Archives AI News

arXiv:2509.02278v1 Announce Type: cross Abstract: Singing-driven 3D head animation is a challenging yet promising task with applications in virtual avatars, entertainment, and education. Unlike speech, singing involves richer emotional nuance, dynamic prosody, and lyric-based semantics, requiring the synthesis of fine-grained, temporally coherent facial motion. Existing speech-driven approaches often produce oversimplified, emotionally flat, and semantically inconsistent results, which are insufficient for singing animation. To address this, we propose Think2Sing, a diffusion-based framework that leverages pretrained large language models to generate semantically coherent and temporally consistent 3D head animations, conditioned on both lyrics and acoustics. A key innovation is the introduction of motion subtitles, an auxiliary semantic representation derived through a novel Singing Chain-of-Thought reasoning process combined with acoustic-guided retrieval. These subtitles contain precise timestamps and region-specific motion descriptions, serving as interpretable motion priors. We frame the task as a motion intensity prediction problem, enabling finer control over facial regions and improving the modeling of expressive motion. To support this, we create a multimodal singing dataset with synchronized video, acoustic descriptors, and motion subtitles, enabling diverse and expressive motion learning. Extensive experiments show that Think2Sing outperforms state-of-the-art methods in realism, expressiveness, and emotional fidelity, while also offering flexible, user-controllable animation editing.

arXiv:2509.01964v1 Announce Type: cross Abstract: Gaussian Splatting (GS), a recent technique for converting discrete points into continuous spatial representations, has shown promising results in 3D scene modeling and 2D image super-resolution. In this paper, we explore its untapped potential for image inpainting, which demands both locally coherent pixel synthesis and globally consistent semantic restoration. We propose the first image inpainting framework based on 2D Gaussian Splatting, which encodes incomplete images into a continuous field of 2D Gaussian splat coefficients and reconstructs the final image via a differentiable rasterization process. The continuous rendering paradigm of GS inherently promotes pixel-level coherence in the inpainted results. To improve efficiency and scalability, we introduce a patch-wise rasterization strategy that reduces memory overhead and accelerates inference. For global semantic consistency, we incorporate features from a pretrained DINO model. We observe that DINO's global features are naturally robust to small missing regions and can be effectively adapted to guide semantic alignment in large-mask scenarios, ensuring that the inpainted content remains contextually consistent with the surrounding scene. Extensive experiments on standard benchmarks demonstrate that our method achieves competitive performance in both quantitative metrics and perceptual quality, establishing a new direction for applying Gaussian Splatting to 2D image processing.

arXiv:2501.19306v4 Announce Type: replace Abstract: Recent advancements in Large Language Models (LLMs) have created new opportunities to enhance performance on complex reasoning tasks by leveraging test-time computation. However, existing scaling methods have key limitations: parallel methods like repeated sampling are often inefficient and quickly saturate, while sequential methods like SELF-REFINE struggle to improve after a few rounds. Although combining these approaches shows promise, current methods require fine-tuned reward and revision models. This paper proposes Self-Enhanced Test-Time Scaling (SETS), a simple yet effective approach that overcomes these limitations by strategically combining parallel and sequential techniques and fully leveraging LLMs' self-improvement abilities. SETS exploits the inherent self-verification and self-correction capabilities of LLMs, unifying sampling, verification, and correction within a single framework. This facilitates efficient and scalable test-time computation for enhanced performance on complex tasks without any model training. Our comprehensive experimental results on challenging benchmarks spanning planning, reasoning, math, and coding demonstrate that SETS achieves significant performance improvements and more advantageous test-time scaling behavior than the alternatives.

