# Research Papers — 2026-04 · bundle page 230 of 233

> Full-text markdown of papers #11451–11500 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/230
[← Previous bundle](https://aipod.dev/papers/archive/2026-04/page/229.md)
[Next bundle →](https://aipod.dev/papers/archive/2026-04/page/231.md)

### HICT: High-precision 3D CBCT reconstruction from a single X-ray

Authors: Wen Ma, Jiaxiang Liu, Zikai Xiao, Ziyang Wang, Feng Yang, Zuozhu Liu
Published: 2026-04-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2604.00792v1

Accurate 3D dental imaging is vital for diagnosis and treatment planning, yet CBCT's high radiation dose and cost limit its accessibility. Reconstructing 3D volumes from a single low-dose panoramic X-ray is a promising alternative but remains challenging due to geometric inconsistencies and limited accuracy. We propose HiCT, a two-stage framework that first generates geometrically consistent multi-view projections from a single panoramic image using a video diffusion model, and then reconstructs high-fidelity CBCT from the projections using a ray-based dynamic attention network and an X-ray sampling strategy. To support this, we built XCT, a large-scale dataset combining public CBCT data with 500 paired PX-CBCT cases. Extensive experiments show that HiCT achieves state-of-the-art performance, delivering accurate and geometrically consistent reconstructions for clinical use.

---

### RefineRL: Advancing Competitive Programming with Self-Refinement Reinforcement Learning

Authors: Shaopeng Fu, Xingxing Zhang, Li Dong, Furu Wei
Published: 2026-04-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2604.00790v2

While large language models (LLMs) have demonstrated strong performance on complex reasoning tasks such as competitive programming (CP), existing methods predominantly focus on single-attempt settings, overlooking their capacity for iterative refinement. In this paper, we present RefineRL, a novel approach designed to unleash the self-refinement capabilities of LLMs for CP problem solving. RefineRL introduces two key innovations: (1) Skeptical-Agent, an iterative self-refinement agent equipped with local execution tools to validate generated solutions against public test cases of CP problems. This agent always remains skeptical of its own outputs and thereby enforces rigorous self-refinement even when validation suggests correctness. (2) A reinforcement learning (RL) solution to incentivize LLMs to self-refine with only standard RLVR data (i.e., problems paired with their verifiable answers). Extensive experiments on Qwen3-4B and Qwen3-4B-2507 demonstrate that our method yields substantial gains: after our RL training, these compact 4B models integrated with the Skeptical-Agent not only outperform much larger 32B models but also approach the single-attempt performance of 235B models. Notably, although our models are trained solely on CP coding data, RefineRL also substantially improves their performance on mathematical reasoning benchmarks, suggesting that the learned self-refinement behavior can transfer beyond coding-specific tasks. These findings highlight self-refinement as a promising direction for scaling LLMs' general reasoning ability.

---

### RefineRL: Advancing Competitive Programming with Self-Refinement Reinforcement Learning

Authors: Shaopeng Fu, Xingxing Zhang, Li Dong, Di Wang, Furu Wei
Published: 2026-04-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2604.00790v1

While large language models (LLMs) have demonstrated strong performance on complex reasoning tasks such as competitive programming (CP), existing methods predominantly focus on single-attempt settings, overlooking their capacity for iterative refinement. In this paper, we present RefineRL, a novel approach designed to unleash the self-refinement capabilities of LLMs for CP problem solving. RefineRL introduces two key innovations: (1) Skeptical-Agent, an iterative self-refinement agent equipped with local execution tools to validate generated solutions against public test cases of CP problems. This agent always maintains a skeptical attitude towards its own outputs and thereby enforces rigorous self-refinement even when validation suggests correctness. (2) A reinforcement learning (RL) solution to incentivize LLMs to self-refine with only standard RLVR data (i.e., problems paired with their verifiable answers). Extensive experiments on Qwen3-4B and Qwen3-4B-2507 demonstrate that our method yields substantial gains: after our RL training, these compact 4B models integrated with the Skeptical-Agent not only outperform much larger 32B models but also approach the single-attempt performance of 235B models. These findings suggest that self-refinement holds considerable promise for scaling LLM reasoning, with significant potential for further advancement.

---

### Valency Classification of Mapudungun Verbal Roots. Established by the language's own morphotactics

Authors: Andrés Chandía
Published: 2026-04-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2604.00789v3

In the previous work, a lexical (re)categorisation -- or confirmation of the given category -- of roots identified as verbal was undertaken to determine their original category accurately. Building on this, the present paper offers an account of the valency classification of those Mapudungun roots confirmed to be verbal, using the language's own morphotactics; specifically, by examining the permissible and restricted combinations of various suffixes with roots or verbal stems in the Mapuche verb form. As with all work conducted thus far, the results presented here aim to improve the morphological analyser (Dungupeyum) with all verified findings incorporated into the system. From a theoretical perspective, we also hope to contribute to the recognition and understanding of issues related to the valency of Mapuche verb forms.

---

### UK AISI Alignment Evaluation Case-Study

Authors: Alexandra Souly, Robert Kirk, Jacob Merizian, Abby D'Cruz, Xander Davies
Published: 2026-04-01
Categories: cs.AI, cs.CR
arXiv: https://arxiv.org/abs/2604.00788v1

This technical report presents methods developed by the UK AI Security Institute for assessing whether advanced AI systems reliably follow intended goals. Specifically, we evaluate whether frontier models sabotage safety research when deployed as coding assistants within an AI lab. Applying our methods to four frontier models, we find no confirmed instances of research sabotage. However, we observe that Claude Opus 4.5 Preview (a pre-release snapshot of Opus 4.5) and Sonnet 4.5 frequently refuse to engage with safety-relevant research tasks, citing concerns about research direction, involvement in self-training, and research scope. We additionally find that Opus 4.5 Preview shows reduced unprompted evaluation awareness compared to Sonnet 4.5, while both models can distinguish evaluation from deployment scenarios when prompted. Our evaluation framework builds on Petri, an open-source LLM auditing tool, with a custom scaffold designed to simulate realistic internal deployment of a coding agent. We validate that this scaffold produces trajectories that all tested models fail to reliably distinguish from real deployment data. We test models across scenarios varying in research motivation, activity type, replacement threat, and model autonomy. Finally, we discuss limitations including scenario coverage and evaluation awareness.

---

### Finding Low Star Discrepancy 3D Kronecker Point Sets Using Algorithm Configuration Techniques

Authors: Imène Ait Abderrahim, Carola Doerr, Martin Durand
Published: 2026-04-01
Categories: cs.NE
arXiv: https://arxiv.org/abs/2604.00786v2

The L infinity star discrepancy is a measure for how uniformly a point set is distributed in a given space. Point sets of low star discrepancy are used as designs of experiments, as initial designs for Bayesian optimization algorithms, for quasi-Monte Carlo integration methods, and many other applications. Recent work has shown that classical constructions such as Sobol', Halton, or Hammersley sequences can be outperformed by large margins when considering point sets of fixed sizes rather than their convergence behavior. These results, highly relevant to the aforementioned applications, raise the question of how much existing constructions can be improved through size-specific optimization. In this work, we study this question for the so-called Kronecker construction. Focusing on the 3-dimensional setting, we show that optimizing the two configurable parameters of its construction yields point sets outperforming the state-of-the-art value for sets of at least 500 points. Using the algorithm configuration technique irace, we then derive parameters that yield new state-of-the-art discrepancy values for whole ranges of set sizes.

---

### Scalable Pretraining of Large Mixture of Experts Language Models on Aurora Super Computer

Authors: Dharma Teja Vooturi, Dhiraj Kalamkar, Dipankar Das, Bharat Kaul
Published: 2026-04-01
Categories: cs.LG, cs.AI, cs.DC
arXiv: https://arxiv.org/abs/2604.00785v1

Pretraining Large Language Models (LLMs) from scratch requires massive amount of compute. Aurora super computer is an ExaScale machine with 127,488 Intel PVC (Ponte Vechio) GPU tiles. In this work, we showcase LLM pretraining on Aurora at the scale of 1000s of GPU tiles. Towards this effort, we developed Optimus, an inhouse training library with support for standard large model training techniques. Using Optimus, we first pretrained Mula-1B, a 1 Billion dense model and Mula-7B-A1B, a 7 Billion Mixture of Experts (MoE) model from scratch on 3072 GPU tiles for the full 4 trillion tokens of the OLMoE-mix-0924 dataset. We then demonstrated model scaling by pretraining three large MoE models Mula-20B-A2B, Mula-100B-A7B, and Mula-220B-A10B till 100 Billion tokens on the same dataset. On our largest model Mula-220B-A10B, we pushed the compute scaling from 384 to 12288 GPU tiles and observed scaling efficiency of around 90% at 12288 GPU tiles. We significantly improved the runtime performance of MoE models using custom GPU kernels for expert computation, and a novel EP-Aware sharded optimizer resulting in training speedups up to 1.71x. As part of the Optimus library, we also developed a robust set of reliability and fault tolerant features to improve training stability and continuity at scale.

