sequence alignment

This work addresses the inefficiency of traditional Dynamic Time Warping (DTW) in aligning long sequences due to its inherently serial computation, which hinders effective GPU parallelization. The study presents the first systematic exploration of GPU-oriented parallel alternatives to DTW, introducing four novel algorithms. The first three employ rectangular block-based approximations to accelerate computation, while the fourth, termed ParDTW, achieves exact alignment through diagonal-wise parallelization. ParDTW integrates block matrix processing with a diagonal scheduling strategy, preserving full alignment accuracy while delivering 15–100× speedup over existing methods on long sequences. This breakthrough substantially overcomes the performance limitations of conventional DTW, establishing ParDTW as an efficient and practical solution for large-scale sequence alignment tasks.

This work addresses key bottlenecks in motion sequence temporal alignment—namely, reliance on paired data, cross-domain mapping, and supervised training. We propose a zero-shot, multimodal motion matching framework that performs alignment solely via metric distances among motion patches within a single domain, leveraging local distance computation and optimal transport—without any cross-domain modeling or labeled supervision. The framework supports diverse control inputs, including sketches, semantic labels, audio, or reference motions. Critically, it enables robust and efficient motion retargeting without requiring paired samples or supervised training. Experiments demonstrate substantial reductions in both data acquisition and computational costs. Comprehensive evaluations across multiple control tasks confirm strong generalization capability and practical deployability.

Existing video synchronization methods heavily rely on audio or domain-specific visual cues (e.g., human pose), resulting in poor generalization to audio-free, single-/multi-person, and non-human scenarios; moreover, the lack of a universal, reproducible benchmark hinders progress. This paper proposes the first feature-agnostic, preprocessing-robust framework for general multi-view video synchronization, decoupling feature extraction from temporal offset prediction. We identify and rectify a systematic preprocessing bias inherent in the state-of-the-art SeSyn-Net. Furthermore, we introduce the first open-source evaluation benchmark covering diverse content types, equipped with a synthetic data generation pipeline and a bias-aware evaluation protocol. Under fair, controlled comparisons, our method reduces mean synchronization error by 32% over prior approaches. All code, tools, and datasets are publicly released.

This work addresses the limited parallelizability of the classical dynamic time warping (DTW) algorithm, which suffers from quadratic time and memory complexity. The authors propose Segmental DTW, a novel approach that decomposes global sequence alignment into local subsequence DTW computations that can be executed in parallel, followed by a segment-level dynamic programming step to integrate the partial alignments. This method achieves near-full parallelism while preserving alignment accuracy comparable to standard DTW. Theoretical analysis and empirical evaluation on Chopin Mazurka audio alignment tasks demonstrate that one variant of the proposed method outperforms existing approaches in both computational efficiency and alignment performance.

This work addresses the lack of reliability assessment in local alignments produced by Dynamic Time Warping (DTW). It introduces the first unsupervised reliability metric capable of identifying trustworthy alignment regions without requiring ground-truth labels. The method leverages FlexDTW—a variant of DTW with relaxed boundary constraints—to generate alternative alignment paths, which are then compared against the original DTW path to compute a local consistency-based reliability score. Evaluated on audio-to-audio alignment tasks, the approach demonstrates strong performance across diverse scenarios, achieving an average AUROC of 0.97 and effectively quantifying the uncertainty inherent in DTW alignments.

Current evaluation methods for audio-visual talking head generation rely on frame-level metrics that assume strict temporal alignment between generated and reference videos, rendering them sensitive to natural variations in speech rate, rhythm, and stylistic expression, and thereby introducing assessment bias. This work reframes evaluation as a sequence alignment problem and introduces Soft Dynamic Time Warping (Soft DTW) to align feature trajectories temporally, enhancing robustness to timing offsets while preserving sequential constraints. The proposed unified sequence-level evaluation framework subsumes frame-level metrics as a special case of rigid alignment, enabling compatibility with existing perceptual, identity, and synchronization encoders without modification. Large-scale experiments across 20 methods and 7 datasets demonstrate that the approach yields more stable evaluations with higher cross-dataset consistency, clearly disentangling trade-offs between synchronization and realism, as well as expressiveness and stability.

Existing long video generation methods are limited by teacher models that possess only short-term memory, hindering student models from effectively learning global temporal dependencies and compromising long-term consistency. To address this, this work proposes the Context Forcing framework, which enables aligned supervision between teacher and student models over extended contexts for the first time. Additionally, a Slow-Fast Memory mechanism is introduced to efficiently compress historical information, overcoming the context-length bottleneck in streaming training. The proposed approach supports effective context modeling beyond 20 seconds—achieving a 2–10× improvement over prior methods such as LongLive and Infinite-RoPE—and demonstrates significant gains across multiple metrics of long-video temporal consistency.

This work addresses the detrimental impact of annotation errors—such as mislabeling and temporal misalignment—in video datasets on the performance of models for temporally sensitive tasks. To tackle this issue, the authors propose a model-agnostic approach based on dynamic loss trajectory analysis: by tracking the average loss of each frame across multiple training checkpoints, they construct cumulative sample loss (CSL) trajectories that serve as frame-level learnability fingerprints. This method enables the identification of hard-to-learn samples without requiring ground-truth error labels. Experiments on the EgoPER and Cholec80 datasets demonstrate that the proposed technique effectively detects subtle annotation inaccuracies, exhibiting strong generalization capability and practical utility in real-world scenarios.

Autoregressive long video generation is prone to temporal drift and semantic distortion during inference due to error propagation. This work proposes a dynamic inference-time pruning method that operates entirely within the latent space, without modifying the model architecture or training procedure. By detecting deviations in latent token representations, the method identifies and discards unstable tokens, thereby interrupting the propagation of corrupted contextual information. To the best of our knowledge, this is the first approach to mitigate temporal drift solely at inference time, significantly enhancing long-term temporal consistency in generated videos while effectively suppressing error accumulation—all without requiring additional training or architectural changes.

Existing approaches to multimodal joint generation suffer from structural misalignment, slow inference, mode collapse, and suboptimal output quality. This work proposes a Rectified Flow–based multimodal joint inference algorithm that, for the first time, establishes a generative framework from the perspective of structural alignment without relying on Score Distillation Sampling. By jointly transporting images, videos, and 3D shapes within arbitrary structured latent spaces, the method achieves efficient and aligned generation. It attains new state-of-the-art results in image and video synthesis, while matching the 3D generation quality of existing methods but with inference speeds accelerated by several orders of magnitude—demonstrating a significant balance between structural consistency and visual fidelity.