# Research Papers — 2026-04 · bundle page 135 of 233 > Full-text markdown of papers #6701–6750 of 11612 published in 2026-04, newest first. AI-readable mirrors of the full 11612-paper 2026-04 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-04/page/135 [← Previous bundle](https://aipod.dev/papers/archive/2026-04/page/134.md) [Next bundle →](https://aipod.dev/papers/archive/2026-04/page/136.md) ### GeomPrompt: Geometric Prompt Learning for RGB-D Semantic Segmentation Under Missing and Degraded Depth Authors: Krishna Jaganathan, Patricio Vela Published: 2026-04-13 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2604.11585v1 Multimodal perception systems for robotics and embodied AI often assume reliable RGB-D sensing, but in practice, depth is frequently missing, noisy, or corrupted. We thus present GeomPrompt, a lightweight cross-modal adaptation module that synthesizes a task-driven geometric prompt from RGB alone for the fourth channel of a frozen RGB-D semantic segmentation model, without depth supervision. We further introduce GeomPrompt-Recovery, an adaptation module that compensates for degraded depth by predicting the fourth channel correction relevant for the frozen segmenter. Both modules are trained solely with downstream segmentation supervision, enabling recovery of the geometric prior useful for segmentation, rather than estimating depth signals. On SUN RGB-D, GeomPrompt improves over RGB-only inference by +6.1 mIoU on DFormer and +3.0 mIoU on GeminiFusion, while remaining competitive with strong monocular depth estimators. For degraded depth, GeomPrompt-Recovery consistently improves robustness, yielding gains up to +3.6 mIoU under severe depth corruptions. GeomPrompt is also substantially more efficient than monocular depth baselines, reaching 7.8 ms latency versus 38.3 ms and 71.9 ms. These results suggest that task-driven geometric prompting is an efficient mechanism for cross-modal compensation under missing and degraded depth inputs in RGB-D perception. --- ### Computation of Least Trimmed Squares: A Branch-and-Bound framework with Hyperplane Arrangement Enhancements Authors: Xiang Meng, Andrés Gómez, Rahul Mazumder Published: 2026-04-13 Categories: math.OC, cs.LG, math.ST arXiv: https://arxiv.org/abs/2604.11584v2 We study computational aspects of a key problem in robust statistics -- the penalized least trimmed squares (LTS) regression problem, a robust estimator that mitigates the influence of outliers in data by capping residuals with large magnitudes. Although statistically attractive, penalized LTS is NP-hard, and existing mixed-integer optimization (MIO) formulations scale poorly due to weak relaxations and exponential worst-case complexity in the number of observations. We propose a new MIO formulation that embeds hyperplane arrangement logic into a perspective reformulation, explicitly enforcing structural properties of optimal solutions. We show that, if the number of features is fixed, the resulting branch-and-bound tree is of polynomial size in the sample size. Moreover, we develop a tailored branch-and-bound algorithm that uses first-order methods with dual bounds to solve node relaxations efficiently. Computational experiments on synthetic and real datasets demonstrate substantial improvements over existing MIO approaches: on synthetic instances with 5000 samples and 20 features, our tailored solver reaches a 1% gap in 1 minute while competing approaches fail to do so within one hour. These gains enable exact robust regression at significantly larger sample sizes in low-dimensional settings. --- ### A Triadic Suffix Tokenization Scheme for Numerical Reasoning Authors: Olga Chetverina Published: 2026-04-13 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2604.11582v3 Standard subword tokenization methods fragment numbers inconsistently, causing large language models (LLMs) to lose positional and decimal structure - a primary driver of errors in arithmetic and scientific reasoning. We introduce Triadic Suffix Tokenization (TST), a deterministic scheme that partitions digits into three-digit triads and annotates each triad with an explicit magnitude marker. Critically, the scheme defines a fixed, one-to-one mapping between suffixes and orders of magnitude for the integer part (thousands, millions, billions, etc.) and a parallel system of replicated markers for fractional depth (tenths, thousandths, millionths, etc.). Unlike approaches that rely on positional inference, this method provides a consistent gradient signal, which should ensure stable convergence. Two implementation variants are proposed: (1) a vocabulary-based approach that adds at most 10,000 fixed tokens to an existing vocabulary, covering 33 orders of magnitude ($10^{-15}$ to $10^{18}$); and (2) a suffix-marker approach that uses a small set of special tokens to denote magnitude dynamically. Both variants preserve exact digits while making order-of-magnitude relationships transparent at the token level. While we focus on 3-digit groups (Triadic), the framework is inherently scalable to any group size for precise vocabulary optimization. Furthermore, it allows for linear vocabulary expansion to accommodate arbitrary precision and range. TST is architecture-agnostic and can be integrated as a drop-in preprocessing step. Experimental validation is deferred to future work. --- ### Hidden Measurement Error in LLM Pipelines Distorts Annotation, Evaluation, and Benchmarking Authors: Solomon Messing Published: 2026-04-13 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.11581v6 LLM evaluations drive which models get deployed, what safety standards get adopted, which research conclusions get published, and how projections of AI's labor-market impact get made. Yet standard confidence intervals ignore variability from judge model choice, model temperature, and prompt phrasing, producing under-coverage that worsens with more data. The omitted variance can shift results enough to reverse conclusions \citep{baumann2025llmhacking, huang2026dropping}; pipelines that fail to average over it leave the surface that ``benchmark hacking'' exploits \citep{singh2025leaderboard}. This paper decomposes LLM pipeline uncertainty into its sources, distinguishes variance that shrinks with more data from sensitivity to researcher design choices, and uses design-study projections to reduce total evaluation error (TEE). Across the demonstrations, naive standard errors are 40 - 60\% smaller than the TEE-corrected SE. Using Chatbot Arena data, we show naive 95\% CI coverage drops as $n$ grows while TEE-corrected coverage holds at 95\%, and TEE-guided pipelines restrict the benchmark gaming surface from 56 to 32 Elo ($K=27$), below the human-leaderboard baseline. We show further that a small pilot recovers honest CIs and projects which design changes most improve precision. Acting on those projections halves MMLU estimation error against the answer key at equivalent cost, and raises per-match agreement with human votes by 7.9 percentage points on Chatbot Arena. --- ### Seeing Through Touch: Tactile-Driven Visual Localization of Material Regions Authors: Seongyu Kim, Seungwoo Lee, Hyeonggon Ryu, Joon Son Chung, Arda Senocak Published: 2026-04-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.11579v1 We address the problem of tactile localization, where the goal is to identify image regions that share the same material properties as a tactile input. Existing visuo-tactile methods rely on global alignment and thus fail to capture the fine-grained local correspondences required for this task. The challenge is amplified by existing datasets, which predominantly contain close-up, low-diversity images. We propose a model that learns local visuo-tactile alignment via dense cross-modal feature interactions, producing tactile saliency maps for touch-conditioned material segmentation. To overcome dataset constraints, we introduce: (i) in-the-wild multi-material scene images that expand visual diversity, and (ii) a material-diversity pairing strategy that aligns each tactile sample with visually varied yet tactilely consistent images, improving contextual localization and robustness to weak signals. We also construct two new tactile-grounded material segmentation datasets for quantitative evaluation. Experiments on both new and existing benchmarks show that our approach substantially outperforms prior visuo-tactile methods in tactile localization. --- ### Minimizing classical resources in variational measurement-based quantum computation for generative modeling Authors: Arunava Majumder, Hendrik Poulsen Nautrup, Hans J. Briegel Published: 2026-04-13 Categories: quant-ph, cs.AI, cs.LG, stat.ML arXiv: https://arxiv.org/abs/2604.11578v1 Measurement-based quantum computation (MBQC) is a framework for quantum information processing in which a computational task is carried out through one-qubit measurements on a highly entangled resource state. Due to the indeterminacy of the outcomes of a quantum measurement, the random outcomes of these operations, if not corrected, yield a variational quantum channel family. Traditionally, this randomness is corrected through classical processing in order to ensure deterministic unitary computations. Recently, variational measurement-based quantum computation (VMBQC) has been introduced to exploit this measurement-induced randomness to gain an advantage in generative modeling. A limitation of this approach is that the corresponding channel model has twice as many parameters compared to the unitary model, scaling as $N \times D$, where $N$ is the number of logical qubits (width) and $D$ is the depth of the VMBQC model. This can often make optimization more difficult and may lead to poorly trainable models. In this paper, we present a restricted VMBQC model that extends the unitary setting to a channel-based one using only a single additional trainable parameter. We show, both numerically and algebraically, that this minimal extension is sufficient to generate probability distributions that cannot be learned by the corresponding unitary model. --- ### Finetune Like You Pretrain: Boosting Zero-shot Adversarial