# Research Papers — 2026-03 · bundle page 167 of 243

> Full-text markdown of papers #8301–8350 of 12123 published in 2026-03, newest first. AI-readable mirrors of the full 12123-paper 2026-03 archive (50 papers per file).

Canonical page: https://aipod.dev/papers/archive/2026-03/page/167
[← Previous bundle](https://aipod.dev/papers/archive/2026-03/page/166.md)
[Next bundle →](https://aipod.dev/papers/archive/2026-03/page/168.md)

### Sabiá-4 Technical Report

Authors: Thiago Laitz, Thales Sales Almeida, Hugo Abonizio, Roseval Malaquias Junior, Giovana Kerche Bonás, Marcos Piau, Celio Larcher, Ramon Pires, Rodrigo Nogueira
Published: 2026-03-10
Categories: cs.CL
arXiv: https://arxiv.org/abs/2603.10213v1

This technical report presents Sabiá-4 and Sabiazinho-4, a new generation of Portuguese language models with a focus on Brazilian Portuguese language. The models were developed through a four-stage training pipeline: continued pre-training on Portuguese and Brazilian legal corpora, long-context extension to 128K tokens, supervised fine-tuning on instruction data spanning chat, code, legal tasks, and function calling, and preference alignment. We evaluate the models on six benchmark categories: conversational capabilities in Brazilian Portuguese, knowledge of Brazilian legislation, long-context understanding, instruction following, standardized exams, and agentic capabilities including tool use and web navigation. Results show that Sabiá-4 and Sabiazinho-4 achieve a favorable cost-performance trade-off compared to other models, positioning them in the upper-left region of the pricing-accuracy chart. The models show improvements over previous generations in legal document drafting, multi-turn dialogue quality, and agentic task completion.

---

### FusionNet: a frame interpolation network for 4D heart models

Authors: Chujie Chang, Shoko Miyauchi, Ken'ichi Morooka, Ryo Kurazume, Oscar Martinez Mozos
Published: 2026-03-10
Categories: cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2603.10212v1

Cardiac magnetic resonance (CMR) imaging is widely used to visualise cardiac motion and diagnose heart disease. However, standard CMR imaging requires patients to lie still in a confined space inside a loud machine for 40-60 min, which increases patient discomfort. In addition, shorter scan times decrease either or both the temporal and spatial resolutions of cardiac motion, and thus, the diagnostic accuracy of the procedure. Of these, we focus on reduced temporal resolution and propose a neural network called FusionNet to obtain four-dimensional (4D) cardiac motion with high temporal resolution from CMR images captured in a short period of time. The model estimates intermediate 3D heart shapes based on adjacent shapes. The results of an experimental evaluation of the proposed FusionNet model showed that it achieved a performance of over 0.897 in terms of the Dice coefficient, confirming that it can recover shapes more precisely than existing methods. This code is available at: https://github.com/smiyauchi199/FusionNet.git

---

### ViDia2Std: A Parallel Corpus and Methods for Low-Resource Vietnamese Dialect-to-Standard Translation

Authors: Khoa Anh Ta, Nguyen Van Dinh, Kiet Van Nguyen
Published: 2026-03-10
Categories: cs.CL
arXiv: https://arxiv.org/abs/2603.10211v1

Vietnamese exhibits extensive dialectal variation, posing challenges for NLP systems trained predominantly on standard Vietnamese. Such systems often underperform on dialectal inputs, especially from underrepresented Central and Southern regions. Previous work on dialect normalization has focused narrowly on Central-to-Northern dialect transfer using synthetic data and limited dialectal diversity. These efforts exclude Southern varieties and intra-regional variants within the North. We introduce ViDia2Std, the first manually annotated parallel corpus for dialect-to-standard Vietnamese translation covering all 63 provinces. Unlike prior datasets, ViDia2Std includes diverse dialects from Central, Southern, and non-standard Northern regions often absent from existing resources, making it the most dialectally inclusive corpus to date. The dataset consists of over 13,000 sentence pairs sourced from real-world Facebook comments and annotated by native speakers across all three dialect regions. To assess annotation consistency, we define a semantic mapping agreement metric that accounts for synonymous standard mappings across annotators. Based on this criterion, we report agreement rates of 86% (North), 82% (Central), and 85% (South). We benchmark several sequence-to-sequence models on ViDia2Std. mBART-large-50 achieves the best results (BLEU 0.8166, ROUGE-L 0.9384, METEOR 0.8925), while ViT5-base offers competitive performance with fewer parameters. ViDia2Std demonstrates that dialect normalization substantially improves downstream tasks, highlighting the need for dialect-aware resources in building robust Vietnamese NLP systems.

---

### Delta-K: Boosting Multi-Instance Generation via Cross-Attention Augmentation

Authors: Zitong Wang, Zijun Shen, Haohao Xu, Zhengjie Luo, Weibin Wu
Published: 2026-03-10
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2603.10210v1

While Diffusion Models excel in text-to-image synthesis, they often suffer from concept omission when synthesizing complex multi-instance scenes. Existing training-free methods attempt to resolve this by rescaling attention maps, which merely exacerbates unstructured noise without establishing coherent semantic representations. To address this, we propose Delta-K, a backbone-agnostic and plug-and-play inference framework that tackles omission by operating directly in the shared cross-attention Key space. Specifically, with Vision-language model, we extract a differential key $ΔK$ that encodes the semantic signature of missing concepts. This signal is then injected during the early semantic planning stage of the diffusion process. Governed by a dynamically optimized scheduling mechanism, Delta-K grounds diffuse noise into stable structural anchors while preserving existing concepts. Extensive experiments demonstrate the generality of our approach: Delta-K consistently improves compositional alignment across both modern DiT models and classical U-Net architectures, without requiring spatial masks, additional training, or architectural modifications.

---

### Flexible Cutoff Learning: Optimizing Machine Learning Potentials After Training

Authors: Rick Oerder, Jan Hamaekers
Published: 2026-03-10
Categories: cond-mat.mtrl-sci, cs.LG
arXiv: https://arxiv.org/abs/2603.10205v1

We introduce Flexible Cutoff Learning (FCL), a method for training machine learning interatomic potentials (MLIPs) whose cutoff radii can be adjusted after training. Unlike conventional MLIPs that fix the cutoff radius during training, FCL models are trained by randomly sampling cutoff radii independently for each atom. The resulting model can then be deployed with different per-atom cutoff radii depending on the application, enabling application-specific optimization of the accuracy-cost tradeoff. Using a differentiable cost model, these per-atom cutoffs can be optimized for specific target systems after training. We demonstrate FCL with a modified MACE architecture trained on the MAD dataset. For a subset featuring molecular crystals, optimized per-atom cutoffs reduce computational cost by more than 60% while increasing force errors by less than 1%. These results show that FCL enables training of a single general-purpose MLIP that can be adapted to diverse applications through post-training cutoff optimization, eliminating the need for retraining.

---

### Hybrid Hidden Markov Model for Modeling Equity Excess Growth Rate Dynamics: A Discrete-State Approach with Jump-Diffusion

Authors: Abdulrahman Alswaidan, Jeffrey D. Varner
Published: 2026-03-10
Categories: q-fin.ST, cs.LG, q-fin.RM
arXiv: https://arxiv.org/abs/2603.10202v2

Generating synthetic financial time series that preserve the statistical properties of real market data is essential for stress testing, risk model validation, and scenario design. Existing approaches struggle to simultaneously reproduce heavy-tailed distributions, negligible linear autocorrelation, and persistent volatility clustering. We developed a hybrid hidden Markov framework that discretized excess growth rates into Laplace quantile-defined states and augmented regime switching with a Poisson jump-duration mechanism to enforce realistic tail-state dwell times. Parameters were estimated by direct transition counting, bypassing the Baum-Welch EM algorithm and scaling to a 424-asset pipeline. Applied to ten years of daily equity data, the framework achieved high distributional pass rates both in-sample and out-of-sample while partially reproducing the volatility clustering that standard regime-switching models miss. No single model was best at everything: GARCH(1,1) better reproduced volatility clustering but failed distributional tests, while the standard HMM without jumps passed more distributional tests but could not generate volatility clustering. The proposed framework delivered the most balanced performance overall. For multi-asset generation, copula-based dependence models that preserved each asset's marginal HMM distribution substantially outperformed a Single-Index Model factor baseline on both per-asset distributional accuracy and correlation reproduction.