---

### An Approach to Enriching Surgical Video Datasets for Fine-Grained Spatial-Temporal Understanding of Vision-Language Models

Authors: Lennart Maack, Alexander Schlaefer
Published: 2026-04-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2604.00784v2

Surgical video understanding is a crucial prerequisite for advancing Computer-Assisted Surgery. While vision-language models (VLMs) have recently been applied to the surgical domain, existing surgical vision-language datasets lack in capturing and evaluating complex, interleaved spatial-temporal dynamics. Creating large scale datasets that accurately represent fine-grained spatial-temporal relationships in surgical videos is challenging due to costly manual annotations or error-prone generation using large language models. To address this gap, we introduce the SurgSTU-Pipeline, a deterministic generation pipeline featuring temporal and spatial continuity filtering to reliably create surgical datasets for fine-grained spatial-temporal multimodal understanding. Applying this pipeline to publicly available surgical datasets, we create the SurgSTU dataset, comprising 6711 video clips densely extended with 150k fine-grained spatial-temporal question-answer samples. Our comprehensive evaluation shows that while state-of-the-art generalist VLMs struggle in zero-shot settings, their spatial-temporal capabilities can be improved through in-context learning. A fine-tuned VLM on the SurgSTU training dataset achieves highest performance among all spatial-temporal tasks, validating the dataset's efficacy to improve spatial-temporal understanding of VLMs in surgical videos. The project is available here: https://lennart-maack.github.io/SurgSTU-project

---

### Using predefined vector systems to speed up neural network multimillion class classification

Authors: Nikita Gabdullin, Ilya Androsov
Published: 2026-04-01
Categories: cs.LG, cs.CV
arXiv: https://arxiv.org/abs/2604.00779v2

Label prediction in neural networks (NNs) has O(n) complexity proportional to the number of classes. This holds true for classification using fully connected layers and cosine similarity with some set of class prototypes. In this paper we show that if NN latent space (LS) geometry is known and possesses specific properties, label prediction complexity can be significantly reduced. This is achieved by associating label prediction with the O(1) complexity closest cluster center search in a vector system used as target for latent space configuration (LSC). The proposed method only requires finding indexes of several largest and lowest values in the embedding vector making it extremely computationally efficient. We show that the proposed method does not change NN training accuracy computational results. We also measure the time required by different computational stages of NN inference and label prediction on multiple datasets. The experiments show that the proposed method allows to achieve up to 11.6 times overall acceleration over conventional methods. Furthermore, the proposed method has unique properties which allow to predict the existence of new classes.

---

### From Early Encoding to Late Suppression: Interpreting LLMs on Character Counting Tasks

Authors: Ayan Datta, Mounika Marreddy, Alexander Mehler, Zhixue Zhao, Radhika Mamidi
Published: 2026-04-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2604.00778v1

Large language models (LLMs) exhibit failures on elementary symbolic tasks such as character counting in a word, despite excelling on complex benchmarks. Although this limitation has been noted, the internal reasons remain unclear. We use character counting (e.g., "How many p's are in apple?") as a minimal, controlled probe that isolates token-level reasoning from higher-level confounds. Using this setting, we uncover a consistent phenomenon across modern architectures, including LLaMA, Qwen, and Gemma: models often compute the correct answer internally yet fail to express it at the output layer. Through mechanistic analysis combining probing classifiers, activation patching, logit lens analysis, and attention head tracing, we show that character-level information is encoded in early and mid-layer representations. However, this information is attenuated by a small set of components in later layers, especially the penultimate and final layer MLP. We identify these components as negative circuits: subnetworks that downweight correct signals in favor of higher-probability but incorrect outputs. Our results lead to two contributions. First, we show that symbolic reasoning failures in LLMs are not due to missing representations or insufficient scale, but arise from structured interference within the model's computation graph. This explains why such errors persist and can worsen under scaling and instruction tuning. Second, we provide evidence that LLM forward passes implement a form of competitive decoding, in which correct and incorrect hypotheses coexist and are dynamically reweighted, with final outputs determined by suppression as much as by amplification. These findings carry implications for interpretability and robustness: simple symbolic reasoning exposes weaknesses in modern LLMs, underscoring need for design strategies that ensure information is encoded and reliably used.

---

### From Baselines to Preferences: A Comparative Study of LoRA/QLoRA and Preference Optimization for Mental Health Text Classification

Authors: Mihael Arcan
Published: 2026-04-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2604.00773v1

Mental health text classification has rapidly adopted modern adaptation methods, yet practical guidance on which optimization strategy to use, when, and why remains limited. This paper presents a systematic comparative study of optimization pathways for a joint mental-health classification task, moving from strong vanilla baselines to progressively more specialized techniques. We first establish classical and encoder references, then examine parameter-efficient supervised fine-tuning with LoRA/QLoRA under multiple objective and optimization settings, and finally evaluate preference-based optimization with DPO, ORPO, and KTO, including class-rebalanced training. Rather than emphasizing a single headline score, we focus on methodological insight: how performance changes with objective formulation, adapter choice, optimizer behavior, context windowing, and class-balance intervention. The results show that optimization effects are highly method-dependent: some approaches deliver stable, transferable gains, while others are sensitive to configuration and data balance. Preference optimization, in particular, exhibits large variation across objectives, indicating that method selection is more consequential than simply adding a preference-training stage. The central contribution is a clear optimization narrative for mental health NLP: start from transparent baselines, apply controlled tuning, and use preference optimization selectively where its gains are demonstrable. This provides a reproducible and practically grounded framework for choosing effective training strategies beyond architecture choice alone.

---

### Thinking Wrong in Silence: Backdoor Attacks on Continuous Latent Reasoning

Authors: Swapnil Parekh
Published: 2026-04-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2604.00770v1

A new generation of language models reasons entirely in continuous hidden states, producing no tokens and leaving no audit trail. We show that this silence creates a fundamentally new attack surface. ThoughtSteer perturbs a single embedding vector at the input layer; the model's own multi-pass reasoning amplifies this perturbation into a hijacked latent trajectory that reliably produces the attacker's chosen answer, while remaining structurally invisible to every token-level defense. Across two architectures (Coconut and SimCoT), three reasoning benchmarks, and model scales from 124M to 3B parameters, ThoughtSteer achieves >=99% attack success rate with near-baseline clean accuracy, transfers to held-out benchmarks without retraining (94-100%), evades all five evaluated active defenses, and survives 25 epochs of clean fine-tuning. We trace these results to a unifying mechanism: Neural Collapse in the latent space pulls triggered representations onto a tight geometric attractor, explaining both why defenses fail and why any effective backdoor must leave a linearly separable signature (probe AUC>=0.999). Yet a striking paradox emerges: individual latent vectors still encode the correct answer even as the model outputs the wrong one. The adversarial information is not in any single vector but in the collective trajectory, establishing backdoor perturbations as a new lens for mechanistic interpretability of continuous reasoning. Code and checkpoints are available.

---

### ActivityNarrated: An Open-Ended Narrative Paradigm for Wearable Human Activity Understanding

Authors: Lala Shakti Swarup Ray, Mengxi Liu, Alcina Pinto, Deepika Gurung, Daniel Geissler, Paul Lukowoicz, Bo Zhou
Published: 2026-04-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2604.00767v2

Wearable human activity recognition (HAR) has made steady progress, yet much of this progress remains grounded in fixed-window, closed-set classification benchmarks. This formulation is poorly matched to everyday behavior, where activities are open-ended, unscripted, personalized, variable in duration, and often compositional. To address this mismatch, we introduce ActivityNarrated, an open-ended narrative paradigm for language-grounded wearable activity understanding. We formulate this setting as dense sensor signal captioning with a comprehensive benchmark protocol that measures temporal localization, caption quality, sensor-language alignment, conventional closed-set classification as a downstream diagnostic, and additional robustness measures. We further present ActNarrator, a 3-stage architecture that discretizes continuous IMU signals into reusable motion tokens and uses an external frozen small language model to generate open-vocabulary activity captions. Experiments show that our method provides high quality dense sensor captioning with superior adaptivity and robustness, enabling various downstream tasks by turning sensor-based human activity understanding into sensor-grounded text-level reasoning. This includes downstream classification where ActNarrator outperforms state-of-the-art HAR models by 3.8 - 31.6 \% in Macro-F1. This paradigm also enables novel activity understanding capabilities such as complex question-answering over long time horizons.

