论文检索

Applications · Language, Speech and Dialog

Artem Dementyev, Wazeer Zulfikar, Sinan Hersek, Pascal Getreuer, Anurag Kumar, Vivek Kumar

Current multimodal LLMs process audio as a mono stream, ignoring the rich spatial information essential for embodied AI. Existing spatial audio models, conversely, are constrained to fixed microphone geometries, preventing deployment across diverse devices. We present PhaseCoder, a transformer-only spatial audio encoder that is agnostic to microphone geometry. PhaseCoder takes raw multichannel audio and microphone coordinates as inputs to perform localization and produces robust spatial embeddings. We integrate PhaseCoder with the Gemma 3n LLM by finetuning it to reason over ``spatial audio tokens''. We show our encoder achieves state-of-the-art results on microphone-invariant localization benchmarks and, for the first time, enables an LLM to perform complex spatial reasoning and targeted transcription tasks from an arbitrary microphone array.

Theory · Probabilistic Methods

William He, Nicholas Kocurek, William Gay, Ryan O'Donnell

We study the problems of approximate sampling from and distribution testing of \emph{mixture models}, where the modes satisfy a functional inequality called \textit{approximate tensorization of entropy} (ATE). While it is known that ATE makes these tasks more efficient in the unimodal setting, mixtures of few distributions satisfying ATE do not necessarily satisfy ATE overall, leading to a lack of theoretical guarantees for multimodal distributions, which are a key challenging case of modern generative models. We show this gap can be overcome by establishing the following pair of results for mixtures of ATE distributions: 1) We show fast mixing of Glauber dynamics from a \textit{data-based initialization}, with \textit{optimal} sample complexity, for mixtures of distributions satisfying modified log-Sobolev inequalities, building on similar results in \cite{KoehlerV24, HuangMRW24} for mixtures satisfying the weaker Poincaré inequality. 2) Answering an open question from \cite{blanca2023complexity}, we give efficient identity-testers for mixtures of ATE distributions in the coordinate-conditional sampling access model.

Theory · Learning Theory

Guillaume Larue, Louis-Adrien Dufrène, Quentin Lampin, Hadi Ghauch, Ghaya Rekaya-Ben Othman

Parity functions are fundamental Boolean operations with critical applications across machine learning, cryptography, and error correction. Yet, learning high-dimensional parity functions poses significant challenges: in a general setting, standard neural network architectures typically require exponential sample complexity, making gradient-based optimization intractable for large number of inputs $N$. We demonstrate that compact product-based neural architectures combined with stochastic data sparsity (Bernoulli inputs with $p_e \leq 1/N$) and appropriate hyperparameter choice enable efficient parity learning, with theoretical guarantees of convergence. Experiments validate our theory across dimensions up to $N = 10{,}000$, with empirical evidence showing optimal hyperparameter choices for $p_e$ and learning rate $\alpha$, as well as polynomial complexity scaling laws. This work establishes fundamental connections between architectural inductive bias and data sparsity, opening new possibilities for neural arithmetic, structured reasoning, binary neural networks, and machine learning applied to automated protocol discovery.

Deep Learning · Other Representation Learning

Alexey Kravets, Da Li, Chuan Li, Da Chen, Vinay Namboodiri

Recent advances in language model interpretability using sparse autoencoders (SAEs) have yet to effectively translate to the visual domain, mainly due to the difficulty and ambiguity of labeling visual concepts. In this paper, we introduce Visual Interpretability via SAE Transfer Alignment (VISTA), a framework that transfers interpretability from language to vision by constraining a visual projector to map visual tokens into an LLM's pre-existing, labeled textual SAE space. This approach enables visual interpretability without training dedicated vision SAEs. By regularizing the projector using the LLM's SAE reconstruction loss, VISTA achieves a \textbf{threefold} increase in the matching rate which measures how accurately the most activating textual concepts in the SAE space correspond to semantic elements in the image. Using this framework, we further analyze spatial localization properties of different vision encoders and show that DINOv2 features have significantly stronger localization abilities than other encoders. Leveraging this precision, VISTA enables fine-grained, localized concept steering, allowing specific objects to be removed or replaced while preserving the surrounding scene. This results in improvements of \textbf{38\%} in object removal and \textbf{58\%} in object replacement tasks over vision-only baselines. These contributions are validated across multiple LLM architectures.