Robustness in Vision-language Models Authors: Songlong Xing, Weijie Wang, Zhengyu Zhao, Jindong Gu, Philip Torr, Nicu Sebe Published: 2026-04-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.11576v1 Despite their impressive zero-shot abilities, vision-language models such as CLIP have been shown to be susceptible to adversarial attacks. To enhance its adversarial robustness, recent studies finetune the pretrained vision encoder of CLIP with adversarial examples on a proxy dataset such as ImageNet by aligning adversarial images with correct class labels. However, these methods overlook the important roles of training data distributions and learning objectives, resulting in reduced zero-shot capabilities and limited transferability of robustness across domains and datasets. In this work, we propose a simple yet effective paradigm AdvFLYP, which follows the training recipe of CLIP's pretraining process when performing adversarial finetuning to the model. Specifically, AdvFLYP finetunes CLIP with adversarial images created based on image-text pairs collected from the web, and match them with their corresponding texts via a contrastive loss. To alleviate distortion of adversarial image embeddings of noisy web images, we further propose to regularise AdvFLYP by penalising deviation of adversarial image features. We show that logit- and feature-level regularisation terms benefit robustness and clean accuracy, respectively. Extensive experiments on 14 downstream datasets spanning various domains show the superiority of our paradigm over mainstream practices. Our code and model weights are released at https://github.com/Sxing2/AdvFLYP. --- ### MIXAR: Scaling Autoregressive Pixel-based Language Models to Multiple Languages and Scripts Authors: Chen Hu, Yintao Tai, Antonio Vergari, Frank Keller, Alessandro Suglia Published: 2026-04-13 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.11575v1 Pixel-based language models are gaining momentum as alternatives to traditional token-based approaches, promising to circumvent tokenization challenges. However, the inherent perceptual diversity across languages poses a significant hurdle for multilingual generalization in pixel space. This paper introduces MIXAR, the first generative pixel-based language model trained on eight different languages utilizing a range of different scripts. We empirically evaluate MIXAR against previous pixel-based models as well as comparable tokenizer-based models, demonstrating substantial performance improvement on discriminative and generative multilingual tasks. Additionally, we show how MIXAR is robust to languages never seen during the training. These results are further strengthened when scaling the model to 0.5B parameters which not only improves its capabilities in generative tasks like LAMBADA but also its robustness when challenged with input perturbations such as orthographic attacks. --- ### Phonological distances for linguistic typology and the origin of Indo-European languages Authors: Marius Mavridis, Juan De Gregorio, Raul Toral, David Sanchez Published: 2026-04-13 Categories: cs.CL, cond-mat.stat-mech, cs.IT, physics.soc-ph arXiv: https://arxiv.org/abs/2604.11565v1 We show that short-range phoneme dependencies encode large-scale patterns of linguistic relatedness, with direct implications for quantitative typology and evolutionary linguistics. Specifically, using an information-theoretic framework, we argue that phoneme sequences modeled as second-order Markov chains essentially capture the statistical correlations of a phonological system. This finding enables us to quantify distances among 67 modern languages from a multilingual parallel corpus employing a distance metric that incorporates articulatory features of phonemes. The resulting phonological distance matrix recovers major language families and reveals signatures of contact-induced convergence. Remarkably, we obtain a clear correlation with geographic distance, allowing us to constrain a plausible homeland region for the Indo-European family, consistent with the Steppe hypothesis. --- ### Training-Free Model Ensemble for Single-Image Super-Resolution via Strong-Branch Compensation Authors: Gengjia Chang, Xining Ge, Weijun Yuan, Zhan Li, Qiurong Song, Luen Zhu, Shuhong Liu Published: 2026-04-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.11564v2 Single-image super-resolution has progressed from deep convolutional baselines to stronger Transformer and state-space architectures, yet the corresponding performance gains typically come with higher training cost, longer engineering iteration, and heavier deployment burden. In many practical settings, multiple pretrained models with partially complementary behaviors are already available, and the binding constraint is no longer architectural capacity but how effectively their outputs can be combined without additional training. Rather than pursuing further architectural redesign, this paper proposes a training-free output-level ensemble framework. A dual-branch pipeline is constructed in which a Hybrid attention network with TLC inference provides stable main reconstruction, while a MambaIRv2 branch with geometric self-ensemble supplies strong compensation for high-frequency detail recovery. The two branches process the same low-resolution input independently and are fused in the image space via a lightweight weighted combination, without updating any model parameters or introducing an additional trainable module. As our solution to the NTIRE 2026 Image Super-Resolution ($\times 4$) Challenge, the proposed design consistently improves over the base branch and slightly exceeds the pure strong branch in PSNR at the best operating point under a unified DIV2K bicubic $\times 4$ evaluation protocol. Ablation studies confirm that output-level compensation provides a low-overhead and practically accessible upgrade path for existing super-resolution systems. --- ### Synthius-Mem: Brain-Inspired Hallucination-Resistant Persona Memory Achieving 94.4% Memory Accuracy and 99.6% Adversarial Robustness on LoCoMo Authors: Artem Gadzhiev, Andrew Kislov Published: 2026-04-13 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2604.11563v1 Providing AI agents with reliable long-term memory that does not hallucinate remains an open problem. Current approaches to memory for LLM agents -- sliding windows, summarization, embedding-based RAG, and flat fact extraction -- each reduce token cost but introduce catastrophic information loss, semantic drift, or uncontrolled hallucination about the user. The structural reason is architectural: every published memory system on the LoCoMo benchmark treats conversation as a retrieval problem over raw or lightly summarized dialogue segments, and none reports adversarial robustness, the ability to refuse questions about facts the user never disclosed. We present Synthius-Mem, a brain-inspired structured persona memory system that takes a fundamentally different approach. Instead of retrieving what was said, Synthius-Mem extracts what is known about the person: a full persona extraction pipeline decomposes conversations into six cognitive domains (biography, experiences, preferences, social circle, work, psychometrics), consolidates and deduplicates per domain, and retrieves structured facts via CategoryRAG at 21.79 ms latency. On the LoCoMo benchmark (ACL 2024, 10 conversations, 1,813 questions), Synthius-Mem achieves 94.37% accuracy, exceeding all published systems including MemMachine (91.69%, adversarial score is not reported) and human performance (87.9 F1). Core memory fact accuracy reaches 98.64%. Adversarial robustness, the hallucination resistance metric that no competing system reports, reaches 99.55%. Synthius-Mem reduces token consumption by ~5x compared to full-context replay while achieving higher accuracy. Synthius-Mem achieves state-of-the-art results on LoCoMo and is, to our knowledge, the only persona memory system that both exceeds human-level performance and reports adversarial robustness. --- ### The Impact of Federated Learning on Distributed Remote Sensing Archives Authors: Anand Umashankar, Karam Tomotaki-Dawoud, Nicolai Schneider Published: 2026-04-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.11562v1 Remote sensing archives are inherently distributed: Earth observation missions such as Sentinel-1, Sentinel-2, and Sentinel-3 have collectively accumulated more than 5 petabytes of imagery, stored and processed across many geographically dispersed platforms. Training machine learning models on such data in a centralized fashion is impractical due to data volume, sovereignty constraints, and geographic distribution. Federated learning (FL) addresses this by keeping data local and exchanging only model updates. A central challenge for remote sensing is the non-IID nature of Earth observation data: label distributions vary strongly by geographic region, degrading the convergence of standard FL algorithms. In this paper, we conduct a systematic empirical study of three FL strategies -- FedAvg, FedProx, and bulk synchronous parallel (BSP) -- applied to multi-label remote sensing image classification under controlled non-IID label-skew conditions. We evaluate three convolutional neural network (CNN) architectures of increasing depth (LeNet, AlexNet, and ResNet-34) and analyze the joint effect of algorithm choice, model capacity, client fraction, client count, batch size, and communication cost. Experiments on the UC Merced multi-label dataset show that FedProx outperforms FedAvg for deeper architectures under data heterogeneity, that BSP approaches centralized accuracy at the cost of high sequential communication, and that LeNet provides the best accuracy-communication trade-off for the dataset scale considered. --- ### bacpipe: a Python package to make bioacoustic deep learning models accessible Authors: Vincent S. Kather, Sylvain Haupert, Burooj Ghani, Dan Stowell Published: 2026-04-13 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.11560v1 1. Natural sounds have been recorded for millions of hours over the previous decades using passive acoustic monitoring. Improvements in deep learning models have vastly accelerated the analysis of large portions of this data. While new models advance the state-of-the-art, accessing them using tools to harness their full potential is not always straightforward. Here we present bacpipe, a collection of bioacoustic deep learning models and evaluation pipelines accessible through a graphical and programming interface, designed for both ecologists and computer scientists. Bacpipe is a modular software package intended as a point of convergence for bioacoustic models. 