arXiv subjects

Dongjie Fu

Publications and source records attributed to Dongjie Fu.

15 recordsLinked to original sources

arXiv2026

Omni Interaction Agent Technical Report

In this work, we present Gander, an end-to-end model that unifies omni perception, realtime interaction, and agentic capabilities within a single framework. In contrast to turn-based conventional paradigms, Gander continuously receives streaming inputs across multiple modalities, including video, speech, and text, enabling natural full-duplex interaction in both everyday conversations and complex workflow-oriented agent scenarios. Users can interrupt the model at any time, while the model can also proactively provide intermediate feedback or ask follow up questions. To natively support these capabilities, Gander adopts two key architectural designs: 1) It employs a Cerebellum-Brain collaborative framework, in which the Cerebellum is responsible for realtime interaction and omni conversational capabilities, while the Brain handles complex reasoning and higher-level agentic tasks. The two components interact continuously through tool calling and the agent orchestration runtime. 2) The Cerebellum is built upon a streaming Thinker-Talker architecture, user inputs and model outputs are further flattened into an ordered token stream at the chunk level, providing a unified representation for low latency, continuous interaction. We conduct comprehensive evaluations of Gander across four dimensions: conversational ability, omni understanding, interactive capability, and agentic intelligence. Internal human evaluations demonstrate that Gander maintains the natural and expressive spoken dialogue capabilities of SOTA open source models while achieving competitive performance in omni interaction. Gander also demonstrates robustness in challenging real-world scenarios, including background noise interference, multi-party interactions, and backchannel communication. We release Gander together with its models, code, and data to facilitate further research and development in the community.

eess.AS↗

arXiv2026

Magpie: Real-Time World Renderer for Interactive Games

Modern game development relies heavily on conventional graphics pipelines. High-quality visual content requires modeling, material authoring, animation, lighting, effects, and runtime optimization, making asset production expensive and extending the development cycle of game prototypes. Recently, video foundation models are beginning to change film and video production, but games differ from linear media, they require not only continuous and realistic imagery, but also stable and reproducible gameplay rules, object states, and interaction outcomes. We present Magpie, a real-time generative world-rendering system for interactive games. Magpie separates gameplay execution from visual generation. Designers define scenes and rules in a game engine. At runtime, the Game Engine resolves player actions and maintains world state, while an independent Render Server generates visual output from white-box frames produced by the engine. Magpie provides a system-level implementation path for applying generative models to real-time game rendering. It preserves gameplay designability and reproducibility, and reduces the dependence of early game prototypes on complete visual assets.

cs.CV↗

arXiv2026

VoxZip: Semantic-Anchored Temporal KV Cache Compression for Long-Context Audio Inference

Recent advancements in Speech Large Language Models have demonstrated remarkable capabilities in understanding complex audio tasks. Despite this progress, their long-context inference remains severely bottlenecked by prohibitive KV cache memory demands. Existing text-centric compression methods struggle here, often disrupting speech continuity or discarding crucial semantic cues. To address this, we propose VoxZip, a train-free, two-stage semantic-anchored KV cache compression framework. The first stage uses automatic speech recognition (ASR) transcriptions as explicit semantic anchors to temporally align, compress, and fuse audio tokens, significantly reducing the initial KV cache while elevating token information density. To further improve the compression ratio, the second stage employs a dynamic filtering strategy based on temporally decayed accumulated attention to evict non-essential tokens while mitigating early-token bias. Comprehensive evaluations on Qwen3-Omni across six diverse audio benchmarks demonstrate the superiority of our approach. VoxZip excels in long-audio reasoning and consistently maintains high-fidelity perception on short-form tasks. Notably, it sustains over 90\% of the uncompressed baseline performance even under an aggressive 20x KV cache compression in long-context scenarios. Furthermore, at a 4x compression ratio, VoxZip yields a 1.9x increase in inference throughput alongside a 3.3x reduction in peak memory overhead. Code and models will be available at https://github.com/MM-Speech/VoxZip.

cs.AI↗

arXiv2026

X$^3$-OPD: Distilling Reasoning into Large Audio-Language Models via On-Policy Alignment

While large audio-language models have achieved remarkable progress in auditory perception, they still lag behind text-based large language models in deep logical reasoning, primarily due to the scarcity of high-quality audio reasoning data. To bridge this gap, we propose X$^3$-OPD, a cross-modal on-policy distillation framework that transfers reasoning capabilities from a powerful text teacher to an audio-language student. During training, the student generates reasoning trajectories conditioned on its own acoustic perception, while the teacher provides token-level guidance using matched textual inputs and verified answers. We further construct a three-tier symmetric corpus covering textual reasoning rendered into speech, audio-event reasoning grounded in complex acoustic scenes, and spoken-dialogue reasoning involving paralinguistic cues. This design extends cross-modal distillation beyond textually recoverable content to reasoning grounded in non-linguistic events, prosody, and conversational context. Experiments on MMSU, MMAU, BIG Bench Audio, and MMAR demonstrate that X$^3$-OPD substantially improves audio-grounded reasoning and chain-of-thought quality while largely preserving the model's existing capabilities under domain shift.