Social Aspects · Privacy

Peiru Yang, Yi Luo, Zhenfeng Gao, Tong Ju, Haoran Zheng, Linjie Zhu, Hongke Fu, Qing Li, Shangguang Wang, Tao Qi

Retrieval-Augmented Generation (RAG) systems are increasingly deployed to provide query-based access to large knowledge bases, thereby introducing concrete privacy risks whereby the underlying corpus may be partially or fully extracted through the deployed service. Existing extraction attacks typically rely on locally driven search strategies, in which newly extracted content is inferred or expanded based on previously recovered fragments. However, real-world knowledge bases are often multi-source and heterogeneous, with pronounced semantic discontinuities across domains. Such gaps can trap extraction methods that rely on local semantic continuity in local optima, severely limiting large-scale corpus reconstruction. In this paper, we introduce an extraction framework (GeoEx) designed to navigate and reconstruct heterogeneous RAG knowledge bases without any prior knowledge. The framework plans extraction directly in the embedding space of a proxy retrieval model to improve global coverage, and employs an embedding inversion module to convert latent vectors into executable queries. We further propose a composite geometric strategy that combines orthogonal query synthesis for cross-domain exploration with local embedding perturbations for dense extraction within discovered clusters. Experiments on mixed corpora spanning eight diverse domains and multiple retrievers and LLMs show that GeoEx significantly outperforms baselines in both extraction coverage and query efficiency.

Deep Learning · Algorithms

Arthur Lebeurrier, Titouan Vayer, Rémi Gribonval

Despite recent algorithmic advances, we still lack principled ways to leverage the well-documented rescaling symmetries in ReLU neural network parameters. While two properly rescaled weights implement the same function, the training dynamics can be dramatically different. To offer a fresh perspective on exploiting this phenomenon, we build on the recent path-lifting framework, which provides a compact factorization of ReLU networks. We introduce a geometrically motivated criterion to rescale neural network parameters which minimization leads to a conditioning strategy that aligns a kernel in the path-lifting space with a chosen reference. We derive an efficient algorithm to perform this alignment. In the context of random network initialization, we analyze how the architecture and the initialization scale jointly impact the output of the proposed method. Numerical experiments illustrate its potential to speed up training.

Deep Learning · Large Language Models

Zhiqi Bu

Model depth is a double-edged sword in deep learning: deeper models achieve higher accuracy but require higher computational cost. To efficiently train models at scale, progressive training -- an effective strategy where model capacity scales up during training, has emerged to significantly reduce computation with little to none performance degradation. In this work, we study the depth expansion of large-scale models through the lens of optimization theory and feature learning, offering insights on the initialization of new layers, hyperparameter transfer, learning rate schedule, and timing of model expansion. Specifically, we propose zero/one-layer progressive training for the optimal tradeoff between computation and loss. For example, zero/one-layer progressive training on GPT2 can save $\approx 80\%$ compute, or equivalently accelerate by $\approx 5\times$, and achieve a loss comparable to a fully trained 60-layer model with 7B parameters.

Applications · Computer Vision

Prashant Pandey, Himanshu Kumar, Devineni Chowdary, Brejesh Lall

Evolving data streams induce joint nonstationarity in continual semantic segmentation, where semantic classes, input distributions, and supervision availability change simultaneously over time. This setting reflects practical structured prediction systems, yet remains largely unexplored in prior continual learning work, which typically studies these factors in isolation. We formalize continual segmentation under coupled class, domain, and label shifts and investigate learning in heterogeneous dense prediction environments with limited annotations and abundant unlabeled data. To address instability and overfitting arising from few-shot supervision under distribution drift, we introduce gradient-adaptive stabilization, a parameter-wise regularization mechanism implemented via gradient-scaled stochastic perturbations that promotes a principled stability–plasticity tradeoff. We further leverage unlabeled data through semi-supervised learning and introduce prototype anchored supervision that validates pseudo-labels via joint confidence and prototype consistency. Together, these mechanisms enable learning under joint nonstationarity in continual segmentation. Extensive empirical evaluation across class-incremental, domain-incremental, and few-shot regimes demonstrates consistent improvements over prior methods in heterogeneous structured prediction settings. Our results expose fundamental failure modes of existing continual segmentation approaches and provide insight into learning robust dense predictors in dynamically evolving environments.