---

### Actor-Accelerated Policy Dual Averaging for Reinforcement Learning in Continuous Action Spaces

Authors: Ji Gao, Caleb Ju, Guanghui Lan, Zhaohui Tong
Published: 2026-03-10
Categories: cs.LG
arXiv: https://arxiv.org/abs/2603.10199v1

Policy Dual Averaging (PDA) offers a principled Policy Mirror Descent (PMD) framework that more naturally admits value function approximation than standard PMD, enabling the use of approximate advantage (or Q-) functions while retaining strong convergence guarantees. However, applying PDA in continuous state and action spaces remains computationally challenging, since action selection involves solving an optimization sub-problem at each decision step. In this paper, we propose \textit{actor-accelerated PDA}, which uses a learned policy network to approximate the solution of the optimization sub-problems, yielding faster runtimes while maintaining convergence guarantees. We provide a theoretical analysis that quantifies how actor approximation error impacts the convergence of PDA under suitable assumptions. We then evaluate its performance on several benchmarks in robotics, control, and operations research problems. Actor-accelerated PDA achieves superior performance compared to popular on-policy baselines such as Proximal Policy Optimization (PPO). Overall, our results bridge the gap between the theoretical advantages of PDA and its practical deployment in continuous-action problems with function approximation.

---

### Adaptive Activation Cancellation for Hallucination Mitigation in Large Language Models

Authors: Eric Yocam, Varghese Vaidyan, Gurcan Comert, Paris Kalathas, Yong Wang, Judith L. Mwakalonge
Published: 2026-03-10
Categories: cs.CL, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2603.10195v1

Large Language Models frequently generate fluent but factually incorrect text. We propose Adaptive Activation Cancellation (AAC), a real-time inference-time framework that treats hallucination-associated neural activations as structured interference within the transformer residual stream, drawing an explicit analogy to classical adaptive noise cancellation from signal processing. The framework identifies Hallucination Nodes (H-Nodes) via layer-wise linear probing and suppresses them using a confidence-weighted forward hook during auto-regressive generation -- requiring no external knowledge, no fine-tuning, and no additional inference passes. Evaluated across OPT-125M, Phi-3-mini, and LLaMA 3-8B on TruthfulQA and HaluEval, the real-time hook is the only intervention that consistently improves downstream accuracy on all three scales. Critically, the method is strictly surgical: WikiText-103 perplexity and MMLU reasoning accuracy are preserved at exactly 0.0% degradation across all three model scales, a property that distinguishes AAC from interventions that trade fluency or general capability for factual improvement. On the LLaMA 3-8B scale, the hook additionally yields positive generation-level gains (MC1 +0.04; MC2 +0.003; Token-F1 +0.003) while achieving probe-space selectivity 5.94x - 3.5x higher than the ITI baseline -- demonstrating that targeted neuron-level suppression can simultaneously improve factual accuracy and preserve model capability.

---

### MCP-in-SoS: Risk assessment framework for open-source MCP servers

Authors: Pratyay Kumar, Miguel Antonio Guirao Aguilera, Srikathyayani Srikanteswara, Satyajayant Misra, Abu Saleh Md Tayeen
Published: 2026-03-10
Categories: cs.CR, cs.AI
arXiv: https://arxiv.org/abs/2603.10194v1

Model Context Protocol (MCP) servers have rapidly emerged over the past year as a widely adopted way to enable Large Language Model (LLM) agents to access dynamic, real-world tools. As MCP servers proliferate and become easy to adopt via open-source releases, understanding their security risks becomes essential for dependable production agent deployments. Recent work has developed MCP threat taxonomies, proposed mitigations, and demonstrated practical attacks. However, to the best of our knowledge, no prior study has conducted a systematic, large-scale assessment of weaknesses in open-source MCP servers. Motivated by this gap, we apply static code analysis to identify Common Weakness Enumeration (CWE) weaknesses and map them to common attack patterns and threat categories using the MITRE Common Attack Pattern Enumerations and Classifications (CAPEC) to ground risk in real-world threats. We then introduce a risk-assessment framework for the MCP landscape that combines these threats using a multi-metric scoring of likelihood and impact. Our findings show that many open-source MCP servers contain exploitable weaknesses that can compromise confidentiality, integrity, and availability, underscoring the need for secure-by-design MCP server development.

---

### ARCHE: Autoregressive Residual Compression with Hyperprior and Excitation

Authors: Sofia Iliopoulou, Dimitris Ampeliotis, Athanassios Skodras
Published: 2026-03-10
Categories: eess.IV, cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2603.10188v1

Recent progress in learning-based image compression has demonstrated that end-to-end optimization can substantially outperform traditional codecs by jointly learning compact latent representations and probabilistic entropy models. However, many existing approaches achieve high rate-distortion efficiency at the expense of increased computational cost and limited parallelism. This paper presents ARCHE - Autoregressive Residual Compression with Hyperprior and Excitation, an end-to-end learned image compression framework that balances modeling accuracy and computational efficiency. The proposed architecture unifies hierarchical, spatial, and channel-based priors within a single probabilistic framework, capturing both global and local dependencies in the latent representation of the image, while employing adaptive feature recalibration and residual refinement to enhance latent representation quality. Without relying on recurrent or transformer-based components, ARCHE attains state-of-the-art rate-distortion efficiency: it reduces the BD-Rate by approximately 48% relative to the commonly used benchmark model of Balle et al., 30% relative to the channel-wise autoregressive model of Minnen & Singh and 5% against the VVC Intra codec on the Kodak benchmark dataset. The framework maintains computational efficiency with 95M parameters and 222ms running time per image. Visual comparisons confirm sharper textures and improved color fidelity, particularly at lower bit rates, demonstrating that accurate entropy modeling can be achieved through efficient convolutional designs suitable for practical deployment.

---

### Stabilizing Bandits using Regularization: Precise Regret and A Quantitative Central Limit Theorem

Authors: Budhaditya Halder, Ishan Sengupta, Koustav Chowdhury, Samya Praharaj, Koulik Khamaru
Published: 2026-03-10
Categories: stat.ML, cs.LG
arXiv: https://arxiv.org/abs/2603.10184v2

Statistical inference with bandit data presents fundamental challenges owing to adaptive sampling, which violates the independence assumptions underlying classical asymptotic theory. Recent work has identified stability~\citep{laiwei82} as a sufficient condition for valid inference under adaptivity. This paper first provides a refined stability condition, stated in terms of the iterates of an online algorithm, and shows that a large class of regularized stochastic-mirror-descent-style algorithms satisfy it. This refined condition allows us to strengthen the asymptotic results of~\citet{laiwei82} in several ways. First, we derive a non-asymptotic Berry--Esseen bound for the empirical reward estimates under adaptive sampling. Second, we derive matching non-asymptotic upper and lower bounds on the regret of the proposed algorithm, yielding a precise characterization of its regret. Third, we show that these regularized algorithms preserve asymptotic normality and valid inference under a prescribed level of adversarial corruption. Finally, we show that regularization is necessary rather than incidental: Lai--Wei stability is incompatible with the optimal $O(\sqrt{T})$ regret rate -- the rate attained by unregularized algorithms such as EXP3 -- so that a controlled, polylogarithmic inflation in regret is the price of valid inference.

