets show that CGT can produce reliable explanations and is promising for traffic flow prediction.

Ontologies are formal representations of concepts and complex relationships among them. They have been widely used to capture comprehensive domain knowledge in areas such as biology and medicine, where large and complex ontologies can contain hundreds of thousands of concepts. Especially due to the large size of ontologies, visualisation is useful for authoring, exploring and understanding their underlying data. Existing ontology visualisation tools generally focus on the hierarchical structure, giving much less emphasis to non-hierarchical associations. In this paper we present OntoPlot, a novel visualisation specifically designed to facilitate the exploration of all concept associations whilst still showing an ontology's large hierarchical structure. This hybrid visualisation combines icicle plots, visual compression techniques and interactivity, improving space-efficiency and reducing visual structural complexity. We conducted a user study with domain experts to evaluate the usability of OntoPlot, comparing it with the de facto ontology editor Prot{\'e}g{\'e}. The results confirm that OntoPlot attains our design goals for association-related tasks and is strongly favoured by domain experts.

Cross-region dynamic connectivity, which describes spatio-temporal dependence of neural activity among multiple brain regions of interest (ROIs), can provide important information for understanding cognition. For estimating such connectivity, magnetoencephalography (MEG) and electroencephalography (EEG) are well-suited tools because of their millisecond temporal resolution. However, localizing source activity in the brain requires solving an under-determined linear problem. In typical two-step approaches, researchers first solve the linear problem with general priors assuming independence across ROIs, and secondly quantify cross-region connectivity. In this work, we propose a one-step state-space model to improve estimation of dynamic connectivity. The model treats the mean activity in individual ROIs as the state variable, and describes non-stationary dynamic dependence across ROIs using time-varying auto-regression. Compared with a two-step method, which first obtains the commonly used minimum-norm estimates of source activity, and then fits the auto-regressive model, our state-space model yielded smaller estimation errors on simulated data where the model assumptions held. When applied on empirical MEG data from one participant in a scene-processing experiment, our state-space model also demonstrated intriguing preliminary results, indicating leading and lagged linear dependence between the early visual cortex and a higher-level scene-sensitive region, which could reflect feed-forward and feedback information flow within the visual cortex during scene processing.