Applications · Neuroscience, Cognitive Science

Shuqi Zhu, Yi Zhong, Ziyi Ye, Bangde Du, Yujia Zhou, Qingyao Ai, Yiqun LIU

While AI-generated hallucinations pose considerable risks, the underlying cognitive mechanisms by which humans can successfully recognize or be misled by these hallucinations remain unclear. To address this problem, this paper explores humans' neural dynamics to characterize how the brain processes hallucinated content. We record EEG signals from 27 participants while they are performing a verification task to judge the correctness of image descriptions generated by a multi-modal large language model (MLLM). Based on an averaged event-related potential (ERP) study, we reveal that multiple cognitive processes, e.g., semantic integration, inferential processing, memory retrieval, and cognitive load, exhibit distinct patterns when humans process hallucinated versus non-hallucinated content. Notably, neural responses to hallucinations that were misjudged versus correctly judged by human participants showed significant differences. This indicates that misjudged AI-generated hallucinations failed to trigger the standard neurocognitive fact verification pathway. The detailed code can be accessed openly through https://anonymous.4open.science/r/EEG-Hallucination-anonymous-F125.

Jingyang Li, Jian-Feng Cai, Yang Chen, Dong Xia

Large tensor learning algorithms are typically computationally expensive and require storing a vast amount of data. In this paper, we propose a unified online Riemannian gradient descent (oRGrad) algorithm for tensor learning, which is computationally efficient, consumes much less memory, and can handle sequentially arriving data while making timely predictions. The algorithm is applicable to both linear and generalized linear models. If the time horizon T is known, oRGrad achieves statistical optimality by choosing an appropriate fixed step size. We find that noisy tensor completion particularly benefits from online algorithms by avoiding the trimming procedure and ensuring sharp entry-wise statistical error, which is often technically challenging for offline methods. The regret of oRGrad is analyzed, revealing a fascinating trilemma concerning the computational convergence rate, statistical error, and regret bound. By selecting an appropriate constant step size, oRGrad achieves an O(T1/2) regret. We then introduce the adaptive-oRGrad algorithm, which can achieve the optimal O(logT) regret by adaptively selecting step sizes, regardless of whether the time horizon is known. The adaptive-oRGrad algorithm can attain a statistically optimal error rate without knowing the horizon. Comprehensive numerical simulations corroborate our theoretical findings. We show that oRGrad significantly outperforms its offline counterpart in predicting the solar F10.7 index with tensor predictors that monitor space weather impacts.

Deep Learning · Graph Neural Networks

Isaac Reid, Arijit Sehanobish, Cederik Höfs, Bruno Mlodozeniec, Leonhard Vulpius, Federico Barbero, Adrian Weller, Krzysztof Choromanski, Richard E Turner, Petar Veličković

We study the extent to which rotary position encodings (RoPE), a recent transformer position encoding algorithm broadly adopted in large language models (LLMs) and vision transformers (ViTs), can be applied to graph-structured data. We find that rotating tokens depending on the spectrum of the graph Laplacian efficiently injects structural information into the attention mechanism, boosting performance in synthetic and real-world graph learning tasks. This approach, coined _Wave-Induced Rotary Encodings_ (WIRE), enjoys intriguing theoretical properties: it recovers regular RoPE on grids, and depends asymptotically on the graph effective resistance. Unlike bias-based relative position encodings, WIRE is compatible with linear attention.

Applications · Neuroscience, Cognitive Science

Wu S. Zihan, Ariane Delrocq, Wulfram Gerstner, Guillaume Bellec

While end-to-end self-supervised learning with backpropagation (global BP-SSL) has become central for training modern AI systems, theories of local self-supervised learning (local-SSL) have struggled to build functional representations in deep neural networks. To establish a link between global and local rules, we first develop a theory for deep linear networks: We identify conditions for local-SSL algorithms (like Forward-forward or CLAPP) to implement exactly the same weight update as a global BP-SSL. Starting from the theoretical insights, we then develop novel variants of local-SSL algorithms to approximate global BP-SSL in deep non-linear convolutional neural networks. Variants that improve the similarity between updates of local-SSL with those of global BP-SSL also show better performance on image datasets (CIFAR-10, STL-10, and Tiny ImageNet). The best local-SSL rule with the CLAPP loss function matches the performance of a comparable global BP-SSL with InfoNCE or CPC-like loss functions, and improves upon state-of-the-art for layer-wise SSL on these benchmarks.