---

### DT-BEHRT: Disease Trajectory-aware Transformer for Interpretable Patient Representation Learning

Authors: Deyi Li, Zijun Yao, Qi Xu, Muxuan Liang, Lingyao Li, Zijian Xu, Mei Liu
Published: 2026-03-10
Categories: cs.LG
arXiv: https://arxiv.org/abs/2603.10180v1

The growing adoption of electronic health record (EHR) systems has provided unprecedented opportunities for predictive modeling to guide clinical decision making. Structured EHRs contain longitudinal observations of patients across hospital visits, where each visit is represented by a set of medical codes. While sequence-based, graph-based, and graph-enhanced sequence approaches have been developed to capture rich code interactions over time or within the same visits, they often overlook the inherent heterogeneous roles of medical codes arising from distinct clinical characteristics and contexts. To this end, in this study we propose the Disease Trajectory-aware Transformer for EHR (DT-BEHRT), a graph-enhanced sequential architecture that disentangles disease trajectories by explicitly modeling diagnosis-centric interactions within organ systems and capturing asynchronous progression patterns. To further enhance the representation robustness, we design a tailored pre-training methodology that combines trajectory-level code masking with ontology-informed ancestor prediction, promoting semantic alignment across multiple modeling modules. Extensive experiments on multiple benchmark datasets demonstrate that DT-BEHRT achieves strong predictive performance and provides interpretable patient representations that align with clinicians' disease-centered reasoning. The source code is publicly accessible at https://github.com/GatorAIM/DT-BEHRT.git.

---

### Video-Based Reward Modeling for Computer-Use Agents

Authors: Linxin Song, Jieyu Zhang, Huanxin Sheng, Taiwei Shi, Gupta Rahul, Yang Liu, Ranjay Krishna, Jian Kang, Jieyu Zhao
Published: 2026-03-10
Categories: cs.CV, cs.CL
arXiv: https://arxiv.org/abs/2603.10178v1

Computer-using agents (CUAs) are becoming increasingly capable; however, it remains difficult to scale evaluation of whether a trajectory truly fulfills a user instruction. In this work, we study reward modeling from execution video: a sequence of keyframes from an agent trajectory that is independent of the agent's internal reasoning or actions. Although video-execution modeling is method-agnostic, it presents key challenges, including highly redundant layouts and subtle, localized cues that determine success. We introduce Execution Video Reward 53k (ExeVR-53k), a dataset of 53k high-quality video--task--reward triplets. We further propose adversarial instruction translation to synthesize negative samples with step-level annotations. To enable learning from long, high-resolution execution videos, we design spatiotemporal token pruning, which removes homogeneous regions and persistent tokens while preserving decisive UI changes. Building on these components, we fine-tune an Execution Video Reward Model (ExeVRM) that takes only a user instruction and a video-execution sequence to predict task success. Our ExeVRM 8B achieves 84.7% accuracy and 87.7% recall on video-execution assessment, outperforming strong proprietary models such as GPT-5.2 and Gemini-3 Pro across Ubuntu, macOS, Windows, and Android, while providing more precise temporal attribution. These results show that video-execution reward modeling can serve as a scalable, model-agnostic evaluator for CUAs.

---

### Calibration-Reasoning Framework for Descriptive Speech Quality Assessment

Authors: Elizaveta Kostenok, Mathieu Salzmann, Milos Cernak
Published: 2026-03-10
Categories: eess.AS, cs.CL
arXiv: https://arxiv.org/abs/2603.10175v1

Explainable speech quality assessment requires moving beyond Mean Opinion Scores (MOS) to analyze underlying perceptual dimensions. To address this, we introduce a novel post-training method that tailors the foundational Audio Large Language Model for multidimensional reasoning, detection and classification of audio artifacts. First, a calibration stage aligns the model to predict predefined perceptual dimensions. Second, a reinforcement learning stage leverages Group Relative Policy Optimization (GRPO) with dimension-specific rewards to heavily enhance accuracy of descriptions and temporal localization of quality issues. With this approach we reach state-of-the-art results of 0.71 mean PCC score on the multidimensional QualiSpeech benchmark and 13% improvement in MOS prediction driven by RL-based reasoning. Furthermore, our fine-grained GRPO rewards substantially advance the model's ability to pinpoint and classify audio artifacts in time.

---

### OpenClaw-RL: Train Any Agent Simply by Talking

Authors: Yinjie Wang, Xuyang Chen, Xiaolong Jin, Mengdi Wang, Ling Yang
Published: 2026-03-10
Categories: cs.CL, cs.AI, cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2603.10165v2

Every agent interaction generates a next-state signal, namely the user reply, tool output, terminal or GUI state change that follows each action, yet no existing agentic RL system recovers it as a live, online learning source. We present OpenClaw-RL, a framework that employs next-state signals to optimize personal agents online through infrastructure and methodology innovations. On the infrastructure side, we extend existing RL systems to a server-client architecture where the RL server hosts the policy behind an inference API and user terminals stream interaction data back over HTTP. From each observed next state, the system extracts two complementary training signals, evaluative and directive, via a separate asynchronous server so that neither signal extraction nor optimization blocks inference. On the methodology side, we introduce a hybrid RL objective that unifies both signal types in a single update: directive signals provide richer, token-level supervision but are sparser, while evaluative signals are more broadly available. To stabilize distillation under teacher-student mismatch, we propose overlap-guided hint selection, which picks the hint whose induced teacher distribution maximally overlaps with the student's top-$k$ tokens, together with a log-probability-difference clip that bounds per-token advantages. Applied to personal agents, OpenClaw-RL enables an agent to improve simply by being used, recovering conversational signals from user re-queries, corrections, and explicit feedback. Applied to general agents, OpenClaw-RL is the first RL framework to unify real-world agent settings spanning terminal, GUI, SWE, and tool-call environments, where we additionally demonstrate the utility of next-state signals in long-horizon settings.

---

### Compatibility at a Cost: Systematic Discovery and Exploitation of MCP Clause-Compliance Vulnerabilities

Authors: Nanzi Yang, Weiheng Bai, Kangjie Lu
Published: 2026-03-10
Categories: cs.CR, cs.AI
arXiv: https://arxiv.org/abs/2603.10163v1

The Model Context Protocol (MCP) is a recently proposed interoperability standard that unifies how AI agents connect with external tools and data sources. By defining a set of common client-server message exchange clauses, MCP replaces fragmented integrations with a standardized, plug-and-play framework. However, to be compatible with diverse AI agents, the MCP specification relaxes many behavioral constraints into optional clauses, leading to misuse-prone SDK implementation. We identify it as a new attack surface that allows adversaries to achieve multiple attacks (e.g, silent prompt injection, DoS, etc.), named as \emph{compatibility-abusing attacks}. In this work, we present the first systematic framework for analyzing this new attack surface across multi-language MCP SDKs. First, we construct a universal and language-agnostic intermediate representation (IR) generator that normalizes SDKs of different languages. Next, based on the new IR, we propose auditable static analysis with LLM-guided semantic reasoning for cross-language/clause compliance analysis. Third, by formalizing the attack semantics of the MCP clauses, we build three attack modalities and develop a modality-guided pipeline to uncover exploitable non-compliance issues.

---

### ReMix: Reinforcement routing for mixtures of LoRAs in LLM finetuning

Authors: Ruizhong Qiu, Hanqing Zeng, Yinglong Xia, Yiwen Meng, Ren Chen, Jiarui Feng, Dongqi Fu, Qifan Wang, Jiayi Liu, Jun Xiao, Xiangjun Fan, Benyu Zhang et al.
Published: 2026-03-10
Categories: cs.LG, cs.CL
arXiv: https://arxiv.org/abs/2603.10160v1

Low-rank adapters (LoRAs) are a parameter-efficient finetuning technique that injects trainable low-rank matrices into pretrained models to adapt them to new tasks. Mixture-of-LoRAs models expand neural networks efficiently by routing each layer input to a small subset of specialized LoRAs of the layer. Existing Mixture-of-LoRAs routers assign a learned routing weight to each LoRA to enable end-to-end training of the router. Despite their empirical promise, we observe that the routing weights are typically extremely imbalanced across LoRAs in practice, where only one or two LoRAs often dominate the routing weights. This essentially limits the number of effective LoRAs and thus severely hinders the expressive power of existing Mixture-of-LoRAs models. In this work, we attribute this weakness to the nature of learnable routing weights and rethink the fundamental design of the router. To address this critical issue, we propose a new router designed that we call Reinforcement Routing for Mixture-of-LoRAs (ReMix). Our key idea is using non-learnable routing weights to ensure all active LoRAs to be equally effective, with no LoRA dominating the routing weights. However, our routers cannot be trained directly via gradient descent due to our non-learnable routing weights. Hence, we further propose an unbiased gradient estimator for the router by employing the reinforce leave-one-out (RLOO) technique, where we regard the supervision loss as the reward and the router as the policy in reinforcement learning. Our gradient estimator also enables to scale up training compute to boost the predictive performance of our ReMix. Extensive experiments demonstrate that our proposed ReMix significantly outperform state-of-the-art parameter-efficient finetuning methods under a comparable number of activated parameters.