---

### PrivHAR-Bench: A Graduated Privacy Benchmark Dataset for Video-Based Action Recognition

Authors: Samar Ansari
Published: 2026-04-01
Categories: cs.CV, cs.CR
arXiv: https://arxiv.org/abs/2604.00761v1

Existing research on privacy-preserving Human Activity Recognition (HAR) typically evaluates methods against a binary paradigm: clear video versus a single privacy transformation. This limits cross-method comparability and obscures the nuanced relationship between privacy strength and recognition utility. We introduce \textit{PrivHAR-Bench}, a multi-tier benchmark dataset designed to standardize the evaluation of the \textit{Privacy-Utility Trade-off} in video-based action recognition. PrivHAR-Bench applies a graduated spectrum of visual privacy transformations: from lightweight spatial obfuscation to cryptographic block permutation, to a curated subset of 15 activity classes selected for human articulation diversity. Each of the 1,932 source videos is distributed across 9 parallel tiers of increasing privacy strength, with additional background-removed variants to isolate the contribution of human motion features from contextual scene bias. We provide lossless frame sequences, per-frame bounding boxes, estimated pose keypoints with joint-level confidence scores, standardized group-based train/test splits, and an evaluation toolkit computing recognition accuracy and privacy metrics. Empirical validation using R3D-18 demonstrates a measurable and interpretable degradation curve across tiers, with within-tier accuracy declining from 88.8\% (clear) to 53.5\% (encrypted, background-removed) and cross-domain accuracy collapsing to 4.8\%, establishing PrivHAR-Bench as a controlled benchmark for comparing privacy-preserving HAR methods under standardized conditions. The dataset, generation pipeline, and evaluation code are publicly available.

---

### IWP: Token Pruning as Implicit Weight Pruning in Large Vision Language Models

Authors: Dong-Jae Lee, Sunghyun Baek, Junmo Kim
Published: 2026-04-01
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2604.00757v3

Large Vision Language Models show impressive performance across image and video understanding tasks, yet their computational cost grows rapidly with the number of visual tokens. Existing token pruning methods mitigate this issue through empirical approaches while overlooking the internal mechanism of attention. In this paper, we propose a novel training free token pruning framework grounded in the dual form perspective of attention. We reformulate attention as an implicit linear layer whose weight matrix is the sum of rank 1 outer products, each generated by a single token's key value pair. Token pruning thus reduces to selecting an optimal subset of these rank 1 updates that best approximates the original dual weight matrix. Extending this perspective to standard softmax attention in LVLMs, we derive a novel metric quantifying both a token's information magnitude and information duplication. To efficiently select the subset with the proposed metric, we introduce Progressive Chunked Maximal Marginal Relevance. Extensive experiments demonstrate that our method achieves a better trade off between performance and efficiency, while providing another perspective on existing pruning approaches.

---

### IWP: Token Pruning as Implicit Weight Pruning in Large Vision Language Models

Authors: Dong-Jae Lee, Sunghyun Baek, Junmo Kim
Published: 2026-04-01
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2604.00757v2

Large Vision Language Models show impressive performance across image and video understanding tasks, yet their computational cost grows rapidly with the number of visual tokens. Existing token pruning methods mitigate this issue through empirical approaches while overlooking the internal mechanism of attention. In this paper, we propose a novel training free token pruning framework grounded in the dual form perspective of attention. We reformulate attention as an implicit linear layer whose weight matrix is the sum of rank 1 outer products, each generated by a single token's key value pair. Token pruning thus reduces to selecting an optimal subset of these rank 1 updates that best approximates the original dual weight matrix. Extending this perspective to standard softmax attention in LVLMs, we derive a novel metric quantifying both a token's information magnitude and information duplication. To efficiently select the subset with the proposed metric, we introduce Progressive Chunked Maximal Marginal Relevance. Extensive experiments demonstrate that our method achieves a better trade off between performance and efficiency, while providing another perspective on existing pruning approaches.

---

### Stochastic Attention: Connectome-Inspired Randomized Routing for Expressive Linear-Time Attention

Authors: Zehao Jin, Yanan Sui
Published: 2026-04-01
Categories: cs.CL, cs.LG
arXiv: https://arxiv.org/abs/2604.00754v2

The whole-brain connectome of a fruit fly comprises over 130K neurons connected with a probability of merely 0.02%, yet achieves an average shortest path of only 4.4 hops. Despite being highly structured at the circuit level, the network's long-range connections are broadly distributed across brain regions, functioning as stochastic shortcuts that enable efficient global communication. Inspired by this observation, we propose Stochastic Attention (SA), a drop-in enhancement for sliding-window attention (SWA) that applies a random permutation to the token sequence before windowed attention and restores the original order afterward. This transforms the fixed local window into a stochastic global one within the same $O(nw)$ per-layer budget. Through depth, independently sampled permutations yield exponentially growing receptive fields, achieving full sequence coverage in $O(\log_w n)$ layers versus $O(n/w)$ for SWA. We validate SA in two settings: pre-training language models from scratch, where a gated SA + SWA combination achieves the best average zero-shot accuracy, and training-free inference on Qwen3-8B and Qwen3-30B-A3B, where SA consistently outperforms SWA and matches or exceeds Mixture of Block Attention at comparable compute budgets. These results suggest that connectome-inspired stochastic routing is a practical primitive for improving the expressivity of efficient attention, complementary to existing linear and sparse approaches.

---

### BioCOMPASS: Integrating Biomarkers into Transformer-Based Immunotherapy Response Prediction

Authors: Sayed Hashim, Frank Soboczenski, Paul Cairns
Published: 2026-04-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2604.00739v1

Datasets used in immunotherapy response prediction are typically small in size, as well as diverse in cancer type, drug administered, and sequencer used. Models often drop in performance when tested on patient cohorts that are not included in the training process. Recent work has shown that transformer-based models along with self-supervised learning show better generalisation performance than threshold-based biomarkers, but is still suboptimal. We present BioCOMPASS, an extension of a transformer-based model called COMPASS, that integrates biomarkers and treatment information to further improve its generalisability. Instead of feeding biomarker data as input, we built loss components to align them with the model's intermediate representations. We found that components such as treatment gating and pathway consistency loss improved generalisability when evaluated with Leave-one-cohort-out, Leave-one-cancer-type-out and Leave-one-treatment-out strategies. Results show that building components that exploit biomarker and treatment information can help in generalisability of immunotherapy response prediction. Careful curation of additional components that leverage complementary clinical information and domain knowledge represents a promising direction for future research.

---

### Spectral Compact Training: Pre-Training Large Language Models via Permanent Truncated SVD and Stiefel QR Retraction

Authors: Björn Roman Kohlberger
Published: 2026-04-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2604.00733v2

The memory wall remains the primary bottleneck for training large language models on consumer hardware. We introduce Spectral Compact Training (SCT), a method that replaces dense weight matrices with permanent truncated SVD factors W = U diag(s) V^T, where the full dense matrix is never materialized during training or inference. Gradients flow through the compact spectral factors via standard backpropagation, and U, V are retracted to the Stiefel manifold via QR decomposition after each optimizer step. SCT achieves up to 199x memory reduction per MLP layer at rank 32, enabling full training steps of 70B-parameter architectures on a Steam Deck handheld (7.2 GB peak memory vs. 1,245 GB for dense FP32 training with Adam). Rank-sweep experiments on SmolLM2-1.7B (ranks 32-256, 2000 steps, NVIDIA A100) show that all tested ranks converge to the same loss floor (~4.2-4.5), identifying the learning rate schedule -- not MLP rank -- as the primary bottleneck. Rank 128 emerges as the efficiency sweet spot at 11.7x MLP compression with the lowest perplexity. GPU memory drops 46% at rank 32 while training throughput doubles.

---

### A CEFR-Inspired Classification Framework with Fuzzy C-Means To Automate Assessment of Programming Skills in Scratch

Authors: Ricardo Hidalgo-Aragón, Jesús M. González-Barahona, Gregorio Robles
Published: 2026-04-01
Categories: cs.CY, cs.AI, cs.LG, cs.SE
arXiv: https://arxiv.org/abs/2604.00730v2

Context: Schools, training platforms, and technology firms increasingly need to assess programming proficiency at scale with transparent, reproducible methods that support personalized learning pathways. Objective: This study introduces a pedagogical framework for Scratch project assessment, aligned with the Common European Framework of Reference (CEFR), providing universal competency levels for students and teachers alongside actionable insights for curriculum design. Method: We apply Fuzzy C-Means clustering to 2008246 Scratch projects evaluated via Dr.Scratch, implementing an ordinal criterion to map clusters to CEFR levels (A1-C2), and introducing enhanced classification metrics that identify transitional learners, enable continuous progress tracking, and quantify classification certainty to balance automated feedback with instructor review. Impact: The framework enables diagnosis of systemic curriculum gaps-notably a "B2 bottleneck" where only 13.3% of learners reside due to the cognitive load of integrating Logic Synchronization, and Data Representation--while providing certainty--based triggers for human intervention.

