arXiv:2503.20921v2 Announce Type: replace Abstract: We present and discuss a generalization of the popular MINI mixed finite element for the 2D Stokes equation by means of conforming virtual elements on polygonal meshes. We prove optimal error estimates for both velocity and pressure. Theoretical results are confirmed by several numerical tests performed with different choices of polynomial accuracy and meshes.
Science Journals
arXiv:2606.08654v1 Announce Type: new Abstract: In this paper, we propose a perturbation-based conformal prediction framework for uncertainty quantification in operator learning, with a focus on the 2D Navier--Stokes equations. While neural operators provide fast surrogates for expensive PDE solvers, they do not by themselves provide calibrated uncertainty for spatiotemporal field predictions. Our approach wraps a trained Fourier Neural Operator (FNO) with split conformal prediction and constructs the local uncertainty scale by comparing the predictions of two operators trained on nearly identical datasets: one on the original labels and one on labels perturbed by small Gaussian noise. We consider this procedure in the data-scarce regime, where the total label budget is fixed and methods that require a separate uncertainty network must divide training data between multiple models. On the 2D Navier--Stokes benchmark, the perturbation-based method produces substantially narrower conformal bands than existing methods under matched total data budgets while maintaining the target simultaneous coverage. These results suggest that perturbation sensitivity is a practical and sample-efficient uncertainty proxy for conformalized neural operators.
arXiv:2606.08656v1 Announce Type: new Abstract: Large language model (LLM) agents are increasingly deployed in long-running settings where improving through experience at test time becomes important. A common approach is to update an explicit memory after each interaction to guide future decisions. However, most existing methods rely on hand-designed prompting rules, making it difficult to align memory updates with downstream objectives over multi-step horizons consistently. We propose MemoPilot, a plug-in memory copilot that explicitly trains the memory update process to improve a frozen LLM's performance across sequential interactions. We formulate memory updating as a multi-turn decision problem and optimize it end-to-end with multi-turn GRPO. Our training recipe introduces (i) a turn-wise reward signal and (ii) a context-independent, turn-level advantage estimation across rollouts, enabling finer-grained credit assignment and more stable training in multi-turn settings. We evaluate MemoPilot on two testbeds: multi-round Rock-Paper-Scissors (RPS) and Limit Texas Hold'em (LHE). Across both environments, MemoPilot substantially improves test-time learning of a frozen player over strong baselines, ranking first in Elo ratings on both games (1762 on LHE and 1590 on RPS) and outperforming all baseline memory methods and proprietary models, including DeepSeek-V3.2.
arXiv:2501.15509v5 Announce Type: replace Abstract: Model fingerprinting has emerged as a crucial mechanism for safeguarding the intellectual property of open-source models, offering a non-intrusive approach that requires no modifications to the protected model. However, our analysis reveals that existing fingerprinting techniques are fundamentally vulnerable to false claim attacks, wherein adversaries can fraudulently assert ownership over independent third-party models. We demonstrate that this vulnerability stems from the untargeted nature of current methods, which evaluate model similarity based on arbitrary sample outputs rather than alignment with a specific, predefined reference. To mitigate this vulnerability, we introduce FIT-Print, a targeted fingerprinting paradigm that actively counters false claim attacks. Specifically, FIT-Print leverages optimization to transform the fingerprint into a verifiable, targeted signature. Building upon this foundation, we propose two black-box fingerprinting methods, the bit-wise FIT-ModelDiff and the list-wise FIT-LIME, which utilize output distances and feature attributions as robust model signatures, respectively. Extensive evaluations across benchmark models and datasets show that our framework perfectly neutralizes false claim attacks (100% defense success rate) and eliminates false alarms on independent models (0.0%), all while maintaining a 100% ownership verification rate against diverse model reuse techniques.