arXiv:2502.20396v2 Announce Type: replace-cross Abstract: Learning generalizable robot manipulation policies, especially for complex multi-fingered humanoids, remains a significant challenge. Existing approaches primarily rely on extensive data collection and imitation learning, which are expensive, labor-intensive, and difficult to scale. Sim-to-real reinforcement learning (RL) offers a promising alternative, but has mostly succeeded in simpler state-based or single-hand setups. How to effectively extend this to vision-based, contact-rich bimanual manipulation tasks remains an open question. In this paper, we introduce a practical sim-to-real RL recipe that trains a humanoid robot to perform three challenging dexterous manipulation tasks: grasp-and-reach, box lift and bimanual handover. Our method features an automated real-to-sim tuning module, a generalized reward formulation based on contact and object goals, a divide-and-conquer policy distillation framework, and a hybrid object representation strategy with modality-specific augmentation. We demonstrate high success rates on unseen objects and robust, adaptive policy behaviors -- highlighting that vision-based dexterous manipulation via sim-to-real RL is not only viable, but also scalable and broadly applicable to real-world humanoid manipulation tasks.

arXiv:2411.02708v2 Announce Type: replace-cross Abstract: Multimodal large language models (MLLMs) have recently achieved state-of-the-art performance on tasks ranging from visual question answering to video understanding. However, existing studies have concentrated mainly on visual-textual misalignment, leaving largely unexplored the MLLMs' ability to preserve an originally correct answer when confronted with misleading information. We reveal a response uncertainty phenomenon: across nine standard datasets, twelve state-of-the-art open-source MLLMs overturn a previously correct answer in 65% of cases after receiving a single deceptive cue. To systematically quantify this vulnerability, we propose a two-stage evaluation pipeline: (1) elicit each model's original response on unperturbed inputs; (2) inject explicit (false-answer hints) and implicit (contextual contradictions) misleading instructions, and compute the misleading rate - the fraction of correct-to-incorrect flips. Leveraging the most susceptible examples, we curate the Multimodal Uncertainty Benchmark (MUB), a collection of image-question pairs stratified into low, medium, and high difficulty based on how many of twelve state-of-the-art MLLMs they mislead. Extensive evaluation on twelve open-source and five closed-source models reveals a high uncertainty: average misleading rates exceed 86%, with explicit cues over 67.19% and implicit cues over 80.67%. To reduce the misleading rate, we then fine-tune all open-source MLLMs on a compact 2000-sample mixed-instruction dataset, reducing misleading rates to 6.97% (explicit) and 32.77% (implicit), boosting consistency by nearly 29.37% on highly deceptive inputs, and slightly improving accuracy on standard benchmarks. Our code is available at https://github.com/Yunkaidang/uncertainty

arXiv:2509.00102v1 Announce Type: new Abstract: Transformer-based foundation models for Electrocardiograms (ECGs) have recently achieved impressive performance in many downstream applications. However, the internal representations of such models across layers have not been fully understood and exploited. An important question arises: Does the final layer of the pre-trained Transformer model, the emph{de facto} representational layer, provide optimal performance for downstream tasks? Although our answer based on empirical and theoretical analyses for this question is negative, we propose a novel approach to leverage the representation diversity of the model's layers effectively. Specifically, we introduce a novel architecture called Post-pretraining Mixture-of-layers Aggregation (PMA), which enables a flexible combination of the layer-wise representations from the layer stack of a Transformer-based foundation model. We first pre-train the model from ECG signals using the 1-dimensional Vision Transformer (ViT) via masked modeling. In downstream applications, instead of relying solely on the last layer of the model, we employ a gating network to selectively fuse the representations from the pretrained model's layers, thereby enhancing representation power and improving performance of the downstream applications. In addition, we extend the proposed method to the pretraining stage by aggregating all representations through group-wise averaging before feeding them into the decoder-based Transformer.