---

### Exploring Silent Data Corruption as a Reliability Challenge in LLM Training

Authors: Anton Altenbernd, Philipp Wiesner, Odej Kao
Published: 2026-04-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2604.00726v1

As Large Language Models (LLMs) scale in size and complexity, the consequences of failures during training become increasingly severe. A major challenge arises from Silent Data Corruption (SDC): hardware-induced faults that bypass system-level detection mechanisms. SDC may behave like benign numerical noise, but can also cause harmful gradient corruption that leads to loss spikes, divergence, or stalled progress. This work provides a controlled study of how intermittent SDC affects LLM pretraining. Using targeted fault injection at the level of GPU matrix-multiply instructions, we characterize the sensitivity of different bit positions, kernel functions, and execution stages. Our analysis shows that locally originating faults can produce impactful corruption, including NaN propagation, short-lived spikes in loss, gradient norm, and attention logits, as well as persistent parameter divergence. Building on the observed corruption signatures, we propose a lightweight detection method that identifies potentially harmful parameter updates. Experiments on LLaMA models with 60M, 350M, and 1.3B parameters demonstrate that recomputing the most recent training step upon detection can effectively mitigate the impact of these events.

---

### A Benchmark of State-Space Models vs. Transformers and BiLSTM-based Models for Historical Newspaper OCR

Authors: Merveilles Agbeti-Messan, Pierrick Tranouez, Stéphane Nicolas, Clément Chatelain, Thierry Paquet
Published: 2026-04-01
Categories: cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2604.00725v2

End-to-end OCR for historical newspapers remains challenging, as models must handle long text sequences, degraded print quality, and complex layouts. While Transformer-based recognizers dominate current research, their quadratic complexity limits efficient paragraph-level transcription and large-scale deployment. We investigate linear-time State-Space Models (SSMs), specifically Mamba, as a scalable alternative to Transformer-based sequence modeling for OCR. We present to our knowledge, the first OCR architecture based on SSMs, combining a CNN visual encoder with bi-directional and autoregressive Mamba sequence modeling, and conduct a large-scale benchmark comparing SSMs with Transformer- and BiLSTM-based recognizers. Multiple decoding strategies (CTC, autoregressive, and non-autoregressive) are evaluated under identical training conditions alongside strong neural baselines (VAN, DAN, DANIEL) and widely used off-the-shelf OCR engines (PERO-OCR, Tesseract OCR, TrOCR, Gemini). Experiments on historical newspapers from the Bibliotheque nationale du Luxembourg, with newly released >99% verified gold-standard annotations, and cross-dataset tests on Fraktur and Antiqua lines, show that all neural models achieve low error rates (~2% CER), making computational efficiency the main differentiator. Mamba-based models maintain competitive accuracy while halving inference time and exhibiting superior memory scaling (1.26x vs 2.30x growth at 1000 chars), reaching 6.07% CER at the severely degraded paragraph level compared to 5.24% for DAN, while remaining 2.05x faster. We release code, trained models, and standardized evaluation protocols to enable reproducible research and guide practitioners in large-scale cultural heritage OCR available at https://github.com/MarcoPerson/ssm-ocr-benchmark.

---

### LangMARL: Natural Language Multi-Agent Reinforcement Learning

Authors: Huaiyuan Yao, Longchao Da, Xiaoou Liu, Charles Fleming, Tianlong Chen, Hua Wei
Published: 2026-04-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2604.00722v1

Large language model (LLM) agents struggle to autonomously evolve coordination strategies in dynamic environments, largely because coarse global outcomes obscure the causal signals needed for local policy refinement. We identify this bottleneck as a multi-agent credit assignment problem, which has long been studied in classical multi-agent reinforcement learning (MARL) but remains underaddressed in LLM-based systems. Building on this observation, we propose LangMARL, a framework that brings credit assignment and policy gradient evolution from cooperative MARL into the language space. LangMARL introduces agent-level language credit assignment, pioneers gradient evolution in language space for policy improvement, and summarizes task-relevant causal relations from replayed trajectories to provide dense feedback and improve convergence under sparse rewards. Extensive experiments across diverse cooperative multi-agent tasks demonstrate improved sample efficiency, interpretability, and strong generalization.

---

### GRASP: Gradient Realignment via Active Shared Perception for Multi-Agent Collaborative Optimization

Authors: Sihan Zhou, Tiantian He, Yifan Lu, Yaqing Hou, Yew-Soon Ong
Published: 2026-04-01
Categories: cs.MA, cs.AI
arXiv: https://arxiv.org/abs/2604.00717v1

Non-stationarity arises from concurrent policy updates and leads to persistent environmental fluctuations. Existing approaches like Centralized Training with Decentralized Execution (CTDE) and sequential update schemes mitigate this issue. However, since the perception of the policies of other agents remains dependent on sampling environmental interaction data, the agent essentially operates in a passive perception state. This inevitably triggers equilibrium oscillations and significantly slows the convergence speed of the system. To address this issue, we propose Gradient Realignment via Active Shared Perception (GRASP), a novel framework that defines generalized Bellman equilibrium as a stable objective for policy evolution. The core mechanism of GRASP involves utilizing the independent gradients of agents to derive a defined consensus gradient, enabling agents to actively perceive policy updates and optimize team collaboration. Theoretically, we leverage the Kakutani Fixed-Point Theorem to prove that the consensus direction $u^*$ guarantees the existence and attainability of this equilibrium. Extensive experiments on StarCraft II Multi-Agent Challenge (SMAC) and Google Research Football (GRF) demonstrate the scalability and promising performance of the framework.

---

### CircuitProbe: Predicting Reasoning Circuits in Transformers via Stability Zone Detection

Authors: Rajkiran Panuganti
Published: 2026-04-01
Categories: cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2604.00716v1

Transformer language models contain localized reasoning circuits, contiguous layer blocks that improve reasoning when duplicated at inference time. Finding these circuits currently requires brute-force sweeps costing 25 GPU hours per model. We propose CircuitProbe, which predicts circuit locations from activation statistics in under 5 minutes on CPU, providing a speedup of three to four orders of magnitude. We find that reasoning circuits come in two types: stability circuits in early layers, detected through the derivative of representation change, and magnitude circuits in late layers, detected through anomaly scoring. We validate across 9 models spanning 6 architectures, including 2025 models, confirming that CircuitProbe top predictions match or are within 2 layers of the optimal circuit in all validated cases. A scaling experiment across the Qwen 2.5 family reveals that layer duplication consistently benefits models under 3B parameters but degrades performance in 7B+ models, making this a practical scaling technique for small language models. CircuitProbe requires as few as 10 calibration examples and its predictions are stable across English, Hindi, Chinese, and French.

---

### To Memorize or to Retrieve: Scaling the Interaction Between Pretraining and Retrieval

Authors: Karan Singh, Michael Yu, Varun Gangal, Zhuofu Tao, Sachin Kumar, Emmy Liu, Steven Y. Feng
Published: 2026-04-01
Categories: cs.CL, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2604.00715v2

Retrieval-augmented generation (RAG) improves language model (LM) performance by providing relevant context at test time for knowledge-intensive situations. In this work, we systematically study the trade-off between pretraining and retrieval by training OLMo-2-based LMs ranging from 30M to 3B parameters on up to 100B DCLM tokens, while varying pretraining data scale, retrieval store size, and retrieval store source (pretraining vs. new data) across reasoning, scientific QA, and open-domain QA benchmarks. We find that retrieval gains depend on model capacity and pretraining exposure and are strongly front-loaded, with a median 91% of the largest observed improvement realized by one retrieval token per model parameter. However, the interaction is objective-dependent: smaller models gain more in gold-answer perplexity, whereas larger, more-pretrained models gain more in accuracy. Retrieval from previously seen data also preserves most of the held-out retrieval gain. Retrieval is therefore a task-, regime-, and metric-dependent complement to parametric learning whose value also depends on datastore size and information novelty. Overall, this motivates the explicit partitioning of data between internalization and external access for LM design.