arXiv:2508.12536v3 Announce Type: replace Abstract: Providing \textit{structured} information to large language models (LLMs) improves multi-step reasoning and factual grounding, and recent retrieval-augmented generation (RAG) systems therefore reconstruct structure from retrieved text on every query. When the corpus is \emph{already} structured -- as in JSON Lines (JSONL), a popular format for LLM prompts, chemical compounds, and geospatial records -- this per-query rebuilding can be replaced by direct \emph{structural retrieval}. The core primitive is \textit{substructure search}: finding all JSON objects in a collection that contain a given query pattern. Existing approaches index each document separately, so both index space and query time grow with the total collection size; XML-based engines add conversion overhead and semantic mismatches. We propose \textbf{jXBW}, a compressed index for fast substructure search over JSONL, combining three innovations: (i) a merged tree representation that consolidates repeated structures across objects, (ii) a succinct tree index based on the eXtended Burrows--Wheeler Transform (XBW), and (iii) a newly developed three-phase substructure search algorithm that runs on this index. Together they achieve \textbf{query-dependent complexity}: the cost is determined by query characteristics rather than collection size, in compressed space. Experiments on seven real-world datasets, including PubChem ($10^6$ compounds) and OpenStreetMap ($6.6 \times 10^6$ objects), show that jXBW outperforms the strongest tree-based baseline by $\mathbf{16\times}$ on the smallest dataset and by up to $\mathbf{2{,}800\times}$ on the largest, and is more than $\mathbf{2 \times 10^6\times}$ faster than the XQuery engine Saxon. jXBW thus brings structural retrieval over million-record JSONL collections into the sub-millisecond range.
arXiv:2602.17307v2 Announce Type: replace Abstract: The Fischlin transform yields non-interactive zero-knowledge proofs with straight-line extractability in the classical random oracle model. This is done by forcing a prover to generate multiple accepting transcripts through a proof-of-work mechanism. Whether the Fischlin transform is straight-line extractable against quantum adversaries has remained open due to the difficulty of reasoning about the likelihood of query transcripts in the quantum-accessible random oracle model (QROM), even when using the compressed oracle methodology. In this work, we prove that the Fischlin transform remains straight-line extractable in the QROM, via an extractor based on the compressed oracle. This establishes the post-quantum security of the Fischlin transform, providing a post-quantum straight-line extractable NIZK alternative to Pass' transform with smaller proof size. Our techniques include tail bounds for sums of independent random variables and for martingales as well as symmetrization, query amplitude and quantum union bound arguments.
arXiv:2606.06915v2 Announce Type: replace Abstract: Test-time compute (TTC) scaling has emerged as a powerful paradigm for improving large language model (LLM) reasoning by allocating additional compute during inference, e.g., via multi-sample generation and verifier-based reranking. Existing TTC scaling strategies and reasoning scorers remain fragmented, evaluated under inconsistent protocols, and are rarely analyzed through the lens of quality-cost trade-offs. We introduce ThinkBooster, a unified framework for seamless test-time compute scaling of LLM reasoning, which consists of (i) a modular Python library implementing state-of-the-art TTC scaling strategy and scorer families, (ii) a benchmark that jointly evaluates performance and computational efficiency, and (iii) a deployable OpenAI-compatible proxy service that enables drop-in integration of adaptive reasoning into real-world applications. We further provide a demo visual debugger for inspecting the reasoning trajectories, intermediate selection decisions, and alternative reasoning paths. Empirical results on mathematical and coding tasks reveal the performance-compute trade-offs of TTC scaling strategies and scoring methods and demonstrate that ThinkBooster provides practical gains in real-world tasks. The code is available online under an MIT license.
arXiv:2511.19376v2 Announce Type: replace-cross Abstract: A Kokotsakis polyhedron is a polyhedral mesh in three-dimensional Euclidean space formed by a central n-gonal face (the base), n quadrilateral faces each sharing one edge with the base, and n triangular faces inserted between every two adjacent quadrilaterals; it is called flexible if it admits a continuous deformation that preserves the rigidity of every face. This work investigates flexible Kokotsakis polyhedra with a quadrangular base (n = 4) of equimodular elliptic type, filling a significant gap in the literature by providing the first explicit constructions of this type together with an explicit algebraic characterization in terms of flat and dihedral angles. A straightforwardly constructible class of polyhedra - called quasi-symmetric nets (QS-nets) - is introduced, characterized by a symmetry relation among flat angles. It is shown that every elliptic QS-net has equimodular elliptic type and is flexible in real three-dimensional Euclidean space (rather than only in complex configuration spaces), except for a few exceptional choices of dihedral angles, and that its flexion admits a closed-form parameterization. Examples are constructed that are non-self-intersecting and belong exclusively to the equimodular elliptic type. To support applications in computational geometry, a numerical pipeline is developed that searches for candidate solutions, verifies them using the explicit algebraic characterization, and constructs and visualizes the resulting polyhedra; numerical validations achieve high precision. Taken together, these results provide constructive criteria, algorithms, and validated examples for the equimodular elliptic type, enabling the design of a broad range of flexible Kokotsakis mechanisms.