arXiv:2406.15836v2 Announce Type: replace Abstract: Learning a world model for model-free Reinforcement Learning (RL) agents can significantly improve the sample efficiency by learning policies in imagination. However, building a world model for Multi-Agent RL (MARL) can be particularly challenging due to the scalability issue in a centralized architecture arising from a large number of agents, and also the non-stationarity issue in a decentralized architecture stemming from the inter-dependency among agents. To address both challenges, we propose a novel world model for MARL that learns decentralized local dynamics for scalability, combined with a centralized representation aggregation from all agents. We cast the dynamics learning as an auto-regressive sequence modeling problem over discrete tokens by leveraging the expressive Transformer architecture, in order to model complex local dynamics across different agents and provide accurate and consistent long-term imaginations. As the first pioneering Transformer-based world model for multi-agent systems, we introduce a Perceiver Transformer as an effective solution to enable centralized representation aggregation within this context. Results on Starcraft Multi-Agent Challenge (SMAC) show that it outperforms strong model-free approaches and existing model-based methods in both sample efficiency and overall performance.

arXiv:2509.00099v1 Announce Type: new Abstract: Quantum annealing offers a promising paradigm for solving NP-hard combinatorial optimization problems, but its practical application is severely hindered by two challenges: the complex, manual process of translating problem descriptions into the requisite Quadratic Unconstrained Binary Optimization (QUBO) format and the scalability limitations of current quantum hardware. To address these obstacles, we propose a novel end-to-end framework, LLM-QUBO, that automates this entire formulation-to-solution pipeline. Our system leverages a Large Language Model (LLM) to parse natural language, automatically generating a structured mathematical representation. To overcome hardware limitations, we integrate a hybrid quantum-classical Benders' decomposition method. This approach partitions the problem, compiling the combinatorial complex master problem into a compact QUBO format, while delegating linearly structured sub-problems to classical solvers. The correctness of the generated QUBO and the scalability of the hybrid approach are validated using classical solvers, establishing a robust performance baseline and demonstrating the framework's readiness for quantum hardware. Our primary contribution is a synergistic computing paradigm that bridges classical AI and quantum computing, addressing key challenges in the practical application of optimization problem. This automated workflow significantly reduces the barrier to entry, providing a viable pathway to transform quantum devices into accessible accelerators for large-scale, real-world optimization challenges.

arXiv:2509.02523v1 Announce Type: cross Abstract: We present the Flavors of Moonshine, a suite of tiny automatic speech recognition (ASR) models specialized for a range of underrepresented languages. Prevailing wisdom suggests that multilingual ASR models outperform monolingual counterparts by exploiting cross-lingual phonetic similarities. We challenge this assumption, showing that for sufficiently small models (27M parameters), training monolingual systems on a carefully balanced mix of high-quality human-labeled, pseudo-labeled, and synthetic data yields substantially superior performance. On average, our models achieve error rates 48% lower than the comparably sized Whisper Tiny model, outperform the 9x larger Whisper Small model, and in most cases match or outperform the 28x larger Whisper Medium model. These results advance the state of the art for models of this size, enabling accurate on-device ASR for languages that previously had limited support. We release Arabic, Chinese, Japanese, Korean, Ukrainian, and Vietnamese Moonshine models under a permissive open-source license.

arXiv:2509.00097v1 Announce Type: new Abstract: Neural network quantization aims to reduce the bit-widths of weights and activations, making it a critical technique for deploying deep neural networks on resource-constrained hardware. Most Quantization-Aware Training (QAT) methods rely on the Straight-Through Estimator (STE) to address the non-differentiability of discretization functions by replacing their derivatives with that of the identity function. While effective, STE overlooks discretization errors between continuous and quantized values, which can lead to accuracy degradation -- especially at extremely low bit-widths. In this paper, we propose Progressive Element-wise Gradient Estimation (PEGE), a simple yet effective alternative to STE, which can be seamlessly integrated with any forward propagation methods and improves the quantized model accuracy. PEGE progressively replaces full-precision weights and activations with their quantized counterparts via a novel logarithmic curriculum-driven mixed-precision replacement strategy. Then it formulates QAT as a co-optimization problem that simultaneously minimizes the task loss for prediction and the discretization error for quantization, providing a unified and generalizable framework. Extensive experiments on CIFAR-10 and ImageNet across various architectures (e.g., ResNet, VGG) demonstrate that PEGE consistently outperforms existing backpropagation methods and enables low-precision models to match or even outperform the accuracy of their full-precision counterparts.