Applications · Language, Speech and Dialog

Bin Kang, Shaoguo Wen, Yang Fan, Shunlong Wu, Junjie Wang, Yulin Li, Junzhi Zhao, Junle Wang, Zhuotao Tian

While existing text-to-speech (TTS) models exhibit high expressiveness, fine-grained control over composite instructions remains challenging due to the structural mismatch between discrete textual intents and continuous acoustic realizations. Inspired by human cognitive decoupling, we introduce AgentSteerTTS, a multi-agent closed-loop framework designed for intent-faithful expressive control of composite instructions. First, in our framework, an adversarial disentanglement module mitigates speaker-emotion leakage through gradient reversal and cross-covariance regularization. Next, a Dual-Stream Anchoring Controller grounds abstract intents using a large-scale acoustic prototype library: a Retrieval Agent selects expressive anchors, while a Synthesis Agent fuses them into continuous control vectors via gated attention. Finally, a Fast–Slow Feedback Agent refines output intensity through latent gradient correction and resolves semantic–acoustic mismatches using high-level perceptual critique. Experiments on a composite-instruction benchmark and public test sets show that AgentSteerTTS yields consistent and significant improvements to the baselines, demonstrating the effectiveness of the proposed method. Our code and models will be released.

Deep Learning · Large Language Models

Meimingwei Li, Yuanhao Ding, Esteban Garces Arias, Christian Heumann

Recent work has identified a counterintuitive phenomenon termed “Hyperfitting", where fine-tuning Large Language Models (LLMs) to near-zero training loss on small datasets surprisingly enhances open-ended generation quality and mitigates repetition in greedy decoding. While effective, the underlying mechanism remains poorly understood, with the extremely low-entropy output distributions suggesting a potential equivalence to simple temperature scaling. In this work, we demonstrate that this phenomenon is fundamentally distinct from distribution sharpening; entropy-matched control experiments reveal that temperature scaling fails to replicate the diversity gains of hyperfitting. Furthermore, we falsify the hypothesis of static vocabulary reweighting, showing through ablation studies that hyperfitting relies on a dynamic, context-dependent rank reordering mechanism. Layer-wise analysis localizes this effect to a “Terminal Expansion" in the final transformer block, where a substantial geometric expansion of the feature space ($\Delta \mathrm{Dim} \approx +80.8$) facilitates the promotion of deep-tail tokens. And we introduce \textbf{Late-Stage LoRA}, a targeted fine-tuning strategy that updates only the final 5 layers, achieving robust generation with minimal parameter updates.

Deep Learning · Large Language Models

Zishan Shao, Lixun Zhang, Kangning Cui, Yixiao Wang, Ting Jiang, Hancheng Ye, Qinsi Wang, Zhixu Du, Yuzhe Fu, Fan Yang 等

Large language models (LLMs) handle many tasks with one set of parameters, but under KV-cached inference it is unclear what task-general structure, if any, is used at $\textit{decode time}$ rather than during $\textit{prefill}$. We propose $\textbf{DecodeShare}$, a protocol that identifies a low-dimensional subspace that is consistently shared across tasks in decode-time hidden states, and then tests its causal role by removing that subspace only during decoding. In our experiments, disturbing the discovered shared subspace degrades decision performance far more than disturbing either a prefill-derived subspace or a random subspace under the same intervention budget. We further find that this decode-shared subspace overlaps common steering vectors, enabling a simple offline adjustment: projecting steering vectors away from the shared subspace can reduce template sensitivity while preserving non-random task utility, with task-dependent trade-offs. Despite being compact, the shared subspace can serve as a high-leverage causal channel at decode time.

General Machine Learning · Evaluation

Michael Hardy, Sang Truong, Anka Reuel, Lijin Zhang, Jodi Casabianca, Yash Dave, Hansol Lee, Ben Domingue, Sanmi Koyejo

AI benchmark ecosystems compress rich evaluation data into aggregate leaderboard scores, but these scores contain substantial measurement noise whose sources and magnitudes remain unquantified. Without systematic methods to measure this noise and separate signal from artifact, it is unclear when benchmark rankings reflect genuine capability differences versus measurement error. We introduce a psychometric framework to methodically test hypotheses about benchmark ecosystem structure and quantify the reliability of common benchmark-derived claims. Applying Confirmatory Factor Analysis and Generalizability Theory to 4,000+ models from the Open LLM Leaderboard, we decompose sources of ranking variance and find that human contributors account for more variance (9\%) than model architecture (4.8\%), revealing that benchmark noise stems as much from contributor practices as from model characteristics. We further demonstrate methods to assess the reliability of scaling laws by controlling for model size and other confounds. Our findings provide actionable diagnostics for when benchmark rankings can be trusted and establish a measurement framework for evaluating the validity of AI evaluation claims.