arXiv:2606.08702v1 Announce Type: new Abstract: Recent advances have improved the adaptive capabilities of LLM-based multi-agent systems (MAS) through memory-, skill-, and learning-based approaches, yet these approaches remain challenged by noisy trajectories, insufficient modeling of memory-skill relations, and reliance on additional training or high-quality supervision. To address these limitations, we propose ConMem, a relation-aware and training-free framework that enables efficient multi-agent adaptation through cross-experience coordination. Specifically, ConMem distills historical interaction trajectories into structured memory cards to capture reusable strategies and cues, organizing them into a relation-aware memory graph. At runtime, ConMem retrieves cards according to task needs and coordinates them through the card graph to resolve strategy conflicts and recover their dependencies. Combined, these modules yield structured and relation-aware guidance, enabling robust, lightweight adaptation in multi-agent systems without additional training. Extensive experiments across multiple benchmarks and mainstream MAS architectures show consistent gains over existing memory architectures, with improved inference-time efficiency through pruning more than 50% of expanded candidates and reducing planning overhead by over 80%. Our codes are available at https://anonymous.4open.science/r/ConMemCode
arXiv:2606.08718v1 Announce Type: new Abstract: While Deep Active Learning (DAL) effectively reduces human annotation costs, its efficacy is constrained by human annotation errors. This is because the data sampled for active learning is assumed to be highly informative for training. When human annotators introduce errors into this informative data at a certain rate, the active learning performance drops significantly and, in some cases, even exhibits worse outcomes than passive learning. In this paper, we first analyze the impact of human annotation errors in the DAL setting. Then we propose a framework to address the human annotation noise problem for DAL. Informed by human learning patterns, the core idea of our proposed solution involves allocating a portion of the human annotation budget to re-annotate data that has already been labeled. Previous theoretical work suggests that when the model possesses a certain level of ability to identify potentially noisy data, even re-labeling a small fraction of the data can effectively remove noise from the active training set. To achieve this, we implement two active noise sampling strategies to detect noise under different circumstances and allocate a part of the annotation budget to re-annotate these instances. Our approach imbues active learning with a revisiting and introspective behavior. Our experiments demonstrate that, under the same annotation budget, our method is more data-efficient and yields a relatively noise-free annotation dataset in the end.
arXiv:2606.08729v1 Announce Type: new Abstract: Simulation plays a key role in automated robotics research supported by large language models (LLMs). However, existing simulators often require custom code or complex interfaces, creating a barrier to rapid prototyping and automated algorithm development. To this end, we propose the Intelligent Robot Simulator (IR-SIM), a lightweight skill-native navigation simulator designed for rapid scenario construction, benchmarking, and robot learning. In IR-SIM, scenarios are entirely defined by YAML configuration files that specify mobile robot kinematics, geometric collision checking, LiDAR sensing, visualization, and behavior modules. This design makes robotic simulation fully describable and reproducible, allowing scenarios to be generated and modified from text prompts through the proposed IR-SIM agent skills. The resulting scenarios can be used for automated benchmarking of navigation algorithms and for automated generation of training data for learning methods. Furthermore, IR-SIM provides bridges to high fidelity simulators and real world deployment, allowing users to validate their algorithms in more realistic settings after prototyping without extra coding. The experiments showcase the convenience and versatility of IR-SIM in multiple tasks: constructing navigation scenarios from natural language, training a collision avoidance policy, benchmarking social navigation policies, and bridging to high fidelity simulators and real world deployment. The project website is available at https://github.com/hanruihua/ir-sim.