2. Bacpipe streamlines the usage of state-of-the-art models on custom audio datasets, generating acoustic feature vectors (embeddings) and classifier predictions. A modular design allows evaluation and benchmarking of models through interactive visualizations, clustering and probing. 3. We believe that access to new deep learning models is important. By designing bacpipe to target a wide audience, researchers will be enabled to answer new ecological and evolutionary questions in bioacoustics. 4. In conclusion, we believe accessibility to developments in deep learning to a wider audience benefits the ecological questions we are trying to answer. --- ### Progressively Texture-Aware Diffusion for Contrast-Enhanced Sparse-View CT Authors: Tianqi Wang, Wenchao Du, Hongyu Yang Published: 2026-04-13 Categories: cs.CV, physics.med-ph arXiv: https://arxiv.org/abs/2604.11559v1 Diffusion-based sparse-view CT (SVCT) imaging has achieved remarkable advancements in recent years, thanks to its more stable generative capability. However, recovering reliable image content and visually consistent textures is still a crucial challenge. In this paper, we present a Progressively Texture-aware Diffusion (PTD) model, a coarse-to-fine learning framework tailored for SVCT. Specifically, PTD comprises a basic reconstructive module PTD$_{\textit{rec}}$ and a conditional diffusion module PTD$_{\textit{diff}}$. PTD$_{\textit{rec}}$ first learns a deterministic mapping to recover the majority of the underlying low-frequency signals (i.e., coarse content with smoothed textures), which serves as the initial estimation to enable fidelity. Moreover, PTD$_{\textit{diff}}$ aims to reconstruct high-fidelity details for coarse prediction, which explores a dual-domain guided conditional diffusion to generate reliable and consistent textures. Extensive experiments on sparse-view CT reconstruction demonstrate that our PTD achieves superior performance in terms of structure similarity and visual appeal with only a few sampling steps, which mitigates the randomness inherent in general diffusion models and enables a better trade-off between visual quality and fidelity of high-frequency details. --- ### UniToolCall: Unifying Tool-Use Representation, Data, and Evaluation for LLM Agents Authors: Yijuan Liang, Xinghao Chen, Yifan Ge, Ziyi Wu, Hao Wu, Changyu Zeng, Wei Xing, Xiaoyu Shen Published: 2026-04-13 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.11557v3 Tool-use capability is a fundamental component of LLM agents, enabling them to interact with external systems through structured function calls. However, existing research exhibits inconsistent interaction representations, largely overlooks the structural distribution of tool-use trajectories, and relies on incompatible evaluation benchmarks. We present UniToolCall, a unified framework for tool learning that standardizes the entire pipeline from toolset construction and dataset generation to evaluation. The framework curates a large tool pool of 22k+ tools and constructs a hybrid training corpus of 390k+ instances by combining 10 standardized public datasets with structurally controlled synthetic trajectories. It explicitly models diverse interaction patterns, including single-hop vs. multi-hop and single-turn vs. multi-turn, while capturing both serial and parallel execution structures. To support coherent multi-turn reasoning, we further introduce an Anchor Linkage mechanism that enforces cross-turn dependencies. Furthermore, we convert 7 public benchmarks into a unified Query--Action--Observation--Answer (QAOA) representation with fine-grained evaluation at the function-call, turn, and conversation levels. Experiments show that fine-tuning Qwen3-8B on our dataset substantially improves tool-use performance. Under the distractor-heavy Hybrid-20 setting, achieves 93.0% single-turn Strict Precision, outperforming commercial models including GPT, Gemini, and Claude. --- ### UniToolCall: Unifying Tool-Use Representation, Data, and Evaluation for LLM Agents Authors: Yijuan Liang, Xinghao Chen, Yifan Ge, Ziyi Wu, Hao Wu, Changyu Zeng, Wei Xing, Xiaoyu Shen Published: 2026-04-13 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.11557v2 Tool-use capability is a fundamental component of LLM agents, enabling them to interact with external systems through structured function calls. However, existing research exhibits inconsistent interaction representations, largely overlooks the structural distribution of tool-use trajectories, and relies on incompatible evaluation benchmarks. We present UniToolCall, a unified framework for tool learning that standardizes the entire pipeline from toolset construction and dataset generation to evaluation. The framework curates a large tool pool of 22k+ tools and constructs a hybrid training corpus of 390k+ instances by combining 10 standardized public datasets with structurally controlled synthetic trajectories. It explicitly models diverse interaction patterns, including single-hop vs. multi-hop and single-turn vs. multi-turn, while capturing both serial and parallel execution structures. To support coherent multi-turn reasoning, we further introduce an Anchor Linkage mechanism that enforces cross-turn dependencies. Furthermore, we convert 7 public benchmarks into a unified Query--Action--Observation--Answer (QAOA) representation with fine-grained evaluation at the function-call, turn, and conversation levels. Experiments show that fine-tuning Qwen3-8B on our dataset substantially improves tool-use performance. Under the distractor-heavy Hybrid-20 setting, achieves 93.0% single-turn Strict Precision, outperforming commercial models including GPT, Gemini, and Claude. --- ### FM-Agent: Scaling Formal Methods to Large Systems via LLM-Based Hoare-Style Reasoning Authors: Haoran Ding, Zhaoguo Wang, Haibo Chen Published: 2026-04-13 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2604.11556v2 LLM-assisted software development has become increasingly prevalent, and can generate large-scale systems, such as compilers. It becomes crucial to strengthen the correctness of the generated code. However, automated reasoning for large-scale systems remains challenging due to code complexity. Hoare logic offers an approach to decomposing a large system into smaller components and reasoning about them separately (i.e., compositional reasoning). However, existing works still struggle to scale, because Hoare logic requires writing formal specifications for each function, imposing a heavy human burden. The problem is exacerbated when code is generated by LLMs, as developers lack a deep understanding of each function's expected behavior. This paper presents FM-Agent, the first framework that realizes automated compositional reasoning for large-scale systems. Leveraging LLMs, FM-Agent introduces a top-down paradigm to automatically generate function-level specifications. Specifically, FM-Agent derives the specification of a function from how its callers expect the function to behave, so the generated specifications can reflect the developer's intent of a function even if the implementation is buggy. Developers' intent is usually expressed in natural language, while existing verifiers only support formulas. Therefore, FM-Agent generalizes Hoare-style inference to reason about functions against natural-language specifications. Finally, to confirm bug existence and explain bug causes, FM-Agent automatically generates test cases to trigger potential bugs. In our evaluation, FM-Agent successfully reasons about large-scale systems within 2 days, each of which has up to 143k LoC. These systems have already been tested by their developers, but FM-Agent still finds 522 newly discovered bugs. These bugs can cause serious consequences, including system crashes and incorrect execution results. --- ### Relax: An Asynchronous Reinforcement Learning Engine for Omni-Modal Post-Training at Scale Authors: Liujie Zhang, Benzhe Ning, Rui Yang, Xiaoyan Yu, Jiaxing Li, Lumeng Wu, Jia Liu, Minghao Li, Weihang Chen, Weiqi Hu, Lei Zhang Published: 2026-04-13 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.11554v2 Reinforcement learning (RL) post-training has proven effective at unlocking reasoning, self-reflection, and tool-use capabilities in large language models. As models extend to omni-modal inputs and agentic multi-turn workflows, RL training systems face three interdependent challenges: heterogeneous data flows, operational robustness at scale, and the staleness -- throughput tradeoff. We present \textbf{Relax} (Reinforcement Engine Leveraging Agentic X-modality), an open-source RL training engine that addresses these challenges through three co-designed architectural layers. First, an \emph{omni-native architecture} builds multimodal support into the full stack -- from data preprocessing and modality-aware parallelism to inference generation -- rather than retrofitting it onto a text-centric pipeline. Second, each RL role runs as an independent, fault-isolated service that can be scaled, recovered, and upgraded without global coordination. Third, service-level decoupling enables asynchronous training via the TransferQueue data bus, where a single staleness parameter smoothly interpolates among on-policy, near-on-policy, and fully asynchronous execution. Relax achieves a 1.20$\times$ end-to-end speedup over veRL on Qwen3-4B on-policy training. Its fully async mode delivers a 1.76$\times$ speedup over colocate on Qwen3-4B and a 2.00$\times$ speedup on Qwen3-Omni-30B, while all modes converge to the same reward level. Relax supports R3 (Rollout Routing