---

### Mashup Learning: Faster Finetuning by Remixing Past Checkpoints

Authors: Sofia Maria Lo Cicero Vaina, Artem Chumachenko, Max Ryabinin
Published: 2026-03-10
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.10156v1

Finetuning on domain-specific data is a well-established method for enhancing LLM performance on downstream tasks. Training on each dataset produces a new set of model weights, resulting in a multitude of checkpoints saved in-house or on open-source platforms. However, these training artifacts are rarely reused for subsequent experiments despite containing improved model abilities for potentially similar tasks. In this paper, we propose Mashup Learning, a simple method to leverage the outputs of prior training runs to enhance model adaptation to new tasks. Our procedure identifies the most relevant historical checkpoints for a target dataset, aggregates them with model merging, and uses the result as an improved initialization for training. Across 8 standard LLM benchmarks, four models, and two collections of source checkpoints, Mashup Learning consistently improves average downstream accuracy by 0.5-5 percentage points over training from scratch. It also accelerates convergence, requiring 41-46% fewer training steps and up to 37% less total wall-clock time to match from-scratch accuracy, including all selection and merging overhead.

---

### A neural operator for predicting vibration frequency response curves from limited data

Authors: D. Bluedorn, A. Badawy, B. E. Saunders, D. Roettgen, A. Abdelkefi
Published: 2026-03-10
Categories: cs.LG, eess.SY
arXiv: https://arxiv.org/abs/2603.10149v1

In the design of engineered components, rigorous vibration testing is essential for performance validation and identification of resonant frequencies and amplitudes encountered during operation. Performing this evaluation numerically via machine learning has great potential to accelerate design iteration and make testing workflows more efficient. However, dynamical systems are conventionally difficult to solve via machine learning methods without using physics-based regularizing loss functions. To properly perform this forecasting task, a structure that has an inspectable physical obedience can be devised without the use of regularizing terms from first principles. The method employed in this work is a neural operator integrated with an implicit numerical scheme. This architecture enables operators to learn of the underlying state-space dynamics from limited data, allowing generalization to untested driving frequencies and initial conditions. This network can infer the system's global frequency response by training on a small set of input conditions. As a foundational proof of concept, this investigation verifies the machine learning algorithm with a linear, single-degree-of-freedom system, demonstrating implicit obedience of dynamics. This approach demonstrates 99.87% accuracy in predicting the Frequency Response Curve (FRC), forecasting the frequency and amplitude of linear resonance training on 7% of the bandwidth of the solution. By training machine learning models to internalize physics information rather than trajectory, better generalization accuracy can be realized, vastly improving the timeframe for vibration studies on engineered components.

---

### Social Knowledge for Cross-Domain User Preference Modeling

Authors: Nir Lotan, Adir Solomon, Ido Guy, Einat Minkov
Published: 2026-03-10
Categories: cs.SI, cs.AI
arXiv: https://arxiv.org/abs/2603.10148v1

We demonstrate that user preferences can be represented and predicted across topical domains using large-scale social modeling. Given information about popular entities favored by a user, we project the user into a social embedding space learned from a large-scale sample of the Twitter (now X) network. By representing both users and popular entities in a joint social space, we can assess the relevance of candidate entities (e.g., music artists) using cosine similarity within this embedding space. A comprehensive evaluation using link prediction experiments shows that this method achieves effective personalization in zero-shot setting, when no user feedback is available for entities in the target domain, yielding substantial improvements over a strong popularity-based baseline. In-depth analysis further illustrates that socio-demographic factors encoded in the social embeddings are correlated with user preferences across domains. Finally, we argue and demonstrate that the proposed approach can facilitate social modeling of end users using large language models (LLMs).

---

### Lost in Backpropagation: The LM Head is a Gradient Bottleneck

Authors: Nathan Godey, Yoav Artzi
Published: 2026-03-10
Categories: cs.CL
arXiv: https://arxiv.org/abs/2603.10145v2

The last layer of neural language models (LMs) projects output features of dimension $D$ to logits in dimension $V$, the size of the vocabulary, where usually $D \ll V$. This mismatch is known to raise risks of limited expressivity in neural LMs, creating a so-called softmax bottleneck. We show the softmax bottleneck is not only an expressivity bottleneck but also an optimization bottleneck. Backpropagating $V$-dimensional gradients through a rank-$D$ linear layer induces unavoidable compression, which alters the training feedback provided to the vast majority of the parameters. We present a theoretical analysis of this phenomenon and measure empirically that 95-99% of the gradient norm is suppressed by the output layer, resulting in vastly suboptimal update directions. We conduct controlled pretraining experiments showing that the gradient bottleneck makes trivial patterns unlearnable, and drastically affects the training dynamics of LLMs. We argue that this inherent flaw contributes to training inefficiencies at scale independently of the model architecture, and raises the need for new LM head designs.

---

### Reason and Verify: A Framework for Faithful Retrieval-Augmented Generation

Authors: Eeham Khan, Luis Rodriguez, Marc Queudot
Published: 2026-03-10
Categories: cs.CL
arXiv: https://arxiv.org/abs/2603.10143v1

Retrieval-Augmented Generation (RAG) significantly improves the factuality of Large Language Models (LLMs), yet standard pipelines often lack mechanisms to verify inter- mediate reasoning, leaving them vulnerable to hallucinations in high-stakes domains. To address this, we propose a domain-specific RAG framework that integrates explicit rea- soning and faithfulness verification. Our architecture augments standard retrieval with neural query rewriting, BGE-based cross-encoder reranking, and a rationale generation module that grounds sub-claims in specific evidence spans. We further introduce an eight-category verification taxonomy that enables fine-grained assessment of rationale faithfulness, distinguishing between explicit and implicit support patterns to facilitate structured error diagnosis. We evaluate this framework on the BioASQ and PubMedQA benchmarks, specifically analyzing the impact of dynamic in-context learning and rerank- ing under constrained token budgets. Experiments demonstrate that explicit rationale generation improves accuracy over vanilla RAG baselines, while dynamic demonstration selection combined with robust reranking yields further gains in few-shot settings. Using Llama-3-8B-Instruct, our approach achieves 89.1% on BioASQ-Y/N and 73.0% on Pub- MedQA, competitive with systems using significantly larger models. Additionally, we perform a pilot study combining human expert assessment with LLM-based verification to explore how explicit rationale generation improves system transparency and enables more detailed diagnosis of retrieval failures in biomedical question answering.

---

### The Generation-Recognition Asymmetry: Six Dimensions of a Fundamental Divide in Formal Language Theory

Authors: Romain Peyrichou
Published: 2026-03-10
Categories: cs.CL, cs.AI, cs.CC, cs.FL
arXiv: https://arxiv.org/abs/2603.10139v2

Every formal grammar defines a language and can in principle be used in three ways: to generate strings (production), to recognize them (parsing), or -- given only examples -- to infer the grammar itself (grammar induction). Generation and recognition are extensionally equivalent -- they characterize the same set -- but operationally asymmetric in multiple independent ways. Inference is a qualitatively harder problem: it does not have access to a known grammar. Despite the centrality of this triad to compiler design, natural language processing, and formal language theory, no survey has treated it as a unified, multidimensional phenomenon. We identify six dimensions along which generation and recognition diverge: computational complexity, ambiguity, directionality, information availability, grammar inference, and temporality. We show that the common characterization "generation is easy, parsing is hard" is misleading: unconstrained generation is trivial, but generation under constraints can be NP-hard. The real asymmetry is that parsing is always constrained (the input is given) while generation need not be. Two of these dimensions -- directionality and temporality -- have not previously been identified as dimensions of the generation-recognition asymmetry. We connect the temporal dimension to the surprisal framework of Hale (2001) and Levy (2008), arguing that surprisal formalizes the temporal asymmetry between a generator (surprisal = 0) and a parser that predicts under uncertainty (surprisal > 0). We review bidirectional systems in NLP and observe that bidirectionality has been available for fifty years yet has not transferred to most domain-specific applications. We conclude with a discussion of large language models, which architecturally unify generation and recognition while operationally preserving the asymmetry.