arXiv:2606.08735v1 Announce Type: new Abstract: Quality-diversity reinforcement learning (QD-RL) aims to construct policy repertoires that contain both high-performing and behaviorally diverse policies. Existing QD-RL methods mainly diversify policy instances after rollout evaluation or use learned value information to improve policy quality and behavior targeting, while the learning branches that generate candidate policies remain less explored. This paper proposes SV-QD-RL, a structure-value coupled framework that represents each candidate as a structure-conditioned actor-critic branch. Each branch contains an actor, a structural mask, a branch-specific critic, a replay state, and evaluation attributes including behavior, return, sparsity, and value profile. The structural mask defines the actor subspace in which the branch learns, while the branch-specific critic and replay state shape its value-learning trajectory. A branch-aware QD archive then evaluates and retains branches according to behavioral quality, structural footprint, and value-profile information. Experiments on MuJoCo continuous-control tasks show that SV-QD-RL constructs policy repertoires with strong archive quality and behaviorally useful diversity. Ablation and diagnostic analyses further indicate that structural conditioning, critic differentiation, and memory-consistent refinement make complementary contributions to behavioral specialization. Schedule-aware repertoire evaluation shows that the learned archive provides selectable policy alternatives under changing behavior-level requirements. These results suggest that coupling actor structure with branch-specific value learning is an effective mechanism for generating diverse QD-RL policy repertoires.
arXiv:2605.15491v2 Announce Type: replace Abstract: Layer pruning removes entire Transformer decoder blocks from large language models, but introduces a mismatch between the hidden state received by the next surviving layer and the distribution it was trained to process, leading to significant performance degradation. We propose Ghosted Layers, a training-free recovery module that addresses this issue by solving a boundary activation alignment problem. Our method derives a closed-form optimal linear operator from a small calibration set to reconstruct the activation discrepancy introduced by the pruned layers. We show that this solution corresponds to the unconstrained optimum of the alignment objective, whereas existing methods are restricted to constrained solutions over limited operator subspaces. Experiments across multiple LLM backbones and pruning strategies demonstrate that our method consistently improves accuracy and perplexity over prior training-free baselines, while preserving the efficiency gains of layer pruning. Official code repository: https://github.com/daniel-eai/ghosted_layers_official_repository/.
arXiv:2406.19749v2 Announce Type: replace-cross Abstract: Automatic vessel segmentation plays a pivotal role in the development of next-generation interventional navigation systems for surgical robotics. However, current approaches still suffer from suboptimal segmentation performance under challenging intraoperative conditions, such as low-signal-to-noise ratio (SNR), small or slender vessels, and strong interference. In this study, a novel spatial-frequency learning and graph-based channel interaction network (SPIRONet) is proposed to address the above issues. To address low-SNR vessel appearance and small or slender branches, dual spatial-frequency encoders are utilized, where the frequency encoder captures global vessel continuity that is less affected by local noise fluctuations, while the spatial encoder preserves fine vessel details. A cross-attention fusion module is further introduced to adaptively integrate this complementary spatial and frequency information. Moreover, to suppress interference from non-target vessels and vessel-like structures, a graph-based channel interaction module is designed to model channel-wise correlations, enhancing consistent vessel-related responses while suppressing task-irrelevant activations. Extensive experimental results on five challenging datasets demonstrate that the proposed method achieves competitive and consistently strong performance compared with existing methods. For example, SPIRONet achieves IoU improvements of +0.87%, +0.52%, +0.23%, +1.39%, and +2.22% over the strongest competing methods on CADSA, CAXF, DCA1, XCAD, and ARCADE, respectively. Moreover, SPIRONet achieves an inference speed of 21 FPS with a 512x512 input size, meeting the real-time requirements of interventional scenarios (6-12 FPS). These promising results indicate SPIRONet's potential for integration into interventional navigation systems. Code is available at https://github.com/Dxhuang-CASIA/SPIRONet.