Replay)~\cite{ma2025r3} for MoE models with only 1.9\% overhead, compared to 32\% degradation in veRL under the same configuration. It further demonstrates stable omni-modal RL convergence on Qwen3-Omni across image, text, and audio, sustaining over 2{,}000 steps on video without degradation. Relax is available at https://github.com/rednote-ai/Relax. --- ### MimicLM: Zero-Shot Voice Imitation through Autoregressive Modeling of Pseudo-Parallel Speech Corpora Authors: Tao Feng, Yuxiang Wang, Yuancheng Wang, Xueyao Zhang, Dekun Chen, Chaoren Wang, Xun Guan, Zhizheng Wu Published: 2026-04-13 Categories: cs.SD, cs.CL arXiv: https://arxiv.org/abs/2604.11552v2 Voice imitation aims to transform source speech to match a reference speaker's timbre and speaking style while preserving linguistic content. A straightforward approach is to train on triplets of (source, reference, target), where source and target share the same content but target matches the reference's voice characteristics, yet such data is extremely scarce. Existing approaches either employ carefully designed disentanglement architectures to bypass this data scarcity or leverage external systems to synthesize pseudo-parallel training data. However, the former requires intricate model design, and the latter faces a quality ceiling when synthetic speech is used as training targets. To address these limitations, we propose MimicLM, which takes a novel approach by using synthetic speech as training sources while retaining real recordings as targets. This design enables the model to learn directly from real speech distributions, breaking the synthetic quality ceiling. Building on this data construction approach, we incorporate interleaved text-audio modeling to guide the generation of content-accurate speech and apply post-training with preference alignment to mitigate the inherent distributional mismatch when training on synthetic data. Experiments demonstrate that MimicLM achieves superior voice imitation quality with a simple yet effective architecture, significantly outperforming existing methods in naturalness while maintaining competitive similarity scores across speaker identity, accent, and emotion dimensions. --- ### Human Centered Non Intrusive Driver State Modeling Using Personalized Physiological Signals in Real World Automated Driving Authors: David Puertas-Ramirez, Raul Fernandez-Matellan, David Martin Gomez, Jesus G. Boticario Published: 2026-04-13 Categories: cs.HC, cs.LG, cs.RO arXiv: https://arxiv.org/abs/2604.11549v1 In vehicles with partial or conditional driving automation (SAE Levels 2-3), the driver remains responsible for supervising the system and responding to take-over requests. Therefore, reliable driver monitoring is essential for safe human-automation collaboration. However, most existing Driver Monitoring Systems rely on generalized models that ignore individual physiological variability. In this study, we examine the feasibility of personalized driver state modeling using non-intrusive physiological sensing during real-world automated driving. We conducted experiments in an SAE Level 2 vehicle using an Empatica E4 wearable sensor to capture multimodal physiological signals, including electrodermal activity, heart rate, temperature, and motion data. To leverage deep learning architectures designed for images, we transformed the physiological signals into two-dimensional representations and processed them using a multimodal architecture based on pre-trained ResNet50 feature extractors. Experiments across four drivers demonstrate substantial interindividual variability in physiological patterns related to driver awareness. Personalized models achieved an average accuracy of 92.68%, whereas generalized models trained on multiple users dropped to an accuracy of 54%, revealing substantial limitations in cross-user generalization. These results underscore the necessity of adaptive, personalized driver monitoring systems for future automated vehicles and imply that autonomous systems should adapt to each driver's unique physiological profile. --- ### SemaClaw: A Step Towards General-Purpose Personal AI Agents through Harness Engineering Authors: Ningyan Zhu, Huacan Wang, Jie Zhou, Feiyu Chen, Shuo Zhang, Ge Chen, Chen Liu, Jiarou Wu, Wangyi Chen, Xiaofeng Mou, Yi Xu Published: 2026-04-13 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.11548v1 The rise of OpenClaw in early 2026 marks the moment when millions of users began deploying personal AI agents into their daily lives, delegating tasks ranging from travel planning to multi-step research. This scale of adoption signals that two parallel arcs of development have reached an inflection point. First is a paradigm shift in AI engineering, evolving from prompt and context engineering to harness engineering-designing the complete infrastructure necessary to transform unconstrained agents into controllable, auditable, and production-reliable systems. As model capabilities converge, this harness layer is becoming the primary site of architectural differentiation. Second is the evolution of human-agent interaction from discrete tasks toward a persistent, contextually aware collaborative relationship, which demands open, trustworthy and extensible harness infrastructure. We present SemaClaw, an open-source multi-agent application framework that addresses these shifts by taking a step towards general-purpose personal AI agents through harness engineering. Our primary contributions include a DAG-based two-phase hybrid agent team orchestration method, a PermissionBridge behavioral safety system, a three-tier context management architecture, and an agentic wiki skill for automated personal knowledge base construction. --- ### Eliciting Medical Reasoning with Knowledge-enhanced Data Synthesis: A Semi-Supervised Reinforcement Learning Approach Authors: Haolin Li, Shuyang Jiang, Ruipeng Zhang, Jiangchao Yao, Ya Zhang, Yanfeng Wang Published: 2026-04-13 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2604.11547v1 While large language models hold promise for complex medical applications, their development is hindered by the scarcity of high-quality reasoning data. To address this issue, existing approaches typically distill chain-of-thought reasoning traces from large proprietary models via supervised fine-tuning, then conduct reinforcement learning (RL). These methods exhibit limited improvement on underrepresented domains like rare diseases while incurring substantial costs from generating complex reasoning chains. To efficiently enhance medical reasoning, we propose MedSSR, a Medical Knowledge-enhanced data Synthesis and Semi-supervised Reinforcement learning framework. Our framework first employs rare disease knowledge to synthesize distribution-controllable reasoning questions. We then utilize the policy model itself to generate high-quality pseudo-labels. This enables a two-stage, intrinsic-to-extrinsic training paradigm: self-supervised RL on the pseudo-labeled synthetic data, followed by supervised RL on the human-annotated real data. MedSSR scales model training efficiently without relying on costly trace distillation. Extensive experiments on Qwen and Llama demonstrate that our method outperforms existing methods across ten medical benchmarks, achieving up to +5.93% gain on rare-disease tasks. Our code is available at https://github.com/tdlhl/MedSSR. --- ### Time is Not a Label: Continuous Phase Rotation for Temporal Knowledge Graphs and Agentic Memory Authors: Weixian Waylon Li, Jiaxin Zhang, Xianan Jim Yang, Tiejun Ma, Yiwen Guo Published: 2026-04-13 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.11544v2 Structured memory representations such as knowledge graphs are central to autonomous agents and other long-lived systems. However, most existing approaches model time as discrete metadata, either sorting by recency (burying old-yet-permanent knowledge), simply overwriting outdated facts, or requiring an expensive LLM call at every ingestion step, leaving them unable to distinguish persistent facts from evolving ones. To address this, we introduce RoMem, a drop-in temporal knowledge graph module for structured memory systems, applicable to agentic memory and beyond. A pretrained Semantic Speed Gate maps each relation's text embedding to a volatility score, learning from data that evolving relations (e.g., "president of") should rotate fast while persistent ones (e.g., "born in") should remain stable. Combined with continuous phase rotation, this enables geometric shadowing: obsolete facts are rotated out of phase in complex vector space, so temporally correct facts naturally outrank contradictions without deletion. On temporal knowledge graph completion, RoMem achieves state-of-the-art results on ICEWS05-15 (72.6 MRR). Applied to agentic memory, it delivers 2-3x MRR and answer accuracy on temporal reasoning (MultiTQ), dominates hybrid benchmark (LoCoMo), preserves static memory with zero degradation (DMR-MSC), and generalises zero-shot to unseen financial domains (FinTMMBench). --- ### Time is Not a Label: Continuous Phase Rotation for Temporal Knowledge Graphs and Agentic Memory Authors: Weixian Waylon Li, Jiaxin Zhang, Xianan Jim Yang, Tiejun Ma, Yiwen Guo Published: 2026-04-13 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.11544v1 Structured memory representations such as knowledge graphs are central to autonomous agents and other long-lived systems. However, most existing approaches model time as discrete metadata, either sorting by recency (burying old-yet-permanent knowledge), simply overwriting outdated facts, or requiring an expensive LLM call at every ingestion step, leaving them unable to distinguish persistent facts from evolving ones. To address this, we introduce RoMem, a drop-in temporal knowledge graph module for structured memory systems, applicable to agentic memory and beyond. A pretrained Semantic Speed Gate maps each relation's text embedding to a volatility score, learning from data that evolving relations (e.g., "president of") should rotate fast while persistent ones (e.g., "born in") should remain stable. Combined with continuous phase rotation, this