---

### Agentic Control Center for Data Product Optimization

Authors: Priyadarshini Tamilselvan, Gregory Bramble, Sola Shirai, Ken C. L. Wong, Faisal Chowdhury, Horst Samulowitz
Published: 2026-03-10
Categories: cs.AI
arXiv: https://arxiv.org/abs/2603.10133v1

Data products enable end users to gain greater insights about their data by providing supporting assets, such as example question-SQL pairs which can be answered using the data or views over the database tables. However, producing useful data products is challenging, and typically requires domain experts to hand-craft supporting assets. We propose a system that automates data product improvement through specialized AI agents operating in a continuous optimization loop. By surfacing questions, monitoring multi-dimensional quality metrics, and supporting human-in-the-loop controls, it transforms data into observable and refinable assets that balance automation with trust and oversight.

---

### Unbalanced Optimal Transport Dictionary Learning for Unsupervised Hyperspectral Image Clustering

Authors: Joshua Lentz, Nicholas Karris, Alex Cloninger, James M. Murphy
Published: 2026-03-10
Categories: cs.CV, cs.LG, math.ST
arXiv: https://arxiv.org/abs/2603.10132v1

Hyperspectral images capture vast amounts of high-dimensional spectral information about a scene, making labeling an intensive task that is resistant to out-of-the-box statistical methods. Unsupervised learning of clusters allows for automated segmentation of the scene, enabling a more rapid understanding of the image. Partitioning the spectral information contained within the data via dictionary learning in Wasserstein space has proven an effective method for unsupervised clustering. However, this approach requires balancing the spectral profiles of the data, blurring the classes, and sacrificing robustness to outliers and noise. In this paper, we suggest improving this approach by utilizing unbalanced Wasserstein barycenters to learn a lower-dimensional representation of the underlying data. The deployment of spectral clustering on the learned representation results in an effective approach for the unsupervised learning of labels.

---

### The Prediction-Measurement Gap: Toward Meaning Representations as Scientific Instruments

Authors: Hubert Plisiecki
Published: 2026-03-10
Categories: cs.CL
arXiv: https://arxiv.org/abs/2603.10130v1

Text embeddings have become central to computational social science and psychology, enabling scalable measurement of meaning and mixed-method inference. Yet most representation learning is optimized and evaluated for prediction and retrieval, yielding a prediction-measurement gap: representations that perform well as features may be poorly suited as scientific instruments. The paper argues that scientific meaning analysis motivates a distinct family of objectives - scientific usability - emphasizing geometric legibility, interpretability and traceability to linguistic evidence, robustness to non-semantic confounds, and compatibility with regression-style inference over semantic directions. Grounded in cognitive and neuro-psychological views of meaning, the paper assesses static word embeddings and contextual transformer representations against these requirements: static spaces remain attractive for transparent measurement, whereas contextual spaces offer richer semantics but entangle meaning with other signals and exhibit geometric and interpretability issues that complicate inference. The paper then outlines a course-setting agenda around (i) geometry-first design for gradients and abstraction, including hierarchy-aware spaces constrained by psychologically privileged levels; (ii) invertible post-hoc transformations that recondition embedding geometry and reduce nuisance influence; and (iii) meaning atlases and measurement-oriented evaluation protocols for reliable and traceable semantic inference. As the field debates the limits of scale-first progress, measurement-ready representations offer a principled new frontier.

---

### HG-Lane: High-Fidelity Generation of Lane Scenes under Adverse Weather and Lighting Conditions without Re-annotation

Authors: Daichao Zhao, Qiupu Chen, Feng He, Xin Ning, Qiankun Li
Published: 2026-03-10
Categories: cs.CV
arXiv: https://arxiv.org/abs/2603.10128v2

Lane detection is a crucial task in autonomous driving, as it helps ensure the safe operation of vehicles. However, existing datasets such as CULane and TuSimple contain relatively limited data under extreme weather conditions, including rain, snow, and fog. As a result, detection models trained on these datasets often become unreliable in such environments, which may lead to serious safety-critical failures on the road. To address this issue, we propose HG-Lane, a High-fidelity Generation framework for Lane Scenes under adverse weather and lighting conditions without requiring re-annotation. Based on this framework, we further construct a benchmark that includes adverse weather and lighting scenarios, containing 30,000 images. Experimental results demonstrate that our method consistently and significantly improves the performance of existing lane detection networks. For example, using the state-of-the-art CLRNet, the overall mF1 score on our benchmark increases by 20.87 percent. The F1@50 score for the overall, normal, snow, rain, fog, night, and dusk categories increases by 19.75 percent, 8.63 percent, 38.8 percent, 14.96 percent, 26.84 percent, 21.5 percent, and 12.04 percent, respectively. The code and dataset are available at: https://github.com/zdc233/HG-Lane.

---

### AR-VLA: True Autoregressive Action Expert for Vision-Language-Action Models

Authors: Yutong Hu, Jan-Nico Zaech, Nikolay Nikolov, Yuanqi Yao, Sombit Dey, Giuliano Albanese, Renaud Detry, Luc Van Gool, Danda Paudel
Published: 2026-03-10
Categories: cs.RO, cs.AI
arXiv: https://arxiv.org/abs/2603.10126v2

We propose a standalone autoregressive (AR) Action Expert that generates actions as a continuous causal sequence while conditioning on refreshable vision-language prefixes. In contrast to existing Vision-Language-Action (VLA) models and diffusion policies that reset temporal context with each new observation and predict actions reactively, our Action Expert maintains its own history through a long-lived memory and is inherently context-aware. This structure addresses the frequency mismatch between fast control and slow reasoning, enabling efficient independent pretraining of kinematic syntax and modular integration with heavy perception backbones, naturally ensuring spatio-temporally consistent action generation across frames. To synchronize these asynchronous hybrid V-L-A modalities, we utilize a re-anchoring mechanism that mathematically accounts for perception staleness during both training and inference. Experiments on simulated and real-robot manipulation tasks demonstrate that the proposed method can effectively replace traditional chunk-based action heads for both specialist and generalist policies. AR-VLA exhibits superior history awareness and substantially smoother action trajectories while maintaining or exceeding the task success rates of state-of-the-art reactive VLAs. Overall, our work introduces a scalable, context-aware action generation schema that provides a robust structural foundation for training effective robotic policies. Code and Videos available at https://arvla.insait.ai

---

### 4DEquine: Disentangling Motion and Appearance for 4D Equine Reconstruction from Monocular Video

Authors: Jin Lyu, Liang An, Pujin Cheng, Yebin Liu, Xiaoying Tang
Published: 2026-03-10
Categories: cs.CV
arXiv: https://arxiv.org/abs/2603.10125v1

4D reconstruction of equine family (e.g. horses) from monocular video is important for animal welfare. Previous mainstream 4D animal reconstruction methods require joint optimization of motion and appearance over a whole video, which is time-consuming and sensitive to incomplete observation. In this work, we propose a novel framework called 4DEquine by disentangling the 4D reconstruction problem into two sub-problems: dynamic motion reconstruction and static appearance reconstruction. For motion, we introduce a simple yet effective spatio-temporal transformer with a post-optimization stage to regress smooth and pixel-aligned pose and shape sequences from video. For appearance, we design a novel feed-forward network that reconstructs a high-fidelity, animatable 3D Gaussian avatar from as few as a single image. To assist training, we create a large-scale synthetic motion dataset, VarenPoser, which features high-quality surface motions and diverse camera trajectories, as well as a synthetic appearance dataset, VarenTex, comprising realistic multi-view images generated through multi-view diffusion. While training only on synthetic datasets, 4DEquine achieves state-of-the-art performance on real-world APT36K and AiM datasets, demonstrating the superiority of 4DEquine and our new datasets for both geometry and appearance reconstruction. Comprehensive ablation studies validate the effectiveness of both the motion and appearance reconstruction network. Project page: https://luoxue-star.github.io/4DEquine_Project_Page/.

