ERNIE 5.0 Technical Report

In this report, we introduce ERNIE 5.0, a natively autoregressive foundation model desinged for unified multimodal understanding and generation across text, image, video, and audio. All modalities are trained from scratch under a unified next-group-of-tokens prediction objective, based on an ultra-sparse mixture-of-experts (MoE) architecture with modality-agnostic expert routing. To address practical challenges in large-scale deployment under diverse resource constraints, ERNIE 5.0 adopts a novel elastic training paradigm. Within a single pre-training run, the model learns a family of sub-models with varying depths, expert capacities, and routing sparsity, enabling flexible trade-offs among performance, model size, and inference latency in memory- or time-constrained scenarios. Moreover, we systematically address the challenges of scaling reinforcement learning to unified foundation models, thereby guaranteeing efficient and stable post-training under ultra-sparse MoE architectures and diverse multimodal settings. Extensive experiments demonstrate that ERNIE 5.0 achieves strong and balanced performance across multiple modalities. To the best of our knowledge, among publicly disclosed models, ERNIE 5.0 represents the first production-scale realization of a trillion-parameter unified autoregressive model that supports both multimodal understanding and generation. To facilitate further research, we present detailed visualizations of modality-agnostic expert routing in the unified model, alongside comprehensive empirical analysis of elastic training, aiming to offer profound insights to the community.

DiMo: Discrete Diffusion Modeling for Motion Generation and Understanding

Prior masked modeling motion generation methods predominantly study text-to-motion. We present DiMo, a discrete diffusion-style framework, which extends masked modeling to bidirectional text--motion understanding and generation. Unlike GPT-style autoregressive approaches that tokenize motion and decode sequentially, DiMo performs iterative masked token refinement, unifying Text-to-Motion (T2M), Motion-to-Text (M2T), and text-free Motion-to-Motion (M2M) within a single model. This decoding paradigm naturally enables a quality-latency trade-off at inference via the number of refinement steps.We further improve motion token fidelity with residual vector quantization (RVQ) and enhance alignment and controllability with Group Relative Policy Optimization (GRPO). Experiments on HumanML3D and KIT-ML show strong motion quality and competitive bidirectional understanding under a unified framework. In addition, we demonstrate model ability in text-free motion completion, text-guided motion prediction and motion caption correction without architectural change.Additional qualitative results are available on our project page: https://animotionlab.github.io/DiMo/.

Small Generalizable Prompt Predictive Models Can Steer Efficient RL Post-Training of Large Reasoning Models

Reinforcement learning enhances the reasoning capabilities of large language models but often involves high computational costs due to rollout-intensive optimization. Online prompt selection presents a plausible solution by prioritizing informative prompts to improve training efficiency. However, current methods either depend on costly, exact evaluations or construct prompt-specific predictive models lacking generalization across prompts. This study introduces Generalizable Predictive Prompt Selection (GPS), which performs Bayesian inference towards prompt difficulty using a lightweight generative model trained on the shared optimization history. Intermediate-difficulty prioritization and history-anchored diversity are incorporated into the batch acquisition principle to select informative prompt batches. The small predictive model also generalizes at test-time for efficient computational allocation. Experiments across varied reasoning benchmarks indicate GPS's substantial improvements in training efficiency, final performance, and test-time efficiency over superior baseline methods.

Learning the Mechanism of Catastrophic Forgetting: A Perspective from Gradient Similarity

Catastrophic forgetting during knowledge injection severely undermines the continual learning capability of large language models (LLMs). Although existing methods attempt to mitigate this issue, they often lack a foundational theoretical explanation. We establish a gradient-based theoretical framework to explain catastrophic forgetting. We first prove that strongly negative gradient similarity is a fundamental cause of forgetting. We then use gradient similarity to identify two types of neurons: conflicting neurons that induce forgetting and account for 50%-75% of neurons, and collaborative neurons that mitigate forgetting and account for 25%-50%. Based on this analysis, we propose a knowledge injection method, Collaborative Neural Learning (CNL). By freezing conflicting neurons and updating only collaborative neurons, CNL theoretically eliminates catastrophic forgetting under an infinitesimal learning rate eta and an exactly known mastered set. Experiments on five LLMs, four datasets, and four optimizers show that CNL achieves zero forgetting in in-set settings and reduces forgetting by 59.1%-81.7% in out-of-set settings.