cs.LG↗

arXiv2026

Auto-AEG: Scalable Data Construction for Open-Vocabulary Audio Event Grounding

Large Audio-Language Models (LALMs) reason fluently about sound yet struggle to localize precisely when events occur, while classical Sound Event Detection attains frame-level precision only over a closed label set. At the intersection of these paradigms lies the task of Open-Vocabulary Audio Event Grounding: predicting all time intervals of a target sound event described by an arbitrary natural language query. Progress is bottlenecked by data scarcity: no large-scale resource provides open-vocabulary onset/offset supervision, and manual temporal annotation is prohibitively expensive. To address this, we introduce Auto-AEG, a scalable pipeline that constructs such supervision by automatic data construction and model fine-tuning. It pairs programmatically synthesized clips, which carry placement-exact ground-truth intervals for supervised cold-start, with multi-model pseudo-labels on real-world audio that supply the reward signal for reinforcement learning. Training with this pipeline yields large temporal-localization gains (+73.9% and +23.1% mIoU over zero-shot) on AEGBench, an independent difficulty-stratified benchmark we release, and these gains generalize to held-out SED and other audio grounding benchmarks. Our results show that automatically constructed data, coupled with interval-aware reward design, provides an effective data-side route to expanding the temporal localization capability of LALMs. AEGBench: https://huggingface.co/datasets/zihan-audio/AEGBench

cs.SD↗

arXiv2026

MoGeFlow: Flowing Through Motion Codebook Geometry for Text-to-Motion Generation

Vector-quantized motion tokenizers provide a compact discrete interface for text-to-motion generation, but most motion-code priors treat code indices as unordered categorical labels. This view overlooks a key property of motion codes: they are decoder-bound prototypes of physical movement, and their learned codebooks can carry meaningful local kinematic geometry. We verify this property through codebook diagnostics. Distances between learned PartVQ group-specific codes align with local motion-prototype distances, shuffled controls remove this alignment, and replacing codes with progressively farther neighbors induces monotonically larger decoded motion changes. These results show that motion codebooks exhibit measurable, non-random, and decoder-causal geometry. Based on this observation, we propose \textbf{MoGeFlow}, a text-to-motion model that generates through motion codebook geometry. MoGeFlow represents each motion-code frame as a structured set of PartVQ group-specific code embeddings, learns a text-conditioned continuous flow over these frame states, and projects terminal states back to valid motion codes for frozen decoding. This preserves the compactness and validity of discrete tokenization while replacing categorical code prediction with geometry-aware codebook-space generation. Experiments set new state of the art in R-Precision on HumanML3D and KIT-ML, achieve the best HumanML3D MultiModal Distance and KIT-ML FID among generated methods, and obtain the best MotionMillion R@1, R@2, R@3, and FID under the benchmark protocol.

cs.GR↗

arXiv2026

KV-Control: Parameter-Efficient K/V Injection for Trajectory-Controlled Text-to-Motion

Text-conditioned 3D human motion models now synthesize plausible motions from prompts, but practical animation and embodied-agent workflows rarely stop at text: a character may need to follow a sketched root path, hit an end-effector target, or satisfy a multi-joint trajectory while still preserving the gait, style, and intent described by language. This exposes a control trade-off. A trajectory controller should be precise without overwriting the pretrained text-conditioned motion prior, yet existing solutions either duplicate large portions of the generator to regain per-layer control access or move much of the cost to test-time optimization. We introduce KV-Control, a compact attention-side control interface for frozen masked text-to-motion transformers. The key idea is to make geometric constraints available as memory inside self-attention rather than injecting them through a global pose token or enforcing them only at the output side. To support this interface, we co-design a part-tokenized motion substrate and controller: \textbf{PartVQ} learns anatomy-aligned part codebooks, T-Concat exposes each frame--part token as an attention-addressable site, and KV-Control injects control-conditioned key/value memories at every self-attention layer while preserving the pretrained query stream, text cross-attention, FFN, and all backbone weights. The resulting adapter adds only trainable injection parameters atop a shared trajectory encoder, yet tracks root and multi-joint constraints with sub-centimeter accuracy under the inherited refinement protocol while retaining text-conditioned motion quality. KV-Control reframes trajectory conditioning as lightweight memory retrieval, providing a small, precise, and transparent control interface for text-to-motion generation.