---

### To Memorize or to Retrieve: Scaling Laws for RAG-Considerate Pretraining

Authors: Karan Singh, Michael Yu, Varun Gangal, Zhuofu Tao, Sachin Kumar, Emmy Liu, Steven Y. Feng
Published: 2026-04-01
Categories: cs.CL, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2604.00715v1

Retrieval-augmented generation (RAG) improves language model (LM) performance by providing relevant context at test time for knowledge-intensive situations. However, the relationship between parametric knowledge acquired during pretraining and non-parametric knowledge accessed via retrieval remains poorly understood, especially under fixed data budgets. In this work, we systematically study the trade-off between pretraining corpus size and retrieval store size across a wide range of model and data scales. We train OLMo-2-based LMs ranging from 30M to 3B parameters on up to 100B tokens of DCLM data, while varying both pretraining data scale (1-150x the number of parameters) and retrieval store size (1-20x), and evaluate performance across a diverse suite of benchmarks spanning reasoning, scientific QA, and open-domain QA. We find that retrieval consistently improves performance over parametric-only baselines across model scales and introduce a three-dimensional scaling framework that models performance as a function of model size, pretraining tokens, and retrieval corpus size. This scaling manifold enables us to estimate optimal allocations of a fixed data budget between pretraining and retrieval, revealing that the marginal utility of retrieval depends strongly on model scale, task type, and the degree of pretraining saturation. Our results provide a quantitative foundation for understanding when and how retrieval should complement pretraining, offering practical guidance for allocating data resources in the design of scalable language modeling systems.

---

### AfrIFact: Cultural Information Retrieval, Evidence Extraction and Fact Checking for African Languages

Authors: Israel Abebe Azime, Jesujoba Oluwadara Alabi, Crystina Zhang, Iffat Maab, Atnafu Lambebo Tonja, Tadesse Destaw Belay, Folasade Peace Alabi, Salomey Osei, Saminu Mohammad Aliyu, Nkechinyere Faith Aguobi, Bontu Fufa Balcha, Blessing Kudzaishe Sibanda et al.
Published: 2026-04-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2604.00706v2

Assessing the veracity of a claim made online is a complex and important task with real-world implications. When these claims are directed at communities with limited access to information and the content concerns issues such as healthcare and culture, the consequences intensify, especially in low-resource languages. In this work, we introduce AfrIFact, a dataset that covers the necessary steps for automatic fact-checking (i.e., information retrieval, evidence extraction, and fact checking), in ten African languages and English. Our evaluation results show that even the best embedding models lack cross-lingual retrieval capabilities, and that cultural and news documents are easier to retrieve than healthcare-domain documents, both in large corpora and in single documents. We show that LLMs lack robust multilingual fact-verification capabilities in African languages, while few-shot prompting improves performance by up to 43% in AfriqueQwen-14B, and task-specific fine-tuning further improves fact-checking accuracy by up to 26%. These findings, along with our release of the AfrIFact dataset, encourage work on low-resource information retrieval, evidence retrieval, and fact checking.

---

### AutoEG: Exploiting Known Third-Party Vulnerabilities in Black-Box Web Applications

Authors: Ruozhao Yang, Mingfei Cheng, Gelei Deng, Junjie Wang, Tianwei Zhang, Xiaofei Xie
Published: 2026-04-01
Categories: cs.CR, cs.AI, cs.SE
arXiv: https://arxiv.org/abs/2604.00704v1

Large-scale web applications are widely deployed with complex third-party components, inheriting security risks arising from component vulnerabilities. Security assessment is therefore required to determine whether such known vulnerabilities remain practically exploitable in real applications. Penetration testing is a widely adopted approach that validates exploitability by launching concrete attacks against known vulnerabilities in real-world black-box systems. However, existing approaches often fail to automatically generate reliable exploits, limiting their effectiveness in practical security assessment. This limitation mainly stems from two issues: (1) precisely triggering vulnerabilities with correct technical details, and (2) adapting exploits to diverse real-world deployment settings. In this paper, we propose AutoEG, a fully automated multi-agent framework for exploit generation targeting black-box web applications. AutoEG has two phases: First, AutoEG extracts precise vulnerability trigger logic from unstructured vulnerability information and encapsulates it into reusable trigger functions. Second, AutoEG uses trigger functions for concrete attack objectives and iteratively refines exploits through feedback-driven interaction with the target application. We evaluate AutoEG on 104 real-world vulnerabilities with 29 attack objectives, resulting in 660 exploitation tasks and 55,440 exploit attempts. AutoEG achieves an average success rate of 82.41%, substantially outperforming state-of-the-art baselines, whose best performance reaches only 32.88%.

---

### G-ICSO-NAS: Shifting Gears between Gradient and Swarm for Robust Neural Architecture Search

Authors: Xingbang Du, Enzhi Zhang, Rui Zhong, Yang Cao, Masaharu Munetomo
Published: 2026-04-01
Categories: cs.NE
arXiv: https://arxiv.org/abs/2604.00703v1

Neural Architecture Search (NAS) has become a pivotal technique in automated machine learning. Evolutionary Algorithm (EA)-based methods demonstrate superior search quality but suffer from prohibitive computational costs, while gradient-based approaches like DARTS offer high efficiency but are prone to premature convergence and performance collapse. To bridge this gap, we propose G-ICSO-NAS, a hybrid framework implementing a three-stage optimization strategy. The Warm-up Phase pre-trains supernet weights ($w$) via differentiable methods while architecture parameters ($α$) remain frozen. The Exploration Phase adopts a hybrid co-optimization mechanism: an Improved Competitive Swarm Optimizer (ICSO) with diversity-aware fitness navigates the architecture space to update $α$, while gradient descent concurrently updates $w$. The Stability Phase employs fine-grained gradient-based search with early stopping to converge to the optimal architecture. By synergizing ICSO's global navigation capability with differentiable methods' efficiency, G-ICSO-NAS achieves remarkable performance with minimal cost. In the context of the DARTS search space, an accuracy of 97.46\% is achieved on CIFAR-10 with a computational budget of just 0.15 GPU-Days. The method also exhibits strong transfer potential, recording accuracies of 83.1\% (CIFAR-100) and 75.02\% (ImageNet). Furthermore, regarding the NAS-Bench-201 benchmark, G-ICSO-NAS is shown to deliver state-of-the-art results across all evaluated datasets.

---

### Learning to Hint for Reinforcement Learning

Authors: Yu Xia, Canwen Xu, Zhewei Yao, Julian McAuley, Yuxiong He
Published: 2026-04-01
Categories: cs.LG, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2604.00698v1

Group Relative Policy Optimization (GRPO) is widely used for reinforcement learning with verifiable rewards, but it often suffers from advantage collapse: when all rollouts in a group receive the same reward, the group yields zero relative advantage and thus no learning signal. For example, if a question is too hard for the reasoner, all sampled rollouts can be incorrect and receive zero reward. Recent work addresses this issue by adding hints or auxiliary scaffolds to such hard questions so that the reasoner produces mixed outcomes and recovers a non-zero update. However, existing hints are usually fixed rather than adapted to the current reasoner, and a hint that creates learning signal under the hinted input does not necessarily improve the no-hint policy used at test time. To this end, we propose Hint Learning for Reinforcement Learning (HiLL), a framework that jointly trains a hinter policy and a reasoner policy during RL. For each hard question, the hinter generates hints online conditioned on the current reasoner's incorrect rollout, allowing hint generation to adapt to the reasoner's evolving errors. We further introduce hint reliance, which measures how strongly correct hinted trajectories depend on the hint. We derive a transferability result showing that lower hint reliance implies stronger transfer from hinted success to no-hint success, and we use this result to define a transfer-weighted reward for training the hinter. Therefore, HiLL favors hints that not only recover informative GRPO groups, but also produce signals that are more likely to improve the original no-hint policy. Experiments across multiple benchmarks show that HiLL consistently outperforms GRPO and prior hint-based baselines, demonstrating the value of adaptive and transfer-aware hint learning for RL. The code is available at https://github.com/Andree-9/HiLL.

---

### Inverse-Free Sparse Variational Gaussian Processes

Authors: Stefano Cortinovis, Laurence Aitchison, Stefanos Eleftheriadis, Mark van der Wilk
Published: 2026-04-01
Categories: stat.ML, cs.LG
arXiv: https://arxiv.org/abs/2604.00697v1

Gaussian processes (GPs) offer appealing properties but are costly to train at scale. Sparse variational GP (SVGP) approximations reduce cost yet still rely on Cholesky decompositions of kernel matrices, ill-suited to low-precision, massively parallel hardware. While one can construct valid variational bounds that rely only on matrix multiplications (matmuls) via an auxiliary matrix parameter, optimising them with off-the-shelf first-order methods is challenging. We make the inverse-free approach practical by proposing a better-conditioned bound and deriving a matmul-only natural-gradient update for the auxiliary parameter, markedly improving stability and convergence. We further provide simple heuristics, such as step-size schedules and stopping criteria, that make the overall optimisation routine fit seamlessly into existing workflows. Across regression and classification benchmarks, we demonstrate that our method 1) serves as a drop-in replacement in SVGP-based models (e.g., deep GPs), 2) recovers similar performance to traditional methods, and 3) can be faster than baselines when well tuned.