---

### Lost in the Middle at Birth: An Exact Theory of Transformer Position Bias

Authors: Borun D Chowdhury
Published: 2026-03-10
Categories: cs.LG, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2603.10123v1

The ``Lost in the Middle'' phenomenon -- a U-shaped performance curve where LLMs retrieve well from the beginning and end of a context but fail in the middle -- is widely attributed to learned Softmax artifacts or the distance-decay of positional encodings like RoPE. This paper makes a single, precise claim: \emph{the U-shape is already present at initialization, before any training or positional encoding takes effect.} It is an inherent geometric property of the causal decoder with residual connections. We model multi-layer causal attention as iterated powers of the Cesàro matrix and derive the exact closed-form influence density in the continuous limit. Causal masking forces a logarithmic divergence of gradient influence at the start of the prompt (the Primacy Tail), while residual connections create an isolated $\mathcal{O}(1)$ anchor at the final token (the Recency Delta). Between these extremes lies a factorial dead zone of order $\mathcal{O}(1/(H{-}1)!)$, where $H$ is the network depth, making middle-context retrieval and training structurally hostile. We validate empirically that untrained Qwen2 and GPT-2 architectures exhibit this U-shape at Step~0, and that it is identical with or without RoPE. Comparing initialized and pretrained networks, we show that standard training does not overcome the topological valley, confirming that the U-shape persists as an architectural baseline under standard pretraining objectives. We do not claim that this bias is insurmountable, nor that interventions such as RoPE modifications are useless. We establish what the baseline is and where it comes from, so that future efforts to overcome it can be precisely targeted.

---

### CLIPO: Contrastive Learning in Policy Optimization Generalizes RLVR

Authors: Sijia Cui, Pengyu Cheng, Jiajun Song, Yongbo Gai, Guojun Zhang, Zhechao Yu, Jianhe Lin, Xiaoxi Jiang, Guanjun Jiang
Published: 2026-03-10
Categories: cs.LG, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2603.10101v1

Reinforcement Learning with Verifiable Rewards (RLVR) has significantly advanced the reasoning capacity of Large Language Models (LLMs). However, RLVR solely relies on final answers as outcome rewards, neglecting the correctness of intermediate reasoning steps. Training on these process-wrong but outcome-correct rollouts can lead to hallucination and answer-copying, severely undermining the model's generalization and robustness. To address this, we incorporate a Contrastive Learning mechanism into the Policy Optimization (CLIPO) to generalize the RLVR process. By optimizing a contrastive loss over successful rollouts, CLIPO steers the LLM to capture the invariant structure shared across correct reasoning paths. This provides a more robust cross-trajectory regularization than the original single-path supervision in RLVR, effectively mitigating step-level reasoning inconsistencies and suppressing hallucinatory artifacts. In experiments, CLIPO consistently improves multiple RLVR baselines across diverse reasoning benchmarks, demonstrating uniform improvements in generalization and robustness for policy optimization of LLMs. Our code and training recipes are available at https://github.com/Qwen-Applications/CLIPO.

---

### Hardware Efficient Approximate Convolution with Tunable Error Tolerance for CNNs

Authors: Vishal Shashidhar, Anupam Kumari, Roy P Paily
Published: 2026-03-10
Categories: cs.LG, cs.AI, cs.AR
arXiv: https://arxiv.org/abs/2603.10100v2

Modern CNNs' high computational demands hinder edge deployment, as traditional ``hard'' sparsity (skipping mathematical zeros) loses effectiveness in deep layers or with smooth activations like Tanh. We propose a ``soft sparsity'' paradigm using a hardware efficient Most Significant Bit (MSB) proxy to skip negligible non-zero multiplications. Integrated as a custom RISC-V instruction and evaluated on LeNet-5 (MNIST), this method reduces ReLU MACs by 88.42% and Tanh MACs by 74.87% with zero accuracy loss--outperforming zero-skipping by 5x. By clock-gating inactive multipliers, we estimate power savings of 35.2% for ReLU and 29.96% for Tanh. While memory access makes power reduction sub-linear to operation savings, this approach significantly optimizes resource-constrained inference.

---

### Denoising the US Census: Succinct Block Hierarchical Regression

Authors: Badih Ghazi, Pritish Kamath, Ravi Kumar, Pasin Manurangsi, Adam Sealfon
Published: 2026-03-10
Categories: cs.LG, cs.CR
arXiv: https://arxiv.org/abs/2603.10099v1

The US Census Bureau Disclosure Avoidance System (DAS) balances confidentiality and utility requirements for the decennial US Census (Abowd et al., 2022). The DAS was used in the 2020 Census to produce demographic datasets critically used for legislative apportionment and redistricting, federal and state funding allocation, municipal and infrastructure planning, and scientific research. At the heart of DAS is TopDown, a heuristic post-processing method that combines billions of private noisy measurements across six geographic levels in order to produce new estimates that are consistent, more accurate, and satisfy certain structural constraints on the data. In this work, we introduce BlueDown, a new post-processing method that produces more accurate, consistent estimates while satisfying the same privacy guarantees and structural constraints. We obtain especially large accuracy improvements for aggregates at the county and tract levels on evaluation metrics proposed by the US Census Bureau. From a technical perspective, we develop a new algorithm for generalized least-squares regression that leverages the hierarchical structure of the measurements and that is statistically optimal among linear unbiased estimators. This reduces the computational dependence on the number of geographic regions measured from matrix multiplication time, which would be infeasible for census-scale data, to linear time. We incorporate the additional structural constraints by combining this regression algorithm with an optimization routine that extends TDA to support correlated measurements. We further improve the efficiency of our algorithm using succinct linear-algebraic operations that exploit symmetries in the structure of the measurements and constraints. We believe our hierarchical regression and succinct operations to be of independent interest.

---

### Code-Space Response Oracles: Generating Interpretable Multi-Agent Policies with Large Language Models

Authors: Daniel Hennes, Zun Li, John Schultz, Marc Lanctot
Published: 2026-03-10
Categories: cs.GT, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2603.10098v1

Recent advances in multi-agent reinforcement learning, particularly Policy-Space Response Oracles (PSRO), have enabled the computation of approximate game-theoretic equilibria in increasingly complex domains. However, these methods rely on deep reinforcement learning oracles that produce `black-box' neural network policies, making them difficult to interpret, trust or debug. We introduce Code-Space Response Oracles (CSRO), a novel framework that addresses this challenge by replacing RL oracles with Large Language Models (LLMs). CSRO reframes the best response computation as a code generation task, prompting an LLM to generate policies directly as human-readable code. This approach not only yields inherently interpretable policies but also leverages the LLM's pretrained knowledge to discover complex, human-like strategies. We explore multiple ways to construct and enhance an LLM-based oracle: zero-shot prompting, iterative refinement and \emph{AlphaEvolve}, a distributed LLM-based evolutionary system. We demonstrate that CSRO achieves performance competitive with baselines while producing a diverse set of explainable policies. Our work presents a new perspective on multi-agent learning, shifting the focus from optimizing opaque policy parameters to synthesizing interpretable algorithmic behavior.

---

### Rethinking Adam for Time Series Forecasting: A Simple Heuristic to Improve Optimization under Distribution Shifts

Authors: Yuze Dong, Jinsong Wu
Published: 2026-03-10
Categories: cs.LG, stat.ML
arXiv: https://arxiv.org/abs/2603.10095v1

Time-series forecasting often faces challenges from non-stationarity, particularly distributional drift, where the data distribution evolves over time. This dynamic behavior can undermine the effectiveness of adaptive optimizers, such as Adam, which are typically designed for stationary objectives. In this paper, we revisit Adam in the context of non-stationary forecasting and identify that its second-order bias correction limits responsiveness to shifting loss landscapes. To address this, we propose TS_Adam, a lightweight variant that removes the second-order correction from the learning rate computation. This simple modification improves adaptability to distributional drift while preserving the optimizer core structure and requiring no additional hyperparameters. TS_Adam integrates easily into existing models and consistently improves performance across long- and short-term forecasting tasks. On the ETT datasets with the MICN model, it achieves an average reduction of 12.8% in MSE and 5.7% in MAE compared to Adam. These results underscore the practicality and versatility of TS_Adam as an effective optimization strategy for real-world forecasting scenarios involving non-stationary data. Code is available at: https://github.com/DD-459-1/TS_Adam.