cs.CV↗

arXiv2026

AnchorRoute: Human Motion Synthesis with Interval-Routed Sparse Contro

Sparse anchors provide a compact interface for human motion authoring: users specify a few root positions, planar trajectory samples, or body-point targets, while the system synthesizes the full-body motion that completes the under-specified intent. We present AnchorRoute, a sparse-anchor motion synthesis framework that uses anchors as a shared scaffold for both generation and refinement. Before generation, AnchorRoute converts sparse anchors into anchor-condition features and injects the resulting condition memory into a frozen Transition Masked Diffusion prior through AnchorKV and dual-context conditioning. This preserves the generation quality of the pretrained text-to-motion prior while learning sparse spatial control. After generation, the same anchors are evaluated as residuals: their timestamps define refinement intervals, and their residuals determine where correction should be concentrated. RouteSolver then refines the motion by projecting soft-token updates onto anchor-defined piecewise-affine interval bases. This couples generation-time anchor conditioning with residual-routed refinement under one anchor scaffold. AnchorRoute supports root-3D, planar-root, and body-point control within the same formulation. In benchmark evaluations, AnchorRoute outperforms prior sparse-control methods under the sparse keyjoint protocol and consistently improves anchor adherence across control families. The results show that the learned anchor-conditioned generator and RouteSolver refinement are complementary: the generator preserves text-motion quality, while RouteSolver provides a controllable path toward stronger anchor adherence.

cs.GR↗

arXiv2026

UMo: Unified Sparse Motion Modeling for Real-Time Co-Speech Avatars

Speech-driven gestures and facial animations are fundamental to expressive digital avatars in games, virtual production, and interactive media. However, existing methods are either limited to a single modality for audio motion alignment, failing to fully utilize the potential of massive human motion data, or are constrained by the representation ability and throughput of multimodal models, which makes it difficult to achieve high-quality motion generation or real-time performance. We present UMo, a unified sparse motion modeling architecture for real-time co-speech avatars, which processes text, audio, and motion tokens within a unified formulation. Leveraging a spatially sparse Mixture-of-Experts framework and a temporally sparse, keyframe-centric design, UMo efficiently performs real-time dense reconstruction, enabling temporally coherent and high-fidelity animation generation for both facial expressions and gestures. Furthermore, we implement a multi-stage training strategy with targeted audio augmentation to enhance acoustic diversity and semantic consistency. Consequently, UMo preserves fine-grained speech-motion alignment even under strict latency constraints. Extensive quantitative and qualitative evaluations show that UMo achieves better output quality under low latency and real-time performance constraints, offering a practical solution for high-fidelity real-time co-speech avatars.

cs.GR↗

arXiv2026

Character Beyond Speech: Leveraging Role-Playing Evaluation in Audio Large Language Models via Reinforcement Learning

The rapid evolution of multimodal large models has revolutionized the simulation of diverse characters in speech dialogue systems, enabling a novel interactive paradigm. Character attributes are manifested not only in textual responses but also through vocal features, as speech conveys rich paralinguistic information that is challenging to quantify. This poses significant difficulties in evaluating the character alignment of role-playing agents. To address these challenges, we present RoleJudge, an evaluation framework that leverages audio large language models to systematically assess the alignment between speech and character across multiple modalities and dimensions. Furthermore, we introduce RoleChat, the first voice role-playing evaluation dataset enriched with chain-of-thought reasoning annotations, comprising a diverse set of authentic and LLM-generated speech samples. Utilizing this dataset, we implement a multi-stage training paradigm and incorporate Standard Alignment in reinforcement learning to mitigate reward misalignment during optimization. Experimental results in terms of accuracy and subjective assessment demonstrate that RoleJudge outperforms various baseline models, validating the effectiveness of our multidimensional evaluation framework.

cs.LG↗

arXiv2026

X-OPD: Cross-Modal On-Policy Distillation for Capability Alignment in Speech LLMs

While the shift from cascaded dialogue systems to end-to-end (E2E) speech Large Language Models (LLMs) improves latency and paralinguistic modeling, E2E models often exhibit a significant performance degradation compared to their text-based counterparts. The standard Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) training methods fail to close this gap. To address this, we propose X-OPD, a novel Cross-Modal On-Policy Distillation framework designed to systematically align the capabilities of Speech LLMs to their text-based counterparts. X-OPD enables the Speech LLM to explore its own distribution via on-policy rollouts, where a text-based teacher model evaluates these trajectories and provides token-level feedback, effectively distilling teacher's capabilities into student's multi-modal representations. Extensive experiments across multiple benchmarks demonstrate that X-OPD significantly narrows the gap in complex tasks while preserving the model's inherent capabilities.