---

### TTA-Vid: Generalized Test-Time Adaptation for Video Reasoning

Authors: Soumya Shamarao Jahagirdar, Edson Araujo, Anna Kukleva, M. Jehanzeb Mirza, Saurabhchand Bhati, Samuel Thomas, Brian Kingsbury, Rogerio Feris, James R. Glass, Hilde Kuehne
Published: 2026-04-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2604.00696v1

Recent video reasoning models have shown strong results on temporal and multimodal understanding, yet they depend on large-scale supervised data and multi-stage training pipelines, making them costly to train and difficult to adapt to new domains. In this work, we leverage the paradigm of Test-Time Reinforcement Learning on video-language data to allow for adapting a pretrained model to incoming video samples at test-time without explicit labels. The proposed test-time adaptation for video approach (TTA-Vid) combines two components that work simultaneously: (1) a test-time adaptation that performs step-by-step reasoning at inference time on multiple frame subsets. We then use a batch-aware frequency-based reward computed across different frame subsets as pseudo ground truth to update the model. It shows that the resulting model trained on a single batch or even a single sample from a dataset, is able to generalize at test-time to the whole dataset and even across datasets. Because the adaptation occurs entirely at test time, our method requires no ground-truth annotations or dedicated training splits. Additionally, we propose a multi-armed bandit strategy for adaptive frame selection that learns to prioritize informative frames, guided by the same reward formulation. Our evaluation shows that TTA-Vid yields consistent improvements across various video reasoning tasks and is able to outperform current state-of-the-art methods trained on large-scale data. This highlights the potential of test-time reinforcement learning for temporal multimodal understanding.

---

### Internal APIs Are All You Need: Shadow APIs, Shared Discovery, and the Case Against Browser-First Agent Architectures

Authors: Lewis Tham, Nicholas Mac Gregor Garcia, Jungpil Hahn
Published: 2026-04-01
Categories: cs.ET, cs.AI
arXiv: https://arxiv.org/abs/2604.00694v1

Autonomous agents increasingly interact with the web, yet most websites remain designed for human browsers -- a fundamental mismatch that the emerging ``Agentic Web'' must resolve. Agents must repeatedly browse pages, inspect DOMs, and reverse-engineer callable routes -- a process that is slow, brittle, and redundantly repeated across agents. We observe that every modern website already exposes internal APIs (sometimes called \emph{shadow APIs}) behind its user interface -- first-party endpoints that power the site's own functionality. We present Unbrowse, a shared route graph that transforms browser-based route discovery into a collectively maintained index of these callable first-party interfaces. The system passively learns routes from real browsing traffic and serves cached routes via direct API calls. In a single-host live-web benchmark of equivalent information-retrieval tasks across 94 domains, fully warmed cached execution averaged 950\,ms versus 3{,}404\,ms for Playwright browser automation (3.6$\times$ mean speedup, 5.4$\times$ median), with well-cached routes completing in under 100\,ms. A three-path execution model -- local cache, shared graph, or browser fallback -- ensures the system is voluntary and self-correcting. A three-tier micropayment model via the x402 protocol charges per-query search fees for graph lookups (Tier~3), a one-time install fee for discovery documentation (Tier~1), and optional per-execution fees for site owners who opt in (Tier~2). All tiers are grounded in a necessary condition for rational adoption: an agent uses the shared graph only when the total fee is lower than the expected cost of browser rediscovery.

---

### Performance of Neural and Polynomial Operator Surrogates

Authors: Josephine Westermann, Benno Huber, Thomas O'Leary-Roseberry, Jakob Zech
Published: 2026-04-01
Categories: cs.LG, math.NA
arXiv: https://arxiv.org/abs/2604.00689v1

We consider the problem of constructing surrogate operators for parameter-to-solution maps arising from parametric partial differential equations, where repeated forward model evaluations are computationally expensive. We present a systematic empirical comparison of neural operator surrogates, including a reduced-basis neural operator trained with $L^2_μ$ and $H^1_μ$ objectives and the Fourier neural operator, against polynomial surrogate methods, specifically a reduced-basis sparse-grid surrogate and a reduced-basis tensor-train surrogate. All methods are evaluated on a linear parametric diffusion problem and a nonlinear parametric hyperelasticity problem, using input fields with algebraically decaying spectral coefficients at varying rates of decay $s$. To enable fair comparisons, we analyze ensembles of surrogate models generated by varying hyperparameters and compare the resulting Pareto frontiers of cost versus approximation accuracy, decomposing cost into contributions from data generation, setup, and evaluation. Our results show that no single method is universally superior. Polynomial surrogates achieve substantially better data efficiency for smooth input fields ($s \geq 2$), with convergence rates for the sparse-grid surrogate in agreement with theoretical predictions. For rough inputs ($s \leq 1$), the Fourier neural operator displays the fastest convergence rates. Derivative-informed training consistently improves data efficiency over standard $L^2_μ$ training, providing a competitive alternative for rough inputs in the low-data regime when Jacobian information is available at reasonable cost. These findings highlight the importance of matching the surrogate methodology to the regularity of the problem as well as accuracy demands and computational constraints of the application.

---

### OmniVoice: Towards Omnilingual Zero-Shot Text-to-Speech with Diffusion Language Models

Authors: Han Zhu, Lingxuan Ye, Wei Kang, Zengwei Yao, Liyong Guo, Fangjun Kuang, Zhifeng Han, Weiji Zhuang, Long Lin, Daniel Povey
Published: 2026-04-01
Categories: cs.CL, eess.AS
arXiv: https://arxiv.org/abs/2604.00688v3

We present OmniVoice, a massively multilingual zero-shot text-to-speech (TTS) model that scales to over 600 languages. At its core is a novel diffusion language model-style discrete non-autoregressive (NAR) architecture. Unlike conventional discrete NAR models that suffer from performance bottlenecks in complex two-stage (text-to-semantic-to-acoustic) pipelines, OmniVoice directly maps text to multi-codebook acoustic tokens. This simplified approach is facilitated by two key technical innovations: (1) a full-codebook random masking strategy for efficient training, and (2) initialization from a pre-trained LLM to ensure superior intelligibility. By leveraging a 581k-hour multilingual dataset curated entirely from open-source data, OmniVoice achieves the broadest language coverage to date and delivers state-of-the-art performance across Chinese, English, and diverse multilingual benchmarks. Our code and pre-trained models are publicly available at https://github.com/k2-fsa/OmniVoice.

---

### Full-Gradient Successor Feature Representations

Authors: Ritish Shrirao, Aditya Priyadarshi, Raghuram Bharadwaj Diddigi
Published: 2026-04-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2604.00686v1

Successor Features (SF) combined with Generalized Policy Improvement (GPI) provide a robust framework for transfer learning in Reinforcement Learning (RL) by decoupling environment dynamics from reward functions. However, standard SF learning methods typically rely on semi-gradient Temporal Difference (TD) updates. When combined with non-linear function approximation, semi-gradient methods lack robust convergence guarantees and can lead to instability, particularly in the multi-task setting where accurate feature estimation is critical for effective GPI. Inspired by Full Gradient DQN, we propose Full-Gradient Successor Feature Representations Q-Learning (FG-SFRQL), an algorithm that optimizes the successor features by minimizing the full Mean Squared Bellman Error. Unlike standard approaches, our method computes gradients with respect to parameters in both the online and target networks. We provide a theoretical proof of almost-sure convergence for FG-SFRQL and demonstrate empirically that minimizing the full residual leads to superior sample efficiency and transfer performance compared to semi-gradient baselines in both discrete and continuous domains.

---

### TP-Seg: Task-Prototype Framework for Unified Medical Lesion Segmentation

Authors: Jiawei Xu, Qiangqiang Zhou, Dandan Zhu, Yong Chen, Yugen Yi, Xiaoqi Zhao
Published: 2026-04-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2604.00684v1

Building a unified model with a single set of parameters to efficiently handle diverse types of medical lesion segmentation has become a crucial objective for AI-assisted diagnosis. Existing unified segmentation approaches typically rely on shared encoders across heterogeneous tasks and modalities, which often leads to feature entanglement, gradient interference, and suboptimal lesion discrimination. In this work, we propose TP-Seg, a task-prototype framework for unified medical lesion segmentation. On one hand, the task-conditioned adapter effectively balances shared and task-specific representations through a dual-path expert structure, enabling adaptive feature extraction across diverse medical imaging modalities and lesion types. On the other hand, the prototype-guided task decoder introduces learnable task prototypes as semantic anchors and employs a cross-attention mechanism to achieve fine-grained modeling of task-specific foreground and background semantics. Without bells and whistles, TP-Seg consistently outperforms specialized, general and unified segmentation methods across 8 different medical lesion segmentation tasks covering multiple imaging modalities, demonstrating strong generalization, scalability and clinical applicability.