---

### Equivariant Asynchronous Diffusion: An Adaptive Denoising Schedule for Accelerated Molecular Conformation Generation

Authors: Junyi An, Chao Qu, Yun-Fei Shi, Zhijian Zhou, Fenglei Cao, Yuan Qi
Published: 2026-03-10
Categories: cs.LG, cs.AI, q-bio.QM
arXiv: https://arxiv.org/abs/2603.10093v2

Recent 3D molecular generation methods primarily use asynchronous auto-regressive or synchronous diffusion models. While auto-regressive models build molecules sequentially, they're limited by a short horizon and a discrepancy between training and inference. Conversely, synchronous diffusion models denoise all atoms at once, offering a molecule-level horizon but failing to capture the causal relationships inherent in hierarchical molecular structures. We introduce Equivariant Asynchronous Diffusion (EAD) to overcome these limitations. EAD is a novel diffusion model that combines the strengths of both approaches: it uses an asynchronous denoising schedule to better capture molecular hierarchy while maintaining a molecule-level horizon. Since these relationships are often complex, we propose a dynamic scheduling mechanism to adaptively determine the denoising timestep. Experimental results show that EAD achieves state-of-the-art performance in 3D molecular generation.

---

### Execution Is the New Attack Surface: Survivability-Aware Agentic Crypto Trading with OpenClaw-Style Local Executors

Authors: Ailiya Borjigin, Igor Stadnyk, Ben Bilski, Serhii Hovorov, Sofiia Pidturkina
Published: 2026-03-10
Categories: cs.CR, cs.AI
arXiv: https://arxiv.org/abs/2603.10092v1

OpenClaw-style agent stacks turn language into privileged execution: LLM intents flow through tool interception, policy gates, and a local executor. In parallel, skill marketplaces such as skills.sh make capability acquisition as easy as installing skills and CLIs, creating a growing capability supply chain. Together, these trends shift the dominant safety failure mode from "wrong answers" to execution-induced loss, where untrusted prompts, compromised skills, or narrative manipulation can trigger real trades and irreversible side effects. We propose Survivability-Aware Execution (SAE), an execution-layer survivability standard for OpenClaw-style systems and skill-enabled agents. SAE sits as middleware between a strategy engine (LLM or non-LLM) and the exchange executor. It defines an explicit execution contract (ExecutionRequest, ExecutionContext, ExecutionDecision) and enforces non-bypassable last-mile invariants: projection-based exposure budgets, cooldown and order-rate limits, slippage bounds, staged execution, and tool/venue allowlists. To make delegated execution testable under supply-chain risk, we operationalize the Delegation Gap (DG) via a logged Intended Policy Spec that enables deterministic out-of-scope labeling and reproducible DG metrics. On an offline replay using official Binance USD-M BTCUSDT/ETHUSDT perpetual data (15m; 2025-09-01--2025-12-01, incl. funding), SAE improves survivability: MDD drops from 0.4643 to 0.0319 (Full; 93.1%), |CVaR_0.99| shrinks from 4.025e-3 to ~1.02e-4 (~97.5%), and DG loss proxy falls from 0.647 to 0.019 (~97.0%). AttackSuccess decreases from 1.00 to 0.728 with zero FalseBlock in this run. Block bootstrap, paired Wilcoxon, and two-proportion tests confirm the shifts. SAE reframes agentic trading safety for the OpenClaw+skills era: treat upstream intent and skills as untrusted, and enforce survivability where actions become side effects.

---

### Multi-Stream Perturbation Attack: Breaking Safety Alignment of Thinking LLMs Through Concurrent Task Interference

Authors: Fan Yang
Published: 2026-03-10
Categories: cs.CR, cs.AI
arXiv: https://arxiv.org/abs/2603.10091v1

The widespread adoption of thinking mode in large language models (LLMs) has significantly enhanced complex task processing capabilities while introducing new security risks. When subjected to jailbreak attacks, the step-by-step reasoning process may cause models to generate more detailed harmful content. We observe that thinking mode exhibits unique vulnerabilities when processing interleaved multiple tasks. Based on this observation, we propose multi-stream perturbation attack, which generates superimposed interference by interweaving multiple task streams within a single prompt. We design three perturbation strategies: multi-stream interleaving, inversion perturbation, and shape transformation, which disrupt the thinking process through concurrent task interleaving, character reversal, and format constraints respectively. On JailbreakBench, AdvBench, and HarmBench datasets, our method achieves attack success rates exceeding most methods across mainstream models including Qwen3 series, DeepSeek, Qwen3-Max, and Gemini 2.5 Flash. Experiments show thinking collapse rates and response repetition rates reach up to 17% and 60% respectively, indicating multi-stream perturbation not only bypasses safety mechanisms but also causes thinking process collapse or repetitive outputs.

---

### A Survey of Weight Space Learning: Understanding, Representation, and Generation

Authors: Xiaolong Han, Zehong Wang, Bo Zhao, Binchi Zhang, Jundong Li, Damian Borth, Rose Yu, Haggai Maron, Yanfang Ye, Lu Yin, Ferrante Neri
Published: 2026-03-10
Categories: cs.LG
arXiv: https://arxiv.org/abs/2603.10090v1

Neural network weights are typically viewed as the end product of training, while most deep learning research focuses on data, features, and architectures. However, recent advances show that the set of all possible weight values (weight space) itself contains rich structure: pretrained models form organized distributions, exhibit symmetries, and can be embedded, compared, or even generated. Understanding such structures has tremendous impact on how neural networks are analyzed and compared, and on how knowledge is transferred across models, beyond individual training instances. This emerging research direction, which we refer to as Weight Space Learning (WSL), treats neural weights as a meaningful domain for analysis and modeling. This survey provides the first unified taxonomy of WSL. We categorize existing methods into three core dimensions: Weight Space Understanding (WSU), which studies the geometry and symmetries of weights; Weight Space Representation (WSR), which learns embeddings over model weights; and Weight Space Generation (WSG), which synthesizes new weights through hypernetworks or generative models. We further show how these developments enable practical applications, including model retrieval, continual and federated learning, neural architecture search, and data-free reconstruction. By consolidating fragmented progress under a coherent framework, this survey highlights weight space as a learnable, structured domain with growing impact across model analysis, transferring, and weight generation. We release an accompanying resource at https://github.com/Zehong-Wang/Awesome-Weight-Space-Learning.

---

### ES-dLLM: Efficient Inference for Diffusion Large Language Models by Early-Skipping

Authors: Zijian Zhu, Fei Ren, Zhanhong Tan, Kaisheng Ma
Published: 2026-03-10
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.10088v1

Diffusion large language models (dLLMs) are emerging as a promising alternative to autoregressive models (ARMs) due to their ability to capture bidirectional context and the potential for parallel generation. Despite the advantages, dLLM inference remains computationally expensive as the full input context is processed at every iteration. In this work, we analyze the generation dynamics of dLLMs and find that intermediate representations, including key, value, and hidden states, change only subtly across successive iterations. Leveraging this insight, we propose \textbf{ES-dLLM}, a training-free inference acceleration framework for dLLM that reduces computation by skipping tokens in early layers based on the estimated importance. Token importance is computed with intermediate tensor variation and confidence scores of previous iterations. Experiments on LLaDA-8B and Dream-7B demonstrate that ES-dLLM achieves throughput of up to 226.57 and 308.51 tokens per second (TPS), respectively, on an NVIDIA H200 GPU, delivering 5.6$\times$ to 16.8$\times$ speedup over the vanilla implementation and up to 1.85$\times$ over the state-of-the-art caching method, while preserving generation quality.