eess.AS↗

arXiv2025

MARS-Sep: Multimodal-Aligned Reinforced Sound Separation

Universal sound separation faces a fundamental misalignment: models optimized for low-level signal metrics often produce semantically contaminated outputs, failing to suppress perceptually salient interference from acoustically similar sources. We introduce a preference alignment perspective, analogous to aligning LLMs with human intent. To address this, we introduce MARS-Sep, a reinforcement learning framework that reformulates separation as decision making. Instead of simply regressing ground-truth masks, MARS-Sep learns a factorized Beta mask policy that is steered by a preference reward model and optimized by a stable, clipped trust-region surrogate. The reward, derived from a progressively-aligned audio-text-vision encoder, directly incentivizes semantic consistency with query prompts. Extensive experiments on multiple benchmarks demonstrate consistent gains in Text-, Audio-, and Image-Queried separation, with notable improvements in signal metrics and semantic quality. Our code is available at https://github.com/mars-sep/MARS-Sep. Sound separation samples are available at https://mars-sep.github.io/.

cs.SD↗

arXiv2025

MOGO: Residual Quantized Hierarchical Causal Transformer for High-Quality and Real-Time 3D Human Motion Generation

Recent advances in transformer-based text-to-motion generation have led to impressive progress in synthesizing high-quality human motion. Nevertheless, jointly achieving high fidelity, streaming capability, real-time responsiveness, and scalability remains a fundamental challenge. In this paper, we propose MOGO (Motion Generation with One-pass), a novel autoregressive framework tailored for efficient and real-time 3D motion generation. MOGO comprises two key components: (1) MoSA-VQ, a motion scale-adaptive residual vector quantization module that hierarchically discretizes motion sequences with learnable scaling to produce compact yet expressive representations; and (2) RQHC-Transformer, a residual quantized hierarchical causal transformer that generates multi-layer motion tokens in a single forward pass, significantly reducing inference latency. To enhance semantic fidelity, we further introduce a text condition alignment mechanism that improves motion decoding under textual control. Extensive experiments on benchmark datasets including HumanML3D, KIT-ML, and CMP demonstrate that MOGO achieves competitive or superior generation quality compared to state-of-the-art transformer-based methods, while offering substantial improvements in real-time performance, streaming generation, and generalization under zero-shot settings.

cs.CV↗

arXiv2025

OmniChat: Enhancing Spoken Dialogue Systems with Scalable Synthetic Data for Diverse Scenarios

With the rapid development of large language models, researchers have created increasingly advanced spoken dialogue systems that can naturally converse with humans. However, these systems still struggle to handle the full complexity of real-world conversations, including audio events, musical contexts, and emotional expressions, mainly because current dialogue datasets are constrained in both scale and scenario diversity. In this paper, we propose leveraging synthetic data to enhance the dialogue models across diverse scenarios. We introduce ShareChatX, the first comprehensive, large-scale dataset for spoken dialogue that spans diverse scenarios. Based on this dataset, we introduce OmniChat, a multi-turn dialogue system with a heterogeneous feature fusion module, designed to optimize feature selection in different dialogue contexts. In addition, we explored critical aspects of training dialogue systems using synthetic data. Through comprehensive experimentation, we determined the ideal balance between synthetic and real data, achieving state-of-the-art results on the real-world dialogue dataset DailyTalk. We also highlight the crucial importance of synthetic data in tackling diverse, complex dialogue scenarios, especially those involving audio and music. For more details, please visit our demo page at \url{https://sharechatx.github.io/}.

cs.CL↗

arXiv2024

Mogo: RQ Hierarchical Causal Transformer for High-Quality 3D Human Motion Generation

In the field of text-to-motion generation, Bert-type Masked Models (MoMask, MMM) currently produce higher-quality outputs compared to GPT-type autoregressive models (T2M-GPT). However, these Bert-type models often lack the streaming output capability required for applications in video game and multimedia environments, a feature inherent to GPT-type models. Additionally, they demonstrate weaker performance in out-of-distribution generation. To surpass the quality of BERT-type models while leveraging a GPT-type structure, without adding extra refinement models that complicate scaling data, we propose a novel architecture, Mogo (Motion Only Generate Once), which generates high-quality lifelike 3D human motions by training a single transformer model. Mogo consists of only two main components: 1) RVQ-VAE, a hierarchical residual vector quantization variational autoencoder, which discretizes continuous motion sequences with high precision; 2) Hierarchical Causal Transformer, responsible for generating the base motion sequences in an autoregressive manner while simultaneously inferring residuals across different layers. Experimental results demonstrate that Mogo can generate continuous and cyclic motion sequences up to 260 frames (13 seconds), surpassing the 196 frames (10 seconds) length limitation of existing datasets like HumanML3D. On the HumanML3D test set, Mogo achieves a FID score of 0.079, outperforming both the GPT-type model T2M-GPT (FID = 0.116), AttT2M (FID = 0.112) and the BERT-type model MMM (FID = 0.080). Furthermore, our model achieves the best quantitative performance in out-of-distribution generation.

cs.CV↗