enables geometric shadowing: obsolete facts are rotated out of phase in complex vector space, so temporally correct facts naturally outrank contradictions without deletion. On temporal knowledge graph completion, RoMem achieves state-of-the-art results on ICEWS05-15 (72.6 MRR). Applied to agentic memory, it delivers 2-3x MRR and answer accuracy on temporal reasoning (MultiTQ), dominates hybrid benchmark (LoCoMo), preserves static memory with zero degradation (DMR-MSC), and generalises zero-shot to unseen financial domains (FinTMMBench). --- ### NovBench: Evaluating Large Language Models on Academic Paper Novelty Assessment Authors: Wenqing Wu, Yi Zhao, Yuzhuo Wang, Siyou Li, Juexi Shao, Yunfei Long, Chengzhi Zhang Published: 2026-04-13 Categories: cs.CL, cs.AI, cs.DL, cs.IR arXiv: https://arxiv.org/abs/2604.11543v1 Novelty is a core requirement in academic publishing and a central focus of peer review, yet the growing volume of submissions has placed increasing pressure on human reviewers. While large language models (LLMs), including those fine-tuned on peer review data, have shown promise in generating review comments, the absence of a dedicated benchmark has limited systematic evaluation of their ability to assess research novelty. To address this gap, we introduce NovBench, the first large-scale benchmark designed to evaluate LLMs' capability to generate novelty evaluations in support of human peer review. NovBench comprises 1,684 paper-review pairs from a leading NLP conference, including novelty descriptions extracted from paper introductions and corresponding expert-written novelty evaluations. We focus on both sources because the introduction provides a standardized and explicit articulation of novelty claims, while expert-written novelty evaluations constitute one of the current gold standards of human judgment. Furthermore, we propose a four-dimensional evaluation framework (including Relevance, Correctness, Coverage, and Clarity) to assess the quality of LLM-generated novelty evaluations. Extensive experiments on both general and specialized LLMs under different prompting strategies reveal that current models exhibit limited understanding of scientific novelty, and that fine--tuned models often suffer from instruction-following deficiencies. These findings underscore the need for targeted fine-tuning strategies that jointly improve novelty comprehension and instruction adherence. --- ### A collaborative agent with two lightweight synergistic models for autonomous crystal materials research Authors: Tongyu Shi, Yutang Li, Zhanyuan Li, Qian Liu, Jie Zhou, Wenhe Xu, Yang Li, Dawei Dai, Rui He, Wenhua Zhou, Jiahong Wang, Xue-Feng Yu Published: 2026-04-13 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.11540v1 Current large language models require hundreds of billions of parameters yet struggle with domain-specific reasoning and tool coordination in materials science. Here, we present MatBrain, a lightweight collaborative agent system with two synergistic models specialization for crystal materials research. MatBrain employs a dual-model architecture: Mat-R1 (30B parameters) as the analytical model providing expert-level domain reasoning, and Mat-T1 (14B parameters) as the executive model orchestrating tool-based actions. Entropy analysis confirms that this architecture resolves the conflict between tool planning and analytical reasoning by decoupling their distinct entropy dynamics. Enabled by this dual-model architecture and structural efficiency, MatBrain significantly outperforms larger general-purpose models while reducing the hardware deployment barrier by over 95%. MatBrain exhibits versatility across structure generation, property prediction, and synthesis planning tasks. Applied to catalyst design, MatBrain generated 30,000 candidate structures and identified 38 promising materials within 48 hours, achieving approximately 100-fold acceleration over traditional approaches. These results demonstrate the potential of lightweight collaborative intelligence for advancing materials research capabilities. --- ### CLAY: Conditional Visual Similarity Modulation in Vision-Language Embedding Space Authors: Sohwi Lim, Lee Hyoseok, Jungjoon Park, Tae-Hyun Oh Published: 2026-04-13 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2604.11539v2 Human perception of visual similarity is inherently adaptive and subjective, depending on the users' interests and focus. However, most image retrieval systems fail to reflect this flexibility, relying on a fixed, monolithic metric that cannot incorporate multiple conditions simultaneously. To address this, we propose CLAY, an adaptive similarity computation method that reframes the embedding space of pretrained Vision-Language Models (VLMs) as a text-conditional similarity space without additional training. This design separates the textual conditioning process and visual feature extraction, allowing highly efficient and multi-conditioned retrieval with fixed visual embeddings. We also construct a synthetic evaluation dataset CLAY-EVAL, for comprehensive assessment under diverse conditioned retrieval settings. Experiments on standard datasets and our proposed dataset show that CLAY achieves high retrieval accuracy and notable computational efficiency compared to previous works. --- ### Problem Reductions at Scale: Agentic Integration of Computationally Hard Problems Authors: Xi-Wei Pan, Shi-Wen An, Jin-Guo Liu Published: 2026-04-13 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.11535v2 Solving an NP-hard optimization problem often requires reformulating it for a specific solver -- quantum hardware, a commercial optimizer, or a domain heuristic. A tool for polynomial-time reductions between hard problems would let practitioners route any supported problem to any supported solver through a single interface. Building such a library at scale, however, has remained out of reach. We show that harness engineering, the practice of designing constraints, verification systems, and feedback loops that channel AI coding agents, can overcome this barrier. Our harness combines a no-code contribution route for domain experts, a multilayer verification stack ranging from type-level checks to agentic feature tests (AI agents role-playing as end users), and a fully automated implementation-review-integration pipeline. In about three months, we built a command-line tool backed by a library of 100+ problem types and 200+ reduction rules in over 170k lines of Rust. The result suggests that a well-engineered harness lets agents build well-tested software at a scale and pace beyond prior reduction-library efforts. Because the reduction graph composes transitively, a new solver registered for any single problem type instantly becomes available to every problem connected by a reduction path. The source code is available at https://github.com/CodingThrust/problem-reductions. --- ### Beyond Attention Scores: SVD-Based Vision Token Pruning for Efficient Vision-Language Models Authors: Yvon Apedo, Martyna Poreba, Michal Szczepanski, Samia Bouchafa Published: 2026-04-13 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2604.11530v3 Vision-Language Models (VLMs) have revolutionized multi-modal learning by jointly processing visual and textual information. Yet, they face significant challenges due to the high computational and memory demands of processing long sequences of vision tokens. Many existing methods rely on local heuristics, such as attention scores or token norms. However, these criteria suffer from positional bias and information dispersion, limiting their ability to preserve essential content at high pruning ratios and leading to performance degradation on visually detailed images. To address these issues, we propose SVD-Prune, a training-free, plug-and-play token pruning method based on Singular Value Decomposition. It decomposes the vision token feature matrix and selects the top-k tokens using statistical leverage scores, ensuring only tokens contributing most to the dominant global variance are preserved. Experiments show that SVD-Prune consistently outperforms prior pruning methods under extreme vision token budgets, maintaining strong performance even with 32 and 16 vision tokens. --- ### TempusBench: An Evaluation Framework for Time-Series Forecasting Authors: Denizalp Goktas, Gerardo Riaño-Briceño, Alif Abdullah, Aryan Nair, Chenkai Shen, Beatriz de Lucio, Alexandra Magnusson, Farhan Mashrur, Ahmed Abdulla, Shawrna Sen, Mahitha Thippireddy, Gregory Schwartz et al. Published: 2026-04-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.11529v2 Foundation models have transformed natural language processing and computer vision, and a rapidly growing literature on time-series foundation models (TSFMs) seeks to replicate this success in forecasting. While recent open-source models demonstrate the promise of TSFMs, the field lacks a comprehensive and community-accepted model evaluation framework. We see at least four major issues impeding progress on the development of such a framework. First, existing evaluation frameworks comprise benchmark forecasting tasks derived from often outdated datasets (e.g., M3), many of which lack clear metadata and overlap with the corpora used to pre-train TSFMs. Second, these frameworks evaluate models along a narrowly defined set of benchmark forecasting tasks, such as forecast horizon length or domain, but overlook core statistical properties such as non-stationarity and seasonality. Third, domain-specific models (e.g., XGBoost) are often compared unfairly, as existing frameworks do not enforce a systematic and consistent hyperparameter tuning convention for all models. Fourth, visualization tools for interpreting comparative performance are lacking. To address these issues, we introduce TempusBench, an open-source evaluation framework for TSFMs. TempusBench consists of 1) new datasets which are not included in existing TSFM pretraining corpora, 2) a set of novel benchmark tasks that go beyond existing ones, 3) a model evaluation pipeline with a standardized hyperparameter tuning protocol, and 4) a tensorboard-based