Lumine: An Open Recipe for Building Generalist Agents in 3D Open Worlds

We introduce Lumine, the first open recipe for developing generalist agents capable of completing hours-long complex missions in real time within challenging 3D open-world environments. Lumine adopts a human-like interaction paradigm that unifies perception, reasoning, and action in an end-to-end manner, powered by a vision-language model. It processes raw pixels at 5 Hz to produce precise 30 Hz keyboard-mouse actions and adaptively invokes reasoning only when necessary. Trained in Genshin Impact, Lumine successfully completes the entire five-hour Mondstadt main storyline on par with human-level efficiency and follows natural language instructions to perform a broad spectrum of tasks in both 3D open-world exploration and 2D GUI manipulation across collection, combat, puzzle-solving, and NPC interaction. In addition to its in-domain performance, Lumine demonstrates strong zero-shot cross-game generalization. Without any fine-tuning, it accomplishes 100-minute missions in Wuthering Waves and the full five-hour first chapter of Honkai: Star Rail. These promising results highlight Lumine's effectiveness across distinct worlds and interaction dynamics, marking a concrete step toward generalist agents in open-ended environments.

Social Agent: Mastering Dyadic Nonverbal Behavior Generation via Conversational LLM Agents

We present Social Agent, a novel framework for synthesizing realistic and contextually appropriate co-speech nonverbal behaviors in dyadic conversations. In this framework, we develop an agentic system driven by a Large Language Model (LLM) to direct the conversation flow and determine appropriate interactive behaviors for both participants. Additionally, we propose a novel dual-person gesture generation model based on an auto-regressive diffusion model, which synthesizes coordinated motions from speech signals. The output of the agentic system is translated into high-level guidance for the gesture generator, resulting in realistic movement at both the behavioral and motion levels. Furthermore, the agentic system periodically examines the movements of interlocutors and infers their intentions, forming a continuous feedback loop that enables dynamic and responsive interactions between the two participants. User studies and quantitative evaluations show that our model significantly improves the quality of dyadic interactions, producing natural, synchronized nonverbal behaviors.

Understanding Transformer from the Perspective of Associative Memory

In this paper, we share our reflections and insights on understanding Transformer architectures through the lens of associative memory--a classic psychological concept inspired by human cognition. We start with the basics of associative memory (think simple linear attention) and then dive into two dimensions: Memory Capacity: How much can a Transformer really remember, and how well? We introduce retrieval SNR to measure this and use a kernel perspective to mathematically reveal why Softmax Attention is so effective. We also show how FFNs can be seen as a type of associative memory, leading to insights on their design and potential improvements. Memory Update: How do these memories learn and evolve? We present a unified framework for understanding how different Transformer variants (like DeltaNet and Softmax Attention) update their "knowledge base". This leads us to tackle two provocative questions: 1. Are Transformers fundamentally limited in what they can express, and can we break these barriers? 2. If a Transformer had infinite context, would it become infinitely intelligent? We want to demystify Transformer architecture, offering a clearer understanding of existing designs. This exploration aims to provide fresh insights and spark new avenues for Transformer innovation.

Seed1.5-VL Technical Report

We present Seed1.5-VL, a vision-language foundation model designed to advance general-purpose multimodal understanding and reasoning. Seed1.5-VL is composed with a 532M-parameter vision encoder and a Mixture-of-Experts (MoE) LLM of 20B active parameters. Despite its relatively compact architecture, it delivers strong performance across a wide spectrum of public VLM benchmarks and internal evaluation suites, achieving the state-of-the-art performance on 38 out of 60 public benchmarks. Moreover, in agent-centric tasks such as GUI control and gameplay, Seed1.5-VL outperforms leading multimodal systems, including OpenAI CUA and Claude 3.7. Beyond visual and video understanding, it also demonstrates strong reasoning abilities, making it particularly effective for multimodal reasoning challenges such as visual puzzles. We believe these capabilities will empower broader applications across diverse tasks. In this report, we mainly provide a comprehensive review of our experiences in building Seed1.5-VL across model design, data construction, and training at various stages, hoping that this report can inspire further research. Seed1.5-VL is now accessible at https://www.volcengine.com/ (Volcano Engine Model ID: doubao-1-5-thinking-vision-pro-250428)

Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens

Recent endeavors in Multimodal Large Language Models (MLLMs) aim to unify visual comprehension and generation by combining LLM and diffusion models, the state-of-the-art in each task, respectively. Existing approaches rely on spatial visual tokens, where image patches are encoded and arranged according to a spatial order (e.g., raster scan). However, we show that spatial tokens lack the recursive structure inherent to languages, hence form an impossible language for LLM to master. In this paper, we build a proper visual language by leveraging diffusion timesteps to learn discrete, recursive visual tokens. Our proposed tokens recursively compensate for the progressive attribute loss in noisy images as timesteps increase, enabling the diffusion model to reconstruct the original image at any timestep. This approach allows us to effectively integrate the strengths of LLMs in autoregressive reasoning and diffusion models in precise image generation, achieving seamless multimodal comprehension and generation within a unified framework. Extensive experiments show that we achieve superior performance for multimodal comprehension and generation simultaneously compared with other MLLMs. Project Page: https://DDT-LLaMA.github.io/.