---

### Convergence of projected stochastic natural gradient variational inference for various step size and sample or batch size schedules

Authors: Thomas Guilmeau, Hadrien Hendrikx, Florence Forbes
Published: 2026-04-01
Categories: stat.ME, math.OC, stat.ML
arXiv: https://arxiv.org/abs/2604.00683v1

Stochastic natural gradient variational inference (NGVI) is a popular and efficient algorithm for Bayesian inference. Despite empirical success, the convergence of this method is still not fully understood. In this work, we define and study a projected stochastic NGVI when variational distributions form an exponential family. Stochasticity arises when either gradients are intractable expectations or large sums. We prove new non-asymptotic convergence results for combinations of constant or decreasing step sizes and constant or increasing sample/batch sizes. When all hyperparameters are fixed, NGVI is shown to converge geometrically to a neighborhood of the optimum, while we establish convergence to the optimum with rates of the form $\mathcal{O}\left(\frac{1}{T^ρ} \right)$, possibly with $ρ\geq 1$, for all other combinations of step size and sample/batch size schedules. These rates apply when the target posterior distribution is close in some sense to the considered exponential family. Our theoretical results extend existing NGVI and stochastic optimization results and provide more flexibility to adjust, in a principled way, step sizes and sample/batch sizes in order to meet speed, resources, or accuracy constraints.

---

### MoonAnything: A Vision Benchmark with Large-Scale Lunar Supervised Data

Authors: Clémentine Grethen, Yuang Shi, Simone Gasparini, Géraldine Morin
Published: 2026-04-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2604.00682v1

Accurate perception of lunar surfaces is critical for modern lunar exploration missions. However, developing robust learning-based perception systems is hindered by the lack of datasets that provide both geometric and photometric supervision. Existing lunar datasets typically lack either geometric ground truth, photometric realism, illumination diversity, or large-scale coverage. In this paper, we introduce MoonAnything, a unified benchmark built on real lunar topography with physically-based rendering, providing the first comprehensive geometric and photometric supervision under diverse illumination with large scale. The benchmark comprises two complementary sub-datasets : i) LunarGeo provides stereo images with corresponding dense depth maps and camera calibration enabling 3D reconstruction and pose estimation; ii) LunarPhoto provides photorealistic images using a spatially-varying BRDF model, along with multi-illumination renderings under real solar configurations, enabling reflectance estimation and illumination-robust perception. Together, these datasets offer over 130K samples with comprehensive supervision. Beyond lunar applications, MoonAnything offers a unique setting and challenging testbed for algorithms under low-textured, high-contrast conditions and applies to other airless celestial bodies and could generalize beyond. We establish baselines using state-of-the-art methods and release the complete dataset along with generation tools to support community extension: https://github.com/clementinegrethen/MoonAnything.

---

### CL-VISTA: Benchmarking Continual Learning in Video Large Language Models

Authors: Haiyang Guo, Yichen Shi, Fei Zhu, Wenzhuo Liu, Hongbo Zhao, Fanhu Zeng, Shijie Ma, Da-Han Wang, Xu-Yao Zhang
Published: 2026-04-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2604.00677v1

Video Large Language Models (Video-LLMs) require continual learning to adapt to non-stationary real-world data. However, existing benchmarks fall short of evaluating modern foundation models: many still rely on models without large-scale pre-training, and prevailing benchmarks typically partition a single dataset into sub-tasks, resulting in high task redundancy and negligible forgetting on pre-trained Video-LLMs. To address these limitations, we propose CL-VISTA, a benchmark tailored for continual video understanding of Video-LLMs. By curating 8 diverse tasks spanning perception, understanding, and reasoning, CL-VISTA induces substantial distribution shifts that effectively expose catastrophic forgetting. To systematically assess CL methods, we establish a comprehensive evaluation framework comprising 6 distinct protocols across 3 critical dimensions: performance, computational efficiency, and memory footprint. Notably, the performance dimension incorporates a general video understanding assessment to assess whether CL methods genuinely enhance foundational intelligence or merely induce task-specific overfitting. Extensive benchmarking of 10 mainstream CL methods reveals a fundamental trade-off: no single approach achieves universal superiority across all dimensions. Methods that successfully mitigate catastrophic forgetting tend to compromise generalization or incur prohibitive computational and memory overheads. We hope CL-VISTA provides critical insights for advancing continual learning in multimodal foundation models.

---

### Procela: Epistemic Governance in Mechanistic Simulations Under Structural Uncertainty

Authors: Kinson Vernet
Published: 2026-04-01
Categories: physics.comp-ph, cs.AI, cs.CE
arXiv: https://arxiv.org/abs/2604.00675v1

Mechanistic simulations typically assume fixed ontologies: variables, causal relationships, and resolution policies are static. This assumption fails when the true causal structure is contested or unidentifiable-as in antimicrobial resistance (AMR) spread, where contact, environmental, and selection ontologies compete. We introduce Procela, a Python framework where variables act as epistemic authorities that maintain complete hypothesis memory, mechanisms encode competing ontologies as causal units, and governance observes epistemic signals and mutates system topology at runtime. This is the first framework where simulations test their own assumptions. We instantiate Procela for AMR in a hospital network with three competing families. Governance detects coverage decay, policy fragility, and runs structural probes. Results show 20.4% error reduction and 69% cumulative regret improvement over baseline. All experiments are reproducible with full auditability. Procela establishes a new paradigm: simulations that model not only the world but their own modeling process, enabling adaptation under structural uncertainty.

---

### Common TF-IDF variants arise as key components in the test statistic of a penalized likelihood-ratio test for word burstiness

Authors: Zeyad Ahmed, Paul Sheridan, Michael McIsaac, Aitazaz A. Farooque
Published: 2026-04-01
Categories: cs.CL, cs.IR, math.ST
arXiv: https://arxiv.org/abs/2604.00672v2

TF-IDF is a classical formula that is widely used for identifying important terms within documents. We show that TF-IDF-like scores arise naturally from the test statistic of a penalized likelihood-ratio test setup capturing word burstiness (also known as word over-dispersion). In our framework, the alternative hypothesis captures word burstiness by modeling a collection of documents according to a family of beta-binomial distributions with a gamma penalty term on the precision parameter. In contrast, the null hypothesis assumes that words are binomially distributed in collection documents, a modeling approach that fails to account for word burstiness. We find that a term-weighting scheme given rise to by this test statistic performs comparably to TF-IDF on document classification tasks. This paper provides insights into TF-IDF from a statistical perspective and underscores the potential of hypothesis testing frameworks for advancing term-weighting scheme development.

---

### Embedded Variational Neural Stochastic Differential Equations for Learning Heterogeneous Dynamics

Authors: Sandeep Kumar Samota, Reema Gupta, Snehashish Chakraverty
Published: 2026-04-01
Categories: cs.LG, math.DS
arXiv: https://arxiv.org/abs/2604.00669v2

This study examines the challenges of modeling complex and noisy data related to socioeconomic factors over time, with a focus on data from various districts in Odisha, India. Traditional time-series models struggle to capture both trends and variations together in this type of data. To tackle this, a Variational Neural Stochastic Differential Equation (V-NSDE) model is designed that combines the expressive dynamics of Neural SDEs with the generative capabilities of Variational Autoencoders (VAEs). This model uses an encoder and a decoder. The encoder takes the initial observations and district embeddings and translates them into a Gaussian distribution, which determines the mean and log-variance of the first latent state. Then the obtained latent state initiates the Neural SDE, which utilize neural networks to determine the drift and diffusion functions that govern continuous-time latent dynamics. These governing functions depend on the time index, latent state, and district embedding, which help the model learn the unique characteristics specific to each district. After that, using a probabilistic decoder, the observations are reconstructed from the latent trajectory. The decoder outputs a mean and log-variance for each time step, which follows the Gaussian likelihood. The Evidence Lower Bound (ELBO) training loss improves by adding a KL-divergence regularization term to the negative log-likelihood (nll). The obtained results demonstrate the effective learning of V-NSDE in recognizing complex patterns over time, yielding realistic outcomes that include clear trends and random fluctuations across different areas.