Text-Pass Filter: An Efficient Scene Text Detector

To pursue an efficient text assembling process, existing methods detect texts via the shrink-mask expansion strategy. However, the shrinking operation loses the visual features of text margins and confuses the foreground and background difference, which brings intrinsic limitations to recognize text features. We follow this issue and design Text-Pass Filter (TPF) for arbitrary-shaped text detection. It segments the whole text directly, which avoids the intrinsic limitations. It is noteworthy that different from previous whole text region-based methods, TPF can separate adhesive texts naturally without complex decoding or post-processing processes, which makes it possible for real-time text detection. Concretely, we find that the band-pass filter allows through components in a specified band of frequencies, called its passband but blocks components with frequencies above or below this band. It provides a natural idea for extracting whole texts separately. By simulating the band-pass filter, TPF constructs a unique feature-filter pair for each text. In the inference stage, every filter extracts the corresponding matched text by passing its pass-feature and blocking other features. Meanwhile, considering the large aspect ratio problem of ribbon-like texts makes it hard to recognize texts wholly, a Reinforcement Ensemble Unit (REU) is designed to enhance the feature consistency of the same text and to enlarge the filter's recognition field to help recognize whole texts. Furthermore, a Foreground Prior Unit (FPU) is introduced to encourage TPF to discriminate the difference between the foreground and background, which improves the feature-filter pair quality. Experiments demonstrate the effectiveness of REU and FPU while showing the TPF's superiority.

DeepSynth-Eval: Objectively Evaluating Information Consolidation in Deep Survey Writing

The evolution of Large Language Models (LLMs) towards autonomous agents has catalyzed progress in Deep Research. While retrieval capabilities are well-benchmarked, the post-retrieval synthesis stage--where agents must digest massive amounts of context and consolidate fragmented evidence into coherent, long-form reports--remains under-evaluated due to the subjectivity of open-ended writing. To bridge this gap, we introduce DeepSynth-Eval, a benchmark designed to objectively evaluate information consolidation capabilities. We leverage high-quality survey papers as gold standards, reverse-engineering research requests and constructing "Oracle Contexts" from their bibliographies to isolate synthesis from retrieval noise. We propose a fine-grained evaluation protocol using General Checklists (for factual coverage) and Constraint Checklists (for structural organization), transforming subjective judgment into verifiable metrics. Experiments across 96 tasks reveal that synthesizing information from hundreds of references remains a significant challenge. Our results demonstrate that agentic plan-and-write workflows significantly outperform single-turn generation, effectively reducing hallucinations and improving adherence to complex structural constraints.

AbductiveMLLM: Boosting Visual Abductive Reasoning Within MLLMs

Visual abductive reasoning (VAR) is a challenging task that requires AI systems to infer the most likely explanation for incomplete visual observations. While recent MLLMs develop strong general-purpose multimodal reasoning capabilities, they fall short in abductive inference, as compared to human beings. To bridge this gap, we draw inspiration from the interplay between verbal and pictorial abduction in human cognition, and propose to strengthen abduction of MLLMs by mimicking such dual-mode behavior. Concretely, we introduce AbductiveMLLM comprising of two synergistic components: REASONER and IMAGINER. The REASONER operates in the verbal domain. It first explores a broad space of possible explanations using a blind LLM and then prunes visually incongruent hypotheses based on cross-modal causal alignment. The remaining hypotheses are introduced into the MLLM as targeted priors, steering its reasoning toward causally coherent explanations. The IMAGINER, on the other hand, further guides MLLMs by emulating human-like pictorial thinking. It conditions a text-to-image diffusion model on both the input video and the REASONER's output embeddings to "imagine" plausible visual scenes that correspond to verbal explanation, thereby enriching MLLMs' contextual grounding. The two components are trained jointly in an end-to-end manner. Experiments on standard VAR benchmarks show that AbductiveMLLM achieves state-of-the-art performance, consistently outperforming traditional solutions and advanced MLLMs.