Body of Her: A Preliminary Study on End-to-End Humanoid Agent

Interactive virtual humanoid agent is a crucial interface with the physical world. A relatively complete humanoid agent first needs to have face and body, then possess both verbal and non-verbal (such as eye contact, facial expression, lip motion, gesture, and manipulation) abilities, and finally, it is capable of real-time duplex communication, e.g., the ability to actively interrupt conversations. Most prior systems typically only consider a subset of these elements, leaving a gap from realistic humanoid agent. In this work, we propose a real-time, duplex, interactive end-to-end network capable of modeling realistic agent behaviors, including speech, full-body movements for talking, responding, idling, and manipulation. This system is a multimodal model integrating audio and visual inputs, extended from a pre-trained large language model (LLM). We collect approximately 200,000 hours of audio, around 130,000 hours of video data, and about 20,000 alignment samples to build the model. The final model demonstrates capabilities that are difficult to achieve in previous systems, such as generalized object manipulation. This work performs a preliminary exploration of the end-to-end approach in this field, aiming to inspire further research towards scaling up.

Semantic Gesticulator: Semantics-Aware Co-Speech Gesture Synthesis

In this work, we present Semantic Gesticulator, a novel framework designed to synthesize realistic gestures accompanying speech with strong semantic correspondence. Semantically meaningful gestures are crucial for effective non-verbal communication, but such gestures often fall within the long tail of the distribution of natural human motion. The sparsity of these movements makes it challenging for deep learning-based systems, trained on moderately sized datasets, to capture the relationship between the movements and the corresponding speech semantics. To address this challenge, we develop a generative retrieval framework based on a large language model. This framework efficiently retrieves suitable semantic gesture candidates from a motion library in response to the input speech. To construct this motion library, we summarize a comprehensive list of commonly used semantic gestures based on findings in linguistics, and we collect a high-quality motion dataset encompassing both body and hand movements. We also design a novel GPT-based model with strong generalization capabilities to audio, capable of generating high-quality gestures that match the rhythm of speech. Furthermore, we propose a semantic alignment mechanism to efficiently align the retrieved semantic gestures with the GPT's output, ensuring the naturalness of the final animation. Our system demonstrates robustness in generating gestures that are rhythmically coherent and semantically explicit, as evidenced by a comprehensive collection of examples. User studies confirm the quality and human-likeness of our results, and show that our system outperforms state-of-the-art systems in terms of semantic appropriateness by a clear margin.

MoConVQ: Unified Physics-Based Motion Control via Scalable Discrete Representations

In this work, we present MoConVQ, a novel unified framework for physics-based motion control leveraging scalable discrete representations. Building upon vector quantized variational autoencoders (VQ-VAE) and model-based reinforcement learning, our approach effectively learns motion embeddings from a large, unstructured dataset spanning tens of hours of motion examples. The resultant motion representation not only captures diverse motion skills but also offers a robust and intuitive interface for various applications. We demonstrate the versatility of MoConVQ through several applications: universal tracking control from various motion sources, interactive character control with latent motion representations using supervised learning, physics-based motion generation from natural language descriptions using the GPT framework, and, most interestingly, seamless integration with large language models (LLMs) with in-context learning to tackle complex and abstract tasks.

Robust Dancer: Long-term 3D Dance Synthesis Using Unpaired Data

How to automatically synthesize natural-looking dance movements based on a piece of music is an incrementally popular yet challenging task. Most existing data-driven approaches require hard-to-get paired training data and fail to generate long sequences of motion due to error accumulation of autoregressive structure. We present a novel 3D dance synthesis system that only needs unpaired data for training and could generate realistic long-term motions at the same time. For the unpaired data training, we explore the disentanglement of beat and style, and propose a Transformer-based model free of reliance upon paired data. For the synthesis of long-term motions, we devise a new long-history attention strategy. It first queries the long-history embedding through an attention computation and then explicitly fuses this embedding into the generation pipeline via multimodal adaptation gate (MAG). Objective and subjective evaluations show that our results are comparable to strong baseline methods, despite not requiring paired training data, and are robust when inferring long-term music. To our best knowledge, we are the first to achieve unpaired data training - an ability that enables to alleviate data limitations effectively. Our code is released on https://github.com/BFeng14/RobustDancer

GestureDiffuCLIP: Gesture Diffusion Model with CLIP Latents

The automatic generation of stylized co-speech gestures has recently received increasing attention. Previous systems typically allow style control via predefined text labels or example motion clips, which are often not flexible enough to convey user intent accurately. In this work, we present GestureDiffuCLIP, a neural network framework for synthesizing realistic, stylized co-speech gestures with flexible style control. We leverage the power of the large-scale Contrastive-Language-Image-Pre-training (CLIP) model and present a novel CLIP-guided mechanism that extracts efficient style representations from multiple input modalities, such as a piece of text, an example motion clip, or a video. Our system learns a latent diffusion model to generate high-quality gestures and infuses the CLIP representations of style into the generator via an adaptive instance normalization (AdaIN) layer. We further devise a gesture-transcript alignment mechanism that ensures a semantically correct gesture generation based on contrastive learning. Our system can also be extended to allow fine-grained style control of individual body parts. We demonstrate an extensive set of examples showing the flexibility and generalizability of our model to a variety of style descriptions. In a user study, we show that our system outperforms the state-of-the-art approaches regarding human likeness, appropriateness, and style correctness.