visualization interface. We provide access to our code on GitHub: https://github.com/Smlcrm/TempusBench and maintain a live leaderboard at https://benchmark.smlcrm.com/. --- ### Limited Perfect Monotonical Surrogates constructed using low-cost recursive linkage discovery with guaranteed output Authors: M. W. Przewozniczek, F. Chicano, R. Tinós, M. M. Komarnicki Published: 2026-04-13 Categories: cs.AI, cs.DS arXiv: https://arxiv.org/abs/2604.11524v1 Surrogates provide a cheap solution evaluation and offer significant leverage for optimizing computationally expensive problems. Usually, surrogates only approximate the original function. Recently, the perfect linear surrogates were proposed that ideally represent the original function. These surrogates do not mimic the original function. In fact, they are another (correct) representation of it and enable a wide range of possibilities, e.g., discovering the optimized function for problems where the direct transformation of the encoded solution into its evaluation is not available. However, many real-world problems can not be represented by linear models, making the aforementioned surrogates inapplicable. Therefore, we propose the Limited Monotonical Perfect Surrogate (LyMPuS), which overcomes this difficulty and enables the comparison of two solutions that differ by a single variable. Our proposition is suitable for limiting the cost of expensive local search procedures. The proposed surrogate is parameterless and can be trained on the fly without any separate surrogate-building step. It uses only the necessary fitness evaluations, and the already-paid costs are not wasted when the model is updated. Finally, it offers low-cost missing-linkage detection and low-cost linkage discovery, guaranteed to find a missing dependency in no more than $2\lceil\log_2(n)\rceil$ steps. --- ### PAC-BENCH: Evaluating Multi-Agent Collaboration under Privacy Constraints Authors: Minjun Park, Donghyun Kim, Hyeonjong Ju, Seungwon Lim, Dongwook Choi, Taeyoon Kwon, Minju Kim, Jinyoung Yeo Published: 2026-04-13 Categories: cs.AI, cs.MA arXiv: https://arxiv.org/abs/2604.11523v1 We are entering an era in which individuals and organizations increasingly deploy dedicated AI agents that interact and collaborate with other agents. However, the dynamics of multi-agent collaboration under privacy constraints remain poorly understood. In this work, we present $PAC\text{-}Bench$, a benchmark for systematic evaluation of multi-agent collaboration under privacy constraints. Experiments on $PAC\text{-}Bench$ show that privacy constraints substantially degrade collaboration performance and make outcomes depend more on the initiating agent than the partner. Further analysis reveals that this degradation is driven by recurring coordination breakdowns, including early-stage privacy violations, overly conservative abstraction, and privacy-induced hallucinations. Together, our findings identify privacy-aware multi-agent collaboration as a distinct and unresolved challenge that requires new coordination mechanisms beyond existing agent capabilities. --- ### Triviality Corrected Endogenous Reward Authors: Xinda Wang, Zhengxu Hou, Yangshijie Zhang, Bingren Yan, Jialin Liu, Chenzhuo Zhao, Zhibo Yang, Bin-Bin Yang, Feng Xiao Published: 2026-04-13 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.11522v1 Reinforcement learning for open-ended text generation is constrained by the lack of verifiable rewards, necessitating reliance on judge models that require either annotated data or powerful closed-source models. Inspired by recent work on unsupervised reinforcement learning for mathematical reasoning using confidence-based endogenous rewards, we investigate whether this principle can be adapted to open-ended writing tasks. We find that directly applying confidence rewards leads to Triviality Bias: the policy collapses toward high-probability outputs, reducing diversity and meaningful content. We propose TCER (Triviality Corrected Endogenous Reward), which addresses this bias by rewarding the relative information gain between a specialist policy and a generalist reference policy, modulated by a probability-dependent correction mechanism. Across multiple writing benchmarks and model architectures, TCER achieves consistent improvements without external supervision. Furthermore, TCER also transfers effectively to mathematical reasoning, validating the generality of our approach across different generation tasks. --- ### Continuous Adversarial Flow Models Authors: Shanchuan Lin, Ceyuan Yang, Zhijie Lin, Hao Chen, Haoqi Fan Published: 2026-04-13 Categories: cs.LG, cs.CV arXiv: https://arxiv.org/abs/2604.11521v2 We propose continuous adversarial flow models, a type of continuous-time flow model trained with an adversarial objective. Unlike flow matching, which uses a fixed mean-squared-error criterion, our approach introduces a learned discriminator to guide training. This change in objective induces a different generalized distribution, which empirically produces samples that are better aligned with the target data distribution. Our method is primarily proposed for post-training existing flow-matching models, although it can also train models from scratch. On the ImageNet 256px generation task, our post-training substantially improves the guidance-free FID of latent-space SiT from 8.26 to 3.63 and of pixel-space JiT from 7.17 to 3.57. It also improves guided generation, reducing FID from 2.06 to 1.53 for SiT and from 1.86 to 1.80 for JiT. We further evaluate our approach on text-to-image generation, where it achieves improved results on both the GenEval and DPG benchmarks. --- ### Continuous Adversarial Flow Models Authors: Shanchuan Lin, Ceyuan Yang, Zhijie Lin, Hao Chen, Haoqi Fan Published: 2026-04-13 Categories: cs.LG, cs.CV arXiv: https://arxiv.org/abs/2604.11521v1 We propose continuous adversarial flow models, a type of continuous-time flow model trained with an adversarial objective. Unlike flow matching, which uses a fixed mean-squared-error criterion, our approach introduces a learned discriminator to guide training. This change in objective induces a different generalized distribution, which empirically produces samples that are better aligned with the target data distribution. Our method is primarily proposed for post-training existing flow-matching models, although it can also train models from scratch. On the ImageNet 256px generation task, our post-training substantially improves the guidance-free FID of latent-space SiT from 8.26 to 3.63 and of pixel-space JiT from 7.17 to 3.57. It also improves guided generation, reducing FID from 2.06 to 1.53 for SiT and from 1.86 to 1.80 for JiT. We further evaluate our approach on text-to-image generation, where it achieves improved results on both the GenEval and DPG benchmarks. --- ### Generative Path-Finding Method for Wasserstein Gradient Flow Authors: Chengyu Liu, Xiang Zhou Published: 2026-04-13 Categories: cs.LG, math-ph arXiv: https://arxiv.org/abs/2604.11519v1 Wasserstein gradient flows (WGFs) describe the evolution of probability distributions in Wasserstein space as steepest descent dynamics for a free energy functional. Computing the full path from an arbitrary initial distribution to equilibrium is challenging, especially in high dimensions. Eulerian methods suffer from the curse of dimensionality, while existing Lagrangian approaches based on particles or generative maps do not naturally improve efficiency through time step tuning. We propose GenWGP, a generative path finding framework for Wasserstein gradient paths. GenWGP learns a generative flow that transports mass from an initial density to an unknown equilibrium distribution by minimizing a path loss that encodes the full trajectory and its terminal equilibrium condition. The loss is derived from a geometric action functional motivated by Dawson Gartner large deviation theory for empirical distributions of interacting diffusion systems. We formulate both a finite horizon action under physical time parametrization and a reparameterization invariant geometric action based on Wasserstein arclength. Using normalizing flows, GenWGP computes a geometric curve toward equilibrium while enforcing approximately constant intrinsic speed between adjacent network layers, so that discretized distributions remain nearly equidistant in the Wasserstein metric along the path. This avoids delicate time stepping constraints and enables stable training that is largely independent of temporal or geometric discretization. Experiments on Fokker Planck and aggregation type problems show that GenWGP matches or exceeds high fidelity reference solutions with only about a dozen discretization points while capturing complex dynamics. --- ### From Translation to Superset: Benchmark-Driven Evolution of a Production AI Agent from Rust to Python Authors: Jinhua Wang, Biswa Sengupta Published: 2026-04-13 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2604.11518v1 Cross-language migration of large software systems is a persistent engineering challenge, particularly when the source codebase evolves rapidly. We present a methodology for LLM-assisted continuous code translation in which a large language model translates a production Rust codebase (648K LOC, 65 crates) into Python (41K LOC, 28 modules), with public agent benchmarks as the objective function driving iterative refinement. Our subject system is Codex CLI, a production AI coding agent. We demonstrate that: (1) the Python port resolves 59/80 SWE-bench Verified tasks (73.8%) versus Rust's 56/80 (70.0%), and achieves 42.5% on Terminal-Bench versus Rust's 47.5%, confirming near-parity on real-world agentic tasks; (2) benchmark-driven debugging, revealing API protocol mismatches, environment pollution, a silent WebSocket failure mode, and an API 400 crash, is more effective than static testing alone; (3) the architecture supports continuous upstream synchronisation via an LLM-assisted diff-translate-test loop; and (4) the Python port has evolved into a capability superset with 30 feature-flagged extensions (multi-agent orchestration, semantic memory, guardian safety, cost