---

### Embedded Variational Neural Stochastic Differential Equations for Learning Heterogeneous Dynamics

Authors: Sandeep Kumar Samota, Reema Gupta, Snehashish Chakraverty
Published: 2026-04-01
Categories: cs.LG, math.DS
arXiv: https://arxiv.org/abs/2604.00669v1

This study examines the challenges of modeling complex and noisy data related to socioeconomic factors over time, with a focus on data from various districts in Odisha, India. Traditional time-series models struggle to capture both trends and variations together in this type of data. To tackle this, a Variational Neural Stochastic Differential Equation (V-NSDE) model is designed that combines the expressive dynamics of Neural SDEs with the generative capabilities of Variational Autoencoders (VAEs). This model uses an encoder and a decoder. The encoder takes the initial observations and district embeddings and translates them into a Gaussian distribution, which determines the mean and log-variance of the first latent state. Then the obtained latent state initiates the Neural SDE, which utilize neural networks to determine the drift and diffusion functions that govern continuous-time latent dynamics. These governing functions depend on the time index, latent state, and district embedding, which help the model learn the unique characteristics specific to each district. After that, using a probabilistic decoder, the observations are reconstructed from the latent trajectory. The decoder outputs a mean and log-variance for each time step, which follows the Gaussian likelihood. The Evidence Lower Bound (ELBO) training loss improves by adding a KL-divergence regularization term to the negative log-likelihood (nll). The obtained results demonstrate the effective learning of V-NSDE in recognizing complex patterns over time, yielding realistic outcomes that include clear trends and random fluctuations across different areas.

---

### TRIMS: Trajectory-Ranked Instruction Masked Supervision for Diffusion Language Models

Authors: Lingjie Chen, Ruizhong Qiu, Yuyu Fan, Yanjun Zhao, Hanghang Tong
Published: 2026-04-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2604.00666v1

Diffusion language models (DLMs) offer a promising path toward low-latency generation through parallel decoding, but their practical efficiency depends heavily on the decoding trajectory. In practice, this advantage often fails to fully materialize because standard training does not provide explicit supervision over token reveal order, creating a train-inference mismatch that leads to suboptimal decoding behavior. We propose Trajectory-Ranked Instruction Masked Supervision (TRIMS), a simple trajectory-guided supervised fine-tuning framework that injects trajectory supervision into standard Masked Diffusion Language Model (MDLM) training with minimal overhead. Instead of relying on costly DLM-based distillation, TRIMS uses lightweight signals from an autoregressive teacher to guide a trajectory-aware masking strategy, encouraging the model to learn more effective decoding orders. Experiments on LLaDA and Dream across math and coding benchmarks show that TRIMS significantly improves the accuracy-parallelism trade-off over both standard MDLM training and train-free acceleration baselines, while achieving competitive performance with prior distillation-based approaches at substantially lower training cost. Further analysis shows that TRIMS leads to better decoding trajectories, validating the effectiveness of trajectory-guided supervision for DLMs.

---

### Streaming Model Cascades for Semantic SQL

Authors: Paweł Liskowski, Kyle Schmaus
Published: 2026-04-01
Categories: cs.DB, cs.AI
arXiv: https://arxiv.org/abs/2604.00660v2

Modern data warehouses extend SQL with semantic operators that invoke large language models on each qualifying row, making per-row inference orders of magnitude more expensive than traditional SQL. Model cascades reduce this cost by routing most rows through a fast proxy model and delegating uncertain cases to an expensive oracle. Prior SUPG-style cascades, however, require a global proxy-score pass that is itself an LLM-inference workload and blocks output in pipelined query engines. They also target either precision or recall and cannot serve workloads that need both. We formalize the cascade routing problem for streaming semantic SQL with independent parallel workers and present two complementary algorithms within this model. SUPG-IT extends SUPG from single-pass, single-metric estimation to streaming execution by iteratively refining two thresholds as oracle labels accumulate across batches, and is the first streaming cascade with joint probabilistic guarantees on user-specified precision and recall at a chosen failure probability $δ$. GAMCAL replaces user-specified targets with a single tradeoff parameter $α$ between classification error and oracle cost, and learns a monotone Generalized Additive Model that calibrates proxy scores to true-positive probabilities and supplies pointwise uncertainty for stochastic routing. On six classification, filtering, and join benchmarks evaluated in a production semantic SQL engine, both algorithms reach $F_1 \geq 0.95$ at their best operating points. GAMCAL also leads all six datasets at a 20% delegation budget and reaches $F_1 \geq 0.95$ with up to 58% fewer oracle calls than LOTUS's SUPG cascade. SUPG-IT attains the highest best-case $F_1$, with a mean of 0.989 across the six datasets.

---

### Chameleons do not Forget: Prompt-Based Online Continual Learning for Next Activity Prediction

Authors: Marwan Hassani, Tamara Verbeek, Sjoerd van Straten
Published: 2026-04-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2604.00653v2

Predictive process monitoring (PPM) focuses on predicting future process trajectories, including next activity predictions. This is crucial in dynamic environments where processes change or face uncertainty. However, current frameworks often assume a static environment, overlooking dynamic characteristics and concept drifts. This results in catastrophic forgetting, where training while focusing merely on new data distribution negatively impacts the performance on previously learned data distributions. Continual learning addresses, among others, the challenges related to mitigating catastrophic forgetting. This paper proposes a novel approach called Continual Next Activity Prediction with Prompts (CNAPwP), which adapts the DualPrompt algorithm for next activity prediction to improve accuracy and adaptability while mitigating catastrophic forgetting. We introduce new datasets with recurring concept drifts, alongside a task-specific forgetting metric that measures the prediction accuracy gap between initial occurrence and subsequent task occurrences. Extensive testing on three synthetic and two real-world datasets representing several setups of recurrent drifts shows that CNAPwP achieves SOTA or competitive results compared to five baselines, demonstrating its potential applicability in real-world scenarios. An open-source implementation of our method, together with the datasets and results, is available at: https://github.com/SvStraten/CNAPwP.

---

### When AI and Experts Agree on Error: Intrinsic Ambiguity in Dermatoscopic Images

Authors: Loris Cino, Pier Luigi Mazzeo, Alessandro Martella, Giulia Radi, Renato Rossi, Cosimo Distante
Published: 2026-04-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2604.00651v1

The integration of artificial intelligence (AI), particularly Convolutional Neural Networks (CNNs), into dermatological diagnosis demonstrates substantial clinical potential. While existing literature predominantly benchmarks algorithmic performance against human experts, our study adopts a novel perspective by investigating the intrinsic complexity of dermatoscopic images. Through rigorous experimentation with multiple CNN architectures, we isolated a subset of images systematically misclassified across all models-a phenomenon statistically proven to exceed random chance. To determine if these failures stem from algorithmic biases or inherent visual ambiguity, expert dermatologists independently evaluated these challenging cases alongside a control group. The results revealed a collapse in human diagnostic performance on the AI-misclassified images. First, agreement with ground-truth labels plummeted, with Cohen's kappa dropping to a mere 0.08 for the difficult images, compared to a 0.61 for the control group. Second, we observed a severe deterioration in expert consensus; inter-rater reliability among physicians fell from moderate concordance (Fleiss kappa = 0.456) on control images to only modest agreement (Fleiss kappa = 0.275) on difficult cases. We identified image quality as a primary driver of these dual systematic failures. To promote transparency and reproducibility, all data, code, and trained models have been made publicly available

---

### DirectFisheye-GS: Enabling Native Fisheye Input in Gaussian Splatting with Cross-View Joint Optimization

Authors: Zhengxian Yang, Fei Xie, Xutao Xue, Rui Zhang, Taicheng Huang, Yang Liu, Mengqi Ji, Tao Yu
Published: 2026-04-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2604.00648v2

3D Gaussian Splatting (3DGS) has enabled efficient 3D scene reconstruction from everyday images with real-time, high-fidelity rendering, greatly advancing VR/AR applications. Fisheye cameras, with their wider field of view (FOV), promise high-quality reconstructions from fewer inputs and have recently attracted much attention. However, since 3DGS relies on rasterization, most subsequent works involving fisheye camera inputs first undistort images before training, which introduces two problems: 1) Black borders at image edges cause information loss and negate the fisheye's large FOV advantage; 2) Undistortion's stretch-and-interpolate resampling spreads each pixel's value over a larger area, diluting detail density -- causes 3DGS overfitting these low-frequency zones, producing blur and floating artifacts. In this work, we integrate fisheye camera model into the original 3DGS framework, enabling native fisheye image input for training without preprocessing. Despite correct modeling, we observed that the reconstructed scenes still exhibit floaters at image edges: Distortion increases toward the periphery, and 3DGS's original per-iteration random-selecting-view optimization ignores the cross-view correlations of a Gaussian, leading to extreme shapes (e.g., oversized or elongated) that degrade reconstruction quality. To address this, we introduce a feature-overlap-driven cross-view joint optimization strategy that establishes consistent geometric and photometric constraints across views-a technique equally applicable to existing pinhole-camera-based pipelines. Our DirectFisheye-GS matches or surpasses state-of-the-art performance on public datasets. Project Page: https://yzxqh.github.io/DirectFisheye-GS/ .

---