arXiv:2503.22697v3 Announce Type: replace-cross Abstract: Decoding sensory experiences from neural activity to reconstruct human-perceived visual stimuli and semantic content remains a challenge in neuroscience and artificial intelligence. Despite notable progress in current brain decoding models, a critical gap still persists in their systematic integration with established neuroscientific theories and the exploration of underlying neural mechanisms. Here, we present a novel framework that directly decodes fMRI signals into textual descriptions of viewed natural images. Our novel deep learning model, trained without visual information, achieves state-of-the-art semantic decoding performance, generating meaningful captions that capture the core semantic content of complex scenes. Neuroanatomical analysis reveals the critical role of higher-level visual cortices, including MT+ complex, ventral stream visual cortex, and inferior parietal cortex, in visual semantic processing. Furthermore, category-specific analysis demonstrates nuanced neural representations for semantic dimensions like animacy and motion. This work provides a more direct and interpretable framework to the brain's semantic decoding, offering a powerful new methodology for probing the neural basis of complex semantic processing, refining the understanding of the distributed semantic network, and potentially developing brain-inspired language models.
arXiv:2502.15917v3 Announce Type: replace-cross Abstract: Stochastic Unit Commitment (SUC) has been proposed to manage the uncertainties driven by renewable integration, but it leads to significant computational complexity. When accelerated by Benders Decomposition (BD), the master problem becomes binary integer programming, which is still NP-hard and computationally demanding for classical methods. Quantum Annealing (QA), known for efficiently solving Quadratic Unconstrained Binary Optimization (QUBO) problems, presents a potential solution. However, existing quantum algorithms rely on slack variables to handle linear binary inequality constraints, leading to increased qubit consumption and reduced computational efficiency. To solve the problem, this paper introduces the Powell-Hestenes-Rockafellar Augmented Lagrangian Multiplier (PHR-ALM) method to eliminate the need for slack variables, making qubit consumption independent of the increasing number of Benders cuts. To further reduce the qubit overhead, quantum ADMM is applied to break large-scale SUC into smaller blocks for sequential solutions, which does not scale with the number of generators. Finally, the simulation results on both 4-generator and the IEEE bus-118 systems demonstrate the feasibility and scalability of the proposed algorithm, indicating its superior qubit and runtime efficiency over classical and baseline quantum approaches on the D-Wave QPU platform.
arXiv:2506.19094v5 Announce Type: replace-cross Abstract: Neural recording technologies now enable simultaneous recording of population activity across many brain regions, motivating the development of data-driven models of inter-regional communication. However, existing models can struggle to disentangle the influences that drive recorded population activity, leading to inaccurate portraits of communication. Here, we introduce Multi-Region Latent Factor Analysis via Dynamical Systems (MR-LFADS), a sequential variational autoencoder designed to disentangle inter-regional communication, inputs from unobserved regions, and local neural population dynamics. We show that MR-LFADS outperforms existing approaches at identifying communication across dozens of simulations of task-trained multi-region networks. When applied to large-scale electrophysiology, MR-LFADS predicts brain-wide effects of circuit perturbations that were held out during model fitting. These validations on synthetic and real neural data position MR-LFADS as a promising tool for discovering principles of brain-wide information processing.
arXiv:2606.00958v2 Announce Type: replace-cross Abstract: The square-and-multiply algorithm, also known as binary exponentiation or repeated squaring, is a technique for fast exponentiation commonly used in modern cryptography and computational number theory. Despite its prominence, the historical origins of the algorithm are not known with certainty. This paper critically examines the origins and formalization of the algorithm through primary source analysis. We focus on Jamshid al-Kashi's fifteenth-century Miftah al-Hisab where the algorithm is articulated explicitly as a general method and claimed by al-Kashi as his own innovation. To contextualize this, we trace earlier instances of successive squaring in the works of al-Uqlidisi and al-Biruni, who applied these techniques for specific calculations, but did not formalize them into a general procedure. The earliest known work on this method of computation is found in Pingala's prosodic studies in ancient India (c. 200 BCE). Even though it was not fully developed as a general technique, Pingala's work seems to contain the conceptual foundation of the algorithm which is to employ the binary representation of a positive integer. By mapping this intellectual progression, the paper illustrates the historical background of an algorithm that is prominent in modern computation.