Deep Learning · Large Language Models

Tong Wu, Michael Liu, Jun Bai, Zixia Jia, Shuyi Zhang, Ziyong Lin, Yanting Wang, Song-Chun Zhu, Zilong Zheng

We introduce **Native Parallel Reasoner (NPR)**, a teacher-free framework that enables Large Language Models (LLMs) to self-evolve genuine parallel reasoning capabilities. NPR transforms the model from sequential emulation to native parallel cognition through three key innovations: 1) a **self-distilled** progressive training paradigm that transitions from ``cold-start'' format discovery to strict topological constraints without external supervision; 2) a novel **Parallel-Aware Policy Optimization (PAPO)** algorithm that optimizes branching policies directly within the execution graph, allowing the model to learn adaptive decomposition via trial and error; and 3) a robust **NPR Engine** that refactors memory management and flow control of SGLang to enable stable, large-scale parallel RL training. Across eight reasoning benchmarks, NPR trained on Qwen3-4B achieves performance gains of up to 24.5\% and inference speedups up to 4.6$\times$. Unlike prior baselines that often fall back to autoregressive decoding, NPR demonstrates 100\% genuine parallel execution, establishing a new standard for self-evolving, efficient, and scalable agentic reasoning.

Deep Learning · Large Language Models

Sina Mansouri, Mohit Marvania, Abolfazl Safikhani

The widespread adoption of large language models (LLMs) has intensified the demand for principled methods to distinguish human- from machine-generated text. Watermarking provides a promising avenue, yet existing detectors exhibit sharp performance deterioration under multiple paraphrasing and when applied to shorter texts. We introduce Pattern Stability Score (PSS), a novel detection framework that leverages local statistical features and stability dynamics across paraphrased variants. Specifically, the proposed method combines global and local z-score features with higher-order statistics of run-length patterns, enriched by autocorrelation signals and stability scores computed over paraphrase depth. Numerical evaluations are performed on three benchmark datasets (PG-19, CNN/DailyMail, and WikiText) using multiple LLMs (Llama-3-8B, Qwen2-7B) and paraphrasers (Mistral-7B, Qwen2-7B, Gemma-7B), systematically stress-testing robustness under up to eight rounds of paraphrasing. Compared to prior z-score thresholding baselines and some state-of-the-art deep learning methods, our approach improves detection AUC (area under the receiver operating characteristic curve) by over 10-15 percentage points across different token lengths. Additionally, extensive cross-domain experiments demonstrate that a single universal classifier generalizes across different LLMs, paraphrasers, and text domains without retraining, maintaining above 83.7\% AUC even when all components differ from training.

Deep Learning · Foundation Models

Yichen Gong, Zhuohan Cai, Sunhao Dai, Yuqi Zhou, Zhangxuan Gu, Changhua Meng, Shuheng Shen

Existing online benchmarks for mobile GUI agents remain largely app-centric and task-homogeneous, failing to reflect the diversity and instability of real-world mobile usage. To this end, we introduce VenusBench-Mobile, a challenging online benchmark for evaluating general-purpose mobile GUI agents under realistic, user-centric conditions. VenusBench-Mobile builds two core evaluation pillars: defining what to evaluate via user-intent-driven task design that reflects real mobile usage, and how to evaluate through a capability-oriented annotation scheme for fine-grained agent behavior analysis. Extensive evaluation of state-of-the-art mobile GUI agents reveals large performance gaps relative to prior benchmarks, indicating that VenusBench-Mobile poses substantially more challenging and realistic tasks and that current agents remain far from reliable real-world deployment. Diagnostic analysis further shows that failures are dominated by deficiencies in perception and memory, which are largely obscured by coarse-grained evaluations. Moreover, even the strongest agents exhibit near-zero success under environment variations, highlighting their brittleness in realistic settings. Based on these insights, we believe VenusBench-Mobile provides an important stepping stone toward robust real-world deployment of mobile GUI agents.

Deep Learning · Theory

Muhammad Ashiq, Samanyu Arora, Abhinav Narayan Harish, Ishaan Kharbanda, Hung Yun Tseng, Grigorios Chrysos

We theoretically study the hallucination phenomena in two canonical diffusion samplers: the stochastic Denoising Diffusion Probabilistic Model (DDPM) and the deterministic Denoising Diffusion Implicit Model (DDIM). We analyze the reverse ODE (DDIM) and SDE (DDPM) for a Gaussian mixture target, proving that after a critical time $\tau$, (a) DDIM can become stuck on the segment connecting the two nearest modes and (b) the *stochasticity of DDPM* helps DDPM become unstuck from this region, thus avoiding hallucination. Our empirical validation verifies that DDPM has a significantly lower hallucination rate than DDIM when this region is entered. Building on our observations, we exhibit how using additional stochastic steps can help DDIM avoid hallucinations and offer new insights on how to design improved samplers.