CyberAgent (Japan)

Designing policies that are both efficient and acceptable for conversational service robots in open and diverse environments is non-trivial. Unlike fixed, hand-tuned parameters, online learning can adapt to non-stationary conditions. In this paper, we study how to adapt a social robot’s speech policy in the wild. During a 12-day in-situ deployment with over 1,400 public encounters, we cast online policy optimization as a multi-armed bandit problem and use Thompson sampling to select among six actions defined by speech rate (slow/normal/fast) and verbosity (concise/detailed). We compare three complementary binary rewards–Ru (user rating), Rc (conversation closure), and Rt (≥2 turns)–and show that each induces distinct arm distributions and interaction behaviors. We complement the online results with offline evaluations that analyze contextual factors (e.g., crowd level, group size) using video-annotated data. Taken together, we distill ready-to-use design lessons for deploying online optimization of speech policies in real public HRI settings.

We propose DeMapGS, a structured Gaussian Splatting framework that jointly optimizes deformable surfaces and surface-attached 2D Gaussian splats. By anchoring splats to a deformable template mesh, our method overcomes topological inconsistencies and enhances editing flexibility, addressing limitations of prior Gaussian Splatting methods that treat points independently. The unified representation in our method supports extraction of high-fidelity diffuse, normal, and displacement maps, enabling the reconstructed mesh to inherit the photorealistic rendering quality of Gaussian Splatting. To support robust optimization, we introduce a gradient diffusion strategy that propagates supervision across the surface, along with an alternating 2D/3D rendering scheme to handle concave regions. Experiments demonstrate that DeMapGS achieves state-of-the-art mesh reconstruction quality and enables downstream applications for Gaussian splats such as editing and cross-object manipulation through a shared parametric surface.

We introduce MarkupDM, a multimodal markup document model that represents graphic design as an interleaved multimodal document consisting of both markup language and images. Unlike existing holistic approaches that rely on an element-by-attribute grid representation, our representation accommodates variable-length elements, type-dependent attributes, and text content. Inspired by fill-in-the-middle training in code generation, we train the model to complete the missing part of a design document from its surrounding context, allowing it to treat various design tasks in a unified manner. Our model also supports image generation by predicting discrete image tokens through a specialized tokenizer with support for image transparency. We evaluate MarkupDM on three tasks, attribute value, image, and text completion, and demonstrate that it can produce plausible designs consistent with the given context. To further illustrate the flexibility of our approach, we evaluate our approach on a new instruction-guided design completion task where our instruction-tuned MarkupDM compares favorably to state-of-the-art image editing models, especially in textual completion. These findings suggest that multimodal language models with our document representation can serve as a versatile foundation for broad design automation.

Text removal is a crucial task in computer vision with applications such as privacy preservation, image editing, and media reuse. While existing research has primarily focused on scene text removal in natural images, limitations in current datasets hinder out-of-domain generalization or accurate evaluation. In particular, widely used benchmarks such as SCUT-EnsText suffer from ground truth artifacts due to manual editing, overly simplistic text backgrounds, and evaluation metrics that do not capture the quality of generated results. To address these issues, we introduce an approach to synthesizing a text removal benchmark applicable to domains other than scene texts. Our dataset features text rendered on complex backgrounds using object-aware placement and vision-language model-generated content, ensuring clean ground truth and challenging text removal scenarios. The dataset is available at https://huggingface.co/datasets/cyberagent/OTR .

Three-dimensional match-3 puzzle games (Match 3D) are a sub-genre of tile-matching puzzles in which players identify and remove sets of three identical 3D objects. Sustaining long-term player engagement in Match 3D games requires diverse and appropriately difficult levels, yet manual design is resource-intensive. In this paper, we present an on-device level generation model that produces a virtually infinite number of levels with minimal designer intervention. By applying reinforcement learning (RL), our model generates levels designed to meet specified difficulty targets through a reward function based on completion time estimates derived from human play data. We evaluate our system by assessing the accuracy of difficulty matching, the diversity of generated content, and the run-time performance on smartphones via ONNX and Unity Barracuda. Experimental results confirm that the RL approach generates diverse, challenging, and engaging levels, thereby offering a sustainable solution for Match 3D puzzle game design. Furthermore, our proposed system has been seamlessly integrated into a commercial Match 3D game application, providing users with a real-world gameplay experience. To the best of our knowledge, this represents the first successful deployment of PCGRL for on-device, real-time level generation in a commercial mobile puzzle game.

This work leverages shadow cues in a scene to infer the surrounding illumination of the shadow-casting object. Unlike prior works that optimize a discrete environment map, we model scene illumination using a mixture of spherical Gaussians (SGs). SG illumination provides more intuitive relations to shadow appearance and offers a more compact parameterization compared to discrete environment maps. To estimate SG parameters, we employ an SG-based, differentiable, closed-form rendering equation to explain the shading of the shadow plane and minimize a photometric loss between the rendered and observed shadow plane shading. Experiments on synthetic and real-world images under various surrounding illumination demonstrate that our method estimates illumination more accurately than approaches based on discrete environment maps. With our estimated lighting, consistent shadow effects are realized when blending virtual objects into real-world images<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup><sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup>Code is available at https://github.com/CyberAgentAILab/ShadowSG..

対話応答生成の分野において，背景情報や対話履歴を活用して，ユーザ個人の嗜好に合わせた応答生成の実現を目的とするパーソナライゼーションが注目されている．先行研究において，応答内容の適応だけでなく，システムがユーザと類似した発話スタイルを用いることが，ユーザの好感度を高める要因になることが示唆されている．しかしながら，スタイル類似に関する評価の多くは，対話の当事者ではない第三者による客観的な評価が用いられており，ユーザ自身の知覚に基づく主観的な評価との違いが十分に検討されていない．本研究では，非タスク指向の対話設定を対象とし，英語と日本語の両言語において，主観的・客観的スタイル類似度およびユーザによる対話の好ましさに関する人手評価を付与した新たなデータセットを構築した．分析の結果，ユーザによって知覚されたスタイル類似度は対話の好ましさと高い正の相関を示した一方で，客観的なスタイル類似度との間には明確な相関が見られなかった．本研究は，パーソナライゼーションにおけるスタイル評価において，評価主体の切り分けの必要性を示す経験的知見を提供する．

To explore the interaction design of an autonomous social robot stationed in a hotel room, we conducted a Wizard of Oz study. We developed a teleoperated robotic system that appears to move autonomously through voice-to-synthesis processing. Comparing the evaluation of the latest autonomous case with one of these teleoperated cases, the results show that it is possible to construct a robotic system that is more highly rated in terms of warmth, competence, intention to use, enjoyment of use, and recommendation. It was suggested that the required dialogue skills include the cognitive ability to recognize guest behavior, language comprehension, and the ability to respond to a wide range of information and topics. In addition, insights into the content of the interactive service were gained in terms of the importance of the social robot going beyond its role as a service provider to create a more equal relationship with the guest and demonstrate a deeper understanding of the guest. Through this research, it is hoped that with future technological advances, social robots will create a new form of customer service for the lodging experience.