tracking) absent from Rust, while preserving strict parity mode for comparison. Our evaluation shows that for LLM-based agents where API latency dominates, Python's expressiveness yields a 15.9x code reduction with negligible performance cost, while the benchmark-as-objective-function methodology provides a principled framework for growing a cross-language port from parity into an extended platform. --- ### DuET: Dual Execution for Test Output Prediction with Generated Code and Pseudocode Authors: Hojae Han, Jaejin Kim, Seung-won Hwang, Yu Jin Kim, Moontae Lee Published: 2026-04-13 Categories: cs.SE, cs.CL arXiv: https://arxiv.org/abs/2604.11514v1 This work addresses test output prediction, a key challenge in test case generation. To improve the reliability of predicted outputs by LLMs, prior approaches generate code first to ground predictions. One grounding strategy is direct execution of generated code, but even minor errors can cause failures. To address this, we introduce LLM-based pseudocode execution, which grounds prediction on more error-resilient pseudocode and simulates execution via LLM reasoning. We further propose DuET, a dual-execution framework that combines both approaches by functional majority voting. Our analysis shows the two approaches are complementary in overcoming the limitations of direct execution suffering from code errors, and pseudocode reasoning from hallucination. On LiveCodeBench, DuET achieves the state-of-the-art performance, improving Pass@1 by 13.6 pp. --- ### Machine-learning modeling of magnetization dynamics in quasi-equilibrium and driven metallic spin systems Authors: Gia-Wei Chern, Yunhao Fan, Sheng Zhang, Puhan Zhang Published: 2026-04-13 Categories: cond-mat.str-el, cs.LG, physics.comp-ph arXiv: https://arxiv.org/abs/2604.11513v1 We review recent advances in machine-learning (ML) force-field methods for large-scale Landau-Lifshitz-Gilbert (LLG) simulations of metallic spin systems. We generalize the Behler-Parrinello (BP) ML architecture -- originally developed for quantum molecular dynamics -- to construct scalable and transferable ML models capable of capturing the intricate dependence of electron-mediated exchange fields on the local magnetic environment characteristic of itinerant magnets. A central ingredient of this framework is the implementation of symmetry-aware magnetic descriptors based on group-theoretical bispectrum formalisms. Leveraging these ML force fields, LLG simulations faithfully reproduce hallmark non-collinear magnetic orders -- such as the $120^\circ$ and tetrahedral states -- on the triangular lattice, and successfully capture the complex spin textures emerging in the mixed-phase states of a square-lattice double-exchange model under thermal quench. We further discuss a generalized potential theory that extends the BP formalism to incorporate both conservative and nonconservative electronic torques, thereby enabling ML models to learn nonequilibrium exchange fields from computationally demanding microscopic approaches such as nonequilibrium Green's-function techniques. This extension yields quantitatively accurate predictions of voltage-driven domain-wall motion and establishes a foundation for quantum-accurate, multiscale modeling of nonequilibrium spin dynamics and spintronic functionalities. --- ### EdgeCIM: A Hardware-Software Co-Design for CIM-Based Acceleration of Small Language Models Authors: Jinane Bazzi, Mariam Rakka, Fadi Kurdahi, Mohammed E. Fouda, Ahmed Eltawil Published: 2026-04-13 Categories: cs.AR, cs.AI arXiv: https://arxiv.org/abs/2604.11512v1 The growing demand for deploying Small Language Models (SLMs) on edge devices, including laptops, smartphones, and embedded platforms, has exposed fundamental inefficiencies in existing accelerators. While GPUs handle prefill workloads efficiently, the autoregressive decoding phase is dominated by GEMV operations that are inherently memory-bound, resulting in poor utilization and prohibitive energy costs at the edge. In this work, we present EdgeCIM, a hardware-software co-design framework that rethinks accelerator design for end-to-end decoder-only inference. At its core is a CIM macro, implemented in 65nm, coupled with a tile-based mapping strategy that balances pipeline stages, maximizing parallelism while alleviating DRAM bandwidth bottlenecks. Our simulator enables design space exploration of SLMs up to 4B parameters, identifying Pareto-optimal configurations in terms of latency and energy. Compared to an NVIDIA Orin Nano, EdgeCIM achieves up to 7.3x higher throughput and 49.59x better energy efficiency on LLaMA3.2-1B, and delivers 9.95x higher throughput than Qualcomm SA8255P on LLaMA3.2-3B. Extensive benchmarks on TinyLLaMA-1.1B, LLaMA3.2 (1B, 3B), Phi-3.5-mini-3.8B, Qwen2.5 (0.5B, 1.5B, 3B), SmolLM2-1.7B, SmolLM3-3B, and Qwen3 (0.6B, 1.7B, 4B) reveal that our accelerator, under INT4 precision, achieves on average 336.42 tokens/s and 173.02 tokens/J. These results establish EdgeCIM as a compelling solution towards real-time, energy-efficient edge-scale SLM inference. --- ### The Price of Ignorance: Information-Free Quotation for Data Retention in Machine Unlearning Authors: Bin Han, Di Feng, Zexin Fang, Jie Wang, Hans D. Schotten Published: 2026-04-13 Categories: cs.GT, cs.LG arXiv: https://arxiv.org/abs/2604.11511v1 When users exercise data deletion rights under the General Data Protection Regulation (GDPR) and similar regulations, mobile network operators face a tradeoff: excessive machine unlearning degrades model accuracy and incurs retraining costs, yet existing pricing mechanisms for data retention require the server to know every user's private privacy and accuracy preferences, which is infeasible under the very regulations that motivate unlearning. We ask: what is the welfare cost of operating without this private information? We design an information-free ascending quotation mechanism where the server broadcasts progressively higher prices and users self-select their data supply, requiring no knowledge of users' parameters. Under complete information, the protocol admits a unique subgame-perfect Nash equilibrium characterized by single-period selling. We formalize the Price of Ignorance -- the welfare gap between optimal personalized pricing (which knows everything) and our information-free quotation (which knows nothing) -- and prove a three-regime efficiency ordering. Numerical evaluation across seven mechanisms and 5000 Monte Carlo runs shows that this price is near zero: the information-free mechanism achieves >=99% of the welfare of its information-intensive benchmarks, while providing noise-robust guarantees and comparable fairness. --- ### Policy Split: Incentivizing Dual-Mode Exploration in LLM Reinforcement with Dual-Mode Entropy Regularization Authors: Jiashu Yao, Heyan Huang, Daiqing Wu, Zeming Liu, Yuhang Guo Published: 2026-04-13 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2604.11510v2 To encourage diverse exploration in reinforcement learning (RL) for large language models (LLMs) without compromising accuracy, we propose Policy Split, a novel paradigm that bifurcates the policy into normal and high-entropy modes with a high-entropy prompt. While sharing model parameters, the two modes undergo collaborative dual-mode entropy regularization tailored to distinct objectives. Specifically, the normal mode optimizes for task correctness, while the high-entropy mode incorporates a preference for exploration, and the two modes learn collaboratively. Extensive experiments demonstrate that our approach consistently outperforms established entropy-guided RL baselines across various model sizes in general and creative tasks. Further analysis reveals that Policy Split facilitates dual-mode exploration, where the high-entropy mode generates distinct behavioral patterns to the normal mode, providing unique learning signals. --- ### Not All Forgetting Is Equal: Retention Dynamics in Fine-Tuned Image Classifiers Authors: Miit Daga, Swarna Priya Ramu Published: 2026-04-13 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.11508v3 Fine-tuning a pretrained classifier leaves some samples reliably learned and others cycling between correct and incorrect. Curriculum learning, data pruning and dataset cartography assume that pattern is a property of the sample, untested. We record per-sample correctness at every epoch while fine-tuning ResNet-18 and DeiT-Small on an imbalanced retinal OCT dataset and CUB-200-2011, matching samples by image identity and holding the split fixed across seeds. Per-sample retention is reproducible: cross-run Spearman correlation of the fitted decay constant is 0.37 to 0.59 over ten seeds. It is architecture-specific: two runs of one backbone agree more than two backbones on identical data (0.45 and 0.59 within against 0.30 between on OCTDL). Loss after five frozen-backbone epochs predicts a different run's decay constant at 0.29 to 0.43. The Ebbinghaus exponential does not survive: monotone decay, the one shape it can represent, is 0.1% to 0.8% of samples, and on traces that do forget mean R-squared is negative in all four configurations. A power law and a free-asymptote variant fail on the same traces: the defect is monotonicity. Across five sampling arms with matched exposure, prioritisation ratios of 2.7x to 28x, and an online variant, none of 48 comparisons against uniform sampling survives Benjamini-Hochberg correction, though three seeds detect only about four accuracy points. A stable, cheap difficulty score does not buy generalisation through sampling. Patient-grouped splitting, the remedy for a leak reaching 76% to 78% of OCT test images, moves that dataset's headline metrics by less than their run-to-run spread. --- ### Not All Forgetting Is Equal: Architecture-Dependent Retention Dynamics in Fine-Tuned Image Classifiers Authors: Miit Daga, Swarna Priya Ramu Published: 2026-04-13 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.11508v2 Fine-tuning pretrained image classifiers is standard practice, yet which individual samples are forgotten during this process, and whether forgetting patterns are stable or architecture