arXiv:2605.16928v2 Announce Type: replace Abstract: Long-context inference in large language models is bottlenecked by the quadratic cost of full attention. Existing efficient alternatives often rely either on native sparse training or on heuristic token eviction, creating an undesirable trade-off among efficiency, training cost, and accuracy. In this work, we show that full-attention LLMs are already intrinsically sparse and can be transformed into highly sparse models with only minimal adaptation. Our approach is built on three observations: (1) only a small subset of attention heads truly requires full long-context processing; (2) long-range retrieval is governed primarily by a low-dimensional subspace, allowing relevant tokens to be retrieved efficiently with a 16-dimensional indexer; and (3) the useful token budget is strongly query-dependent, making dynamic top-$p$ selection more suitable than fixed top-$k$ sparsification. Based on these insights, we propose RTPurbo, which retains the full KV cache only for retrieval heads and introduces a lightweight token indexer for sparse attention. By exploiting the model's intrinsic sparsity, RTPurbo achieves sparsification with only a few hundred training steps. Experiments on long-context benchmarks and reasoning tasks show that RTPurbo preserves near-lossless accuracy while delivering substantial efficiency gains, including up to a 9.36$\times$ prefill speedup at 1M context and about a 2.01$\times$ decode speedup. These results suggest that strong sparse inference can be obtained from standard full-attention training without expensive native sparse pretraining.
arXiv:2605.17289v2 Announce Type: replace Abstract: Unstructured sparsity is now natively accelerated by recent GPU kernels and dataflow hardware, shifting the bottleneck from inference execution to the pruning algorithm. State-of-the-art methods for unstructured LLM pruning are layer-wise surrogates derived from the Optimal Brain Surgeon principle, and they sacrifice end-to-end accuracy, especially under aggressive sparsity. End-to-end alternatives such as MaskLLM and PATCH show that learnable masks can close this gap, but their categorical-over-patterns parameterization scales with the number of valid masks per row and does not port to the unstructured setting. We introduce LEAP, which replaces this intractable parameterization with a per-weight Bernoulli-via-Gumbel-sigmoid relaxation that makes end-to-end unstructured mask learning tractable. Across five LLM families from 0.5B to 8B parameters at 50% and 60% sparsity, LEAP improves six-task average zero-shot accuracy by +2.59 points on average over ADMM, the best layer-wise baseline in our sweep.
arXiv:2603.26763v2 Announce Type: replace Abstract: Talking-head videos constitute a predominant content type in real-time communication, yet publicly available datasets for video processing research in this domain remain scarce and limited in signal fidelity. In this paper, we open-source a camera-native dataset of 847 talking-head recordings (approximately 212 minutes), each 15s in duration, captured from 805 participants using 446 unique consumer webcam devices in their natural environments. All recordings are stored using the FFV1 lossless codec, preserving the camera-native signal -- uncompressed (24.4%) or MJPEG-encoded (75.6%) -- without additional lossy processing. Each recording is annotated with a Mean Opinion Score (MOS) and ten perceptual quality tokens that jointly explain 64.4% of the MOS variance. From this corpus, we curate a stratified benchmarking subset of 120 clips in three content conditions: original, background blur, and background replacement. Codec efficiency evaluation across four datasets and four codecs, namely H.264, H.265, H.266, and AV1, yields VMAF BD-rate savings up to $-71.3\%$ (H.266) relative to H.264, with significant encoder$\times$dataset ($\eta_p^2 = .112$) and encoder$\times$content condition ($\eta_p^2 = .149$) interactions, demonstrating that both content type and background processing affect compression efficiency. A preliminary super-resolution evaluation with four SR models confirms that the dataset significantly affects absolute performance while preserving model rankings, demonstrating applicability beyond codec benchmarking. The dataset offers 5$\times$ the scale of the largest prior talking-head webcam dataset (847 vs. 160 clips) with lossless signal fidelity, establishing a resource for benchmarking video compression, super-resolution, quality assessment, and enhancement models in real-time communication.