SyncAnyone: Implicit Disentanglement via Progressive Self-Correction for Lip-Syncing in the wild

High-quality AI-powered video dubbing demands precise audio-lip synchronization, high-fidelity visual generation, and faithful preservation of identity and background. Most existing methods rely on a mask-based training strategy, where the mouth region is masked in talking-head videos, and the model learns to synthesize lip movements from corrupted inputs and target audios. While this facilitates lip-sync accuracy, it disrupts spatiotemporal context, impairing performance on dynamic facial motions and causing instability in facial structure and background consistency. To overcome this limitation, we propose SyncAnyone, a novel two-stage learning framework that achieves accurate motion modeling and high visual fidelity simultaneously. In Stage 1, we train a diffusion-based video transformer for masked mouth inpainting, leveraging its strong spatiotemporal modeling to generate accurate, audio-driven lip movements. However, due to input corruption, minor artifacts may arise in the surrounding facial regions and the background. In Stage 2, we develop a mask-free tuning pipeline to address mask-induced artifacts. Specifically, on the basis of the Stage 1 model, we develop a data generation pipeline that creates pseudo-paired training samples by synthesizing lip-synced videos from the source video and random sampled audio. We further tune the stage 2 model on this synthetic data, achieving precise lip editing and better background consistency. Extensive experiments show that our method achieves state-of-the-art results in visual quality, temporal coherence, and identity preservation under in-the wild lip-syncing scenarios.

Knot Forcing: Taming Autoregressive Video Diffusion Models for Real-time Infinite Interactive Portrait Animation

Real-time portrait animation is essential for interactive applications such as virtual assistants and live avatars, requiring high visual fidelity, temporal coherence, ultra-low latency, and responsive control from dynamic inputs like reference images and driving signals. While diffusion-based models achieve strong quality, their non-causal nature hinders streaming deployment. Causal autoregressive video generation approaches enable efficient frame-by-frame generation but suffer from error accumulation, motion discontinuities at chunk boundaries, and degraded long-term consistency. In this work, we present a novel streaming framework named Knot Forcing for real-time portrait animation that addresses these challenges through three key designs: (1) a chunk-wise generation strategy with global identity preservation via cached KV states of the reference image and local temporal modeling using sliding window attention; (2) a temporal knot module that overlaps adjacent chunks and propagates spatio-temporal cues via image-to-video conditioning to smooth inter-chunk motion transitions; and (3) A "running ahead" mechanism that dynamically updates the reference frame's temporal coordinate during inference, keeping its semantic context ahead of the current rollout frame to support long-term coherence. Knot Forcing enables high-fidelity, temporally consistent, and interactive portrait animation over infinite sequences, achieving real-time performance with strong visual stability on consumer-grade GPUs.

SWiT-4D: Sliding-Window Transformer for Lossless and Parameter-Free Temporal 4D Generation

Despite significant progress in 4D content generation, the conversion of monocular videos into high-quality animated 3D assets with explicit 4D meshes remains considerably challenging. The scarcity of large-scale, naturally captured 4D mesh datasets further limits the ability to train generalizable video-to-4D models from scratch in a purely data-driven manner. Meanwhile, advances in image-to-3D generation, supported by extensive datasets, offer powerful prior models that can be leveraged. To better utilize these priors while minimizing reliance on 4D supervision, we introduce SWiT-4D, a Sliding-Window Transformer for lossless, parameter-free temporal 4D mesh generation. SWiT-4D integrates seamlessly with any Diffusion Transformer (DiT)-based image-to-3D generator, adding spatial-temporal modeling across video frames while preserving the original single-image forward process, enabling 4D mesh reconstruction from videos of arbitrary length. To recover global translation, we further introduce an optimization-based trajectory module tailored for static-camera monocular videos. SWiT-4D demonstrates strong data efficiency: with only a single short (<10s) video for fine-tuning, it achieves high-fidelity geometry and stable temporal consistency, indicating practical deployability under extremely limited 4D supervision. Comprehensive experiments on both in-domain zoo-test sets and challenging out-of-domain benchmarks (C4D, Objaverse, and in-the-wild videos) show that SWiT-4D consistently outperforms existing baselines in temporal smoothness. Project page: https://animotionlab.github.io/SWIT4D/