dependent, remains unclear. Understanding these dynamics has direct implications for curriculum design, data pruning, and ensemble construction. We track per-sample correctness at every epoch during fine-tuning of ResNet-18 and DeiT-Small on a retinal OCT dataset (7 classes, 56:1 imbalance) and CUB-200-2011 (200 bird species), fitting Ebbinghaus-style exponential decay curves to each sample's retention trace. Five findings emerge. First, the two architectures forget fundamentally different samples: Jaccard overlap of the top 10 percent most-forgotten is 0.34 on OCTDL and 0.15 on CUB-200. Second, ViT forgetting is more structured (mean $R^2 = 0.74$) than CNN forgetting ($R^2 = 0.52$). Third, per-sample forgetting is stochastic across random seeds (Spearman $ρ\approx 0.01$), challenging the assumption that sample difficulty is an intrinsic property. Fourth, class-level forgetting is consistent and semantically interpretable: visually similar species are forgotten most, distinctive ones least. Fifth, a sample's loss after head warmup predicts its long-term decay constant ($ρ= 0.30$ to $0.50$, $p < 10^{-45}$). These findings suggest that architectural diversity in ensembles provides complementary retention coverage, and that curriculum or pruning methods based on per-sample difficulty may not generalize across runs. A spaced repetition sampler built on these decay constants does not outperform random sampling, indicating that static scheduling cannot exploit unstable per-sample signals. --- ### Deep Learning for Sequential Decision Making under Uncertainty: Foundations, Frameworks, and Frontiers Authors: I. Esra Buyuktahtakin Published: 2026-04-13 Categories: math.OC, cs.AI, cs.LG, eess.SY, stat.ML arXiv: https://arxiv.org/abs/2604.11507v2 Artificial intelligence (AI) is moving increasingly beyond prediction to support decisions in complex, uncertain, and dynamic environments. This shift creates a natural intersection with operations research and management science (OR/MS), which has long provided methodological foundations for sequential decision making under uncertainty. At the same time, deep learning advances, including feedforward neural networks, recurrent architectures, transformers, large language models (LLMs), and deep reinforcement learning, have expanded data-driven modeling for large-scale decisions. This tutorial presents an OR/MS-centered perspective on deep learning for sequential decision making under uncertainty, bridging neural architectures and OR/MS approaches to decision making. Its premise: deep learning complements optimization rather than replacing it. Deep learning brings adaptability and scalable approximation, whereas OR/MS provides the mathematical rigor to represent constraints, recourse, uncertainty, and decision quality. The tutorial reviews key decision making foundations, connects them to the major neural architectures in modern AI, and organizes the field around three central themes: predict-then-optimize and decision-aware learning, learning-based decision generation under constraints for continuous and discrete problems with temporal coupling, and deep reinforcement learning for sequential and combinatorial decision making. Impact spans supply chains, service systems, healthcare and epidemic response, agriculture, energy, environmental sustainability, and autonomous operations. This tutorial frames these developments as part of a shift from predictive AI toward decision-capable AI, highlighting OR/MS's role in shaping the next generation of integrated learning--optimization systems. --- ### Deep Learning for Sequential Decision Making under Uncertainty: Foundations, Frameworks, and Frontiers Authors: I. Esra Buyuktahtakin Published: 2026-04-13 Categories: math.OC, cs.AI, cs.LG, eess.SY, stat.ML arXiv: https://arxiv.org/abs/2604.11507v1 Artificial intelligence (AI) is moving increasingly beyond prediction to support decisions in complex, uncertain, and dynamic environments. This shift creates a natural intersection with operations research and management sciences (OR/MS), which have long offered conceptual and methodological foundations for sequential decision-making under uncertainty. At the same time, recent advances in deep learning, including feedforward neural networks, LSTMs, transformers, and deep reinforcement learning, have expanded the scope of data-driven modeling and opened new possibilities for large-scale decision systems. This tutorial presents an OR/MS-centered perspective on deep learning for sequential decision-making under uncertainty. Its central premise is that deep learning is valuable not as a replacement for optimization, but as a complement to it. Deep learning brings adaptability and scalable approximation, whereas OR/MS provides the structural rigor needed to represent constraints, recourse, and uncertainty. The tutorial reviews key decision-making foundations, connects them to the major neural architectures in modern AI, and discusses leading approaches to integrating learning and optimization. It also highlights emerging impact in domains such as supply chains, healthcare and epidemic response, agriculture, energy, and autonomous operations. More broadly, it frames these developments as part of a wider transition from predictive AI toward decision-capable AI and highlights the role of OR/MS in shaping the next generation of integrated learning--optimization systems. --- ### Lectures on AI for Mathematics Authors: Xiaoyang Chen, Xiaoyang Chen Published: 2026-04-13 Categories: cs.AI, math.AP, math.AT, math.DG arXiv: https://arxiv.org/abs/2604.11504v1 This book provides a comprehensive and accessible introduction to the emerging field of AI for mathematics. It covers the core principles and diverse applications of using artificial intelligence to advance mathematical research. Through clear explanations, the text explores how AI can discover hidden mathematical patterns, assist in proving complicated theorems, and even construct counterexamples to challenge conjectures. --- ### METER: Evaluating Multi-Level Contextual Causal Reasoning in Large Language Models Authors: Pengfeng Li, Chen Huang, Chaoqun Hao, Hongyao Chen, Xiao-Yong Wei, Wenqiang Lei, See-Kiong Ng Published: 2026-04-13 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.11502v2 Contextual causal reasoning is a critical yet challenging capability for Large Language Models (LLMs). Existing benchmarks, however, often evaluate this skill in fragmented settings, failing to ensure context consistency or cover the full causal hierarchy. To address this, we pioneer METER to systematically benchmark LLMs across all three levels of the causal ladder under a unified context setting. Our extensive evaluation of various LLMs reveals a significant decline in proficiency as tasks ascend the causal hierarchy. To diagnose this degradation, we conduct a deep mechanistic analysis via both error pattern identification and internal information flow tracing. Our analysis reveals two primary failure modes: (1) LLMs are susceptible to distraction by causally irrelevant but factually correct information at lower level of causality; and (2) as tasks ascend the causal hierarchy, faithfulness to the provided context degrades, leading to a reduced performance. We belive our work advances our understanding of the mechanisms behind LLM contextual causal reasoning and establishes a critical foundation for future research. Our code and dataset are available at https://github.com/SCUNLP/METER . --- ### Attention-Weighted Value Projection for KV-Cache Compression Authors: Samuel Salfati Published: 2026-04-13 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2604.11501v2 Rank reduction discards dimensions; quantization keeps them at lower precision. Comparing the two requires a choice of what compression should preserve. For attention values, we study reconstruction of the attention output rather than reconstruction of the values alone. With fixed attention weights, the optimal orthogonal rank-$r$ projection uses the leading eigenvectors of $V^\topα^\topαV$, and its error is exactly the discarded eigenvalue sum. We extend this objective to calibration datasets and grouped query attention, and describe rank allocation under an additive local error budget. We also examine the limits of using local error to predict downstream loss. Historical weight-perturbation experiments favor coefficient rounding over the value projections tested, but do not establish a comparison at equal cache storage. The resulting distinction is practical: the projection objective has an exact solution, while a comparison with cache quantization requires separate activation-level experiments. --- ### Quantization Dominates Rank Reduction for KV-Cache Compression Authors: Samuel Salfati Published: 2026-04-13 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2604.11501v1 We compare two strategies for compressing the KV cache in transformer inference: rank reduction (discard dimensions) and quantization (keep all dimensions, reduce precision). At matched storage budgets across five models (124M-14B, MHA and GQA), we find that quantization consistently outperforms rank reduction by 4-364 PPL depending on model and compression level. The gap persists even when rank reduction is combined with quantization in hybrid baselines, and it grows with GQA aggressiveness. On LAMBADA, INT4 matches FP16 accuracy (+0.23 PPL on Mistral 7B, +0.58 on GPT-2) while rank-32 at identical storage collapses to 0.4%. We trace this gap to a structural asymmetry: under softmax attention routing, removing a dimension can flip which token is attended (a discrete failure), while quantization noise is bounded and typically preserves score ordering. We formalize this via a perturbation result showing projection damage exceeds quantization damage by 3 x 2^(2b) per direction under the softmax Fisher metric. A basis ablation confirms the finding is basis-independent (spread <0.4 PPL), establishing that the advantage comes from preserving dimensions, not from a better coordinate system. Joint K+V INT4 quantization achieves 75% total KV reduction at only +0.18 PPL on Mistral 7B. ---