arXiv:2603.27493v2 Announce Type: replace Abstract: Spiking Neural Networks (SNNs), characterized by their event-driven computation and low power consumption, have shown great potential for energy-efficient visual tracking on unmanned aerial vehicles (UAVs). However, existing SNN-based trackers often rely on costly event cameras, which limits their deployment on standard RGB-camera UAV platforms. To address this limitation, we propose STATrack, a fully spiking neural network framework for UAV visual tracking using only RGB inputs. To the best of our knowledge, this is the first study to explore fully spiking neural networks for RGB-based UAV visual tracking. To alleviate target semantic degradation caused by spike discretization and reduce background interference in UAV scenes, we introduce an Adaptive Mutual Information Maximization (AMIM) mechanism. AMIM maximizes the mutual information between template inputs and their deep target-aware features, encouraging the spiking backbone to preserve discriminative target semantics. In addition, a sample-difficulty-aware dynamic weighting strategy is designed to adaptively adjust the mutual information constraint during training. Extensive experiments on four widely used UAV tracking benchmarks demonstrate that STATrack achieves state-of-the-art tracking performance with low theoretical energy consumption, highlighting its potential for energy-constrained UAV applications.
arXiv:2510.04593v3 Announce Type: replace-cross Abstract: Large language models (LLMs) have demonstrated promising performance in both automatic speech recognition (ASR) and text-to-speech (TTS) systems, gradually becoming the mainstream approach. However, most current approaches address these tasks separately rather than through a unified framework. This work aims to integrate these two tasks into one unified model. Although discrete speech tokenization enables joint modeling, its inherent information loss limits performance in both recognition and generation. In this work, we present UniVoice, a unified LLM framework through continuous representations that seamlessly integrates speech recognition and synthesis within a single model. Our approach combines the strengths of autoregressive modeling for speech recognition with flow matching for high-quality generation. To mitigate the inherent divergence between autoregressive and flow-matching models, we further design a dual attention mechanism, which switches between a causal mask for recognition and a bidirectional attention mask for synthesis. Furthermore, the proposed text-prefix-conditioned speech infilling method enables high-fidelity zero-shot voice cloning. Experimental results demonstrate that our method can achieve or exceed current single-task modeling methods in both ASR and zero-shot TTS tasks. This work explores new possibilities for end-to-end speech understanding and generation. Code is available at https://github.com/gwh22/UniVoice.
arXiv:2602.00797v3 Announce Type: replace-cross Abstract: Flow-based methods have achieved significant success in various generative modeling tasks, capturing nuanced details within complex data distributions. However, few existing works have exploited this unique capability to resolve fine-grained structural details beyond generation tasks. This paper presents a flow-inspired framework for representation learning. First, we demonstrate that a rectified flow trained using independent coupling is zero everywhere at $t=0.5$ if and only if the source and target distributions are identical. We term this property the \emph{zero-flow criterion}. Second, we show that this criterion can certify conditional independence, thereby extracting \emph{sufficient information} from the data. Third, we translate this criterion into a tractable, simulation-free loss function that enables learning amortized Markov blankets in graphical models and latent representations in self-supervised learning tasks. Experiments on both simulated and real-world datasets demonstrate the effectiveness of our approach. The code reproducing our experiments can be found at: https://github.com/probabilityFLOW/zfe.
arXiv:2509.10534v3 Announce Type: replace Abstract: The attention mechanism in a Transformer architecture matches key to query based on both content -- the what -- and position in a sequence -- the where. We present an analysis indicating that what and where are entangled in the popular RoPE rotary position embedding. This entanglement can impair performance particularly when decisions require independent matches on these two factors. We propose an improvement to RoPE, which we call Polar Coordinate Position Embeddings or PoPE, that eliminates the what-where confound. PoPE is far superior on a diagnostic task requiring indexing solely by position or by content. On autoregressive sequence modeling in music, genomic, and natural language domains, Transformers using PoPE as the positional encoding scheme outperform baselines using RoPE with respect to evaluation loss (perplexity) and downstream task performance. On language modeling, these gains persist across model scale, from 124M to 774M parameters. Crucially, PoPE shows strong zero-shot length extrapolation capabilities compared not only to RoPE but even a method designed for extrapolation, YaRN, which requires additional fine tuning and frequency interpolation.