---

### KernelSkill: A Multi-Agent Framework for GPU Kernel Optimization

Authors: Qitong Sun, Jun Han, Tianlin Li, Zhe Tang, Sheng Chen, Fei Yang, Aishan Liu, Xianglong Liu, Yang Liu
Published: 2026-03-10
Categories: cs.LG, cs.AI, cs.MA
arXiv: https://arxiv.org/abs/2603.10085v1

Improving GPU kernel efficiency is crucial for advancing AI systems. Recent work has explored leveraging large language models (LLMs) for GPU kernel generation and optimization. However, existing LLM-based kernel optimization pipelines typically rely on opaque, implicitly learned heuristics within the LLMs to determine optimization strategies. This leads to inefficient trial-and-error and weakly interpretable optimizations. Our key insight is to replace implicit heuristics with expert optimization skills that are knowledge-driven and aware of task trajectories. Specifically, we present KernelSkill, a multi-agent framework with a dual-level memory architecture. KernelSkill operates by coordinating agents with long-term memory of reusable expert skills and short-term memory to prevent repetitive backtracking. On KernelBench Levels 1-3, KernelSkill achieves a 100% success rate and average speedups of 5.44x, 2.82x, and 1.92x over Torch Eager on Levels 1, 2, and 3, respectively, outperforming prior baselines. Code is available at https://github.com/0satan0/KernelMem/.

---

### Digging Deeper: Learning Multi-Level Concept Hierarchies

Authors: Oscar Hill, Mateo Espinosa Zarlenga, Mateja Jamnik
Published: 2026-03-10
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.10084v1

Although concept-based models promise interpretability by explaining predictions with human-understandable concepts, they typically rely on exhaustive annotations and treat concepts as flat and independent. To circumvent this, recent work has introduced Hierarchical Concept Embedding Models (HiCEMs) to explicitly model concept relationships, and Concept Splitting to discover sub-concepts using only coarse annotations. However, both HiCEMs and Concept Splitting are restricted to shallow hierarchies. We overcome this limitation with Multi-Level Concept Splitting (MLCS), which discovers multi-level concept hierarchies from only top-level supervision, and Deep-HiCEMs, an architecture that represents these discovered hierarchies and enables interventions at multiple levels of abstraction. Experiments across multiple datasets show that MLCS discovers human-interpretable concepts absent during training and that Deep-HiCEMs maintain high accuracy while supporting test-time concept interventions that can improve task performance.

---

### Mitigating Frequency Learning Bias in Quantum Models via Multi-Stage Residual Learning

Authors: Ammar Daskin
Published: 2026-03-10
Categories: quant-ph, cs.LG
arXiv: https://arxiv.org/abs/2603.10083v2

Quantum machine learning models based on parameterized circuits can be viewed as Fourier series approximators. However, they often struggle to learn functions with multiple frequency components, particularly high-frequency or non-dominant ones; a phenomenon we term the quantum Fourier parameterization bias. Inspired by recent advances in classical Fourier neural operators (FNOs), we adapt the multi-stage residual learning idea to the quantum domain, iteratively training additional quantum modules on the residuals of previous stages. We evaluate our method on a synthetic benchmark composed of spatially localized frequency components with diverse envelope shapes (Gaussian, Lorentzian, triangular). Systematic experiments show that the number of qubits, the encoding scheme, and residual learning are all crucial for resolving multiple frequencies; residual learning alone can improve test MSE significantly over a single-stage baseline trained for the same total number of epochs. Our work provides a practical framework for enhancing the spectral expressivity of quantum models and offers new insights into their frequency-learning behavior.

---

### Amnesia: Adversarial Semantic Layer Specific Activation Steering in Large Language Models

Authors: Ali Raza, Gurang Gupta, Nikolay Matyunin, Jibesh Patra
Published: 2026-03-10
Categories: cs.CR, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2603.10080v2

Warning: This article includes red-teaming experiments, which contain examples of compromised LLM responses that may be offensive or upsetting. Large Language Models (LLMs) have the potential to create harmful content, such as generating sophisticated phishing emails and assisting in writing code of harmful computer viruses. Thus, it is crucial to ensure their safe and responsible response generation. To reduce the risk of generating harmful or irresponsible content, researchers have developed techniques such as reinforcement learning with human feedback to align LLM's outputs with human values and preferences. However, it is still undetermined whether such measures are sufficient to prevent LLMs from generating interesting responses. In this study, we propose Amnesia, a lightweight activation-space adversarial attack that manipulates internal transformer states to bypass existing safety mechanisms in open-weight LLMs. Through experimental analysis on state-of-the-art, open-weight LLMs, we demonstrate that our attack effectively circumvents existing safeguards, enabling the generation of harmful content without the need for any fine-tuning or additional training. Our experiments on benchmark datasets show that the proposed attack can induce various antisocial behaviors in LLMs. These findings highlight the urgent need for more robust security measures in open-weight LLMs and underscore the importance of continued research to prevent their potential misuse.

---

### Large Spikes in Stochastic Gradient Descent: A Large-Deviations View

Authors: Benjamin Gess, Daniel Heydecker
Published: 2026-03-10
Categories: cs.LG, math.PR
arXiv: https://arxiv.org/abs/2603.10079v2

Large loss spikes in stochastic gradient descent are studied through a rigorous large-deviations analysis for a shallow, fully connected network in the NTK scaling. In contrast to full-batch gradient descent, the catapult phase is shown to split into inflationary and deflationary regimes, determined by an explicit log-drift criterion. In both cases, large spikes are shown to be at least polynomially likely. In addition, these spikes are shown to be the dominant mechanism by which sharp minima are escaped and curvature is reduced, thereby favouring flatter solutions. Corresponding results are also obtained for certain ReLU networks, and implications for curriculum learning are derived.

---

### Stochastic Port-Hamiltonian Neural Networks: Universal Approximation with Passivity Guarantees

Authors: Luca Di Persio, Matthias Ehrhardt, Youness Outaleb
Published: 2026-03-10
Categories: cs.LG, math.PR
arXiv: https://arxiv.org/abs/2603.10078v1

Stochastic port-Hamiltonian systems represent open dynamical systems with dissipation, inputs, and stochastic forcing in an energy based form. We introduce stochastic port-Hamiltonian neural networks, SPH-NNs, which parameterize the Hamiltonian with a feedforward network and enforce skew symmetry of the interconnection matrix and positive semidefiniteness of the dissipation matrix. For Itô dynamics we establish a weak passivity inequality in expectation under an explicit generator condition, stated for a stopped process on a compact set. We also prove a universal approximation result showing that, on any compact set and finite horizon, SPH-NNs approximate the coefficients of a target stochastic port-Hamiltonian system with $C^2$ accuracy of the Hamiltonian and yield coupled solutions that remain close in mean square up to the exit time. Experiments on noisy mass spring, Duffing, and Van der Pol oscillators show improved long horizon rollouts and reduced energy error relative to a multilayer perceptron baseline.

---

### TASER: Task-Aware Spectral Energy Refine for Backdoor Suppression in UAV Swarms Decentralized Federated Learning

Authors: Sizhe Huang, Shujie Yang
Published: 2026-03-10
Categories: cs.CR, cs.AI
arXiv: https://arxiv.org/abs/2603.10075v1

As backdoor attacks in UAV-based decentralized federated learning (DFL) grow increasingly stealthy and sophisticated, existing defenses have likewise escalated in complexity. Yet these defenses, which rely heavily on outlier detection, remain vulnerable to carefully crafted backdoors. In UAV-DFL, the lack of global coordination and limited resources further render outlier-based defenses impractical. Against this backdrop, gradient spectral analysis offers a promising alternative. While prior work primarily leverages low-frequency coefficients for pairwise comparisons, it neglects to analyze the intrinsic spectral characteristics of backdoor gradients. Through empirical analysis of existing stealthy attacks, we reveal a key insight: the more effort attackers invest in mimicking benign behaviors, the more distinct the spectral concentration becomes. Motivated by this, we propose Task-Aware Spectral Energy Refine (TASER) -- a decentralized defense framework. To our knowledge, this is the first efficient backdoor defense that utilizes spectral concentration instead of complex outlier detection, enabling mitigation of stealthy attacks by structurally disrupting the backdoor task. To suppress the backdoor task, TASER preserves main-task-relevant frequency coefficients and discards others. We provide theoretical guarantees and demonstrate through experiments that TASER remains effective against stealthy backdoor attacks that bypass outlier-based defenses, achieving attack success rate below 20% and accuracy loss under 5%.

---

### Marginals Before Conditionals

Authors: Mihir Sahasrabudhe
Published: 2026-03-10
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2603.10074v1

We construct a minimal task that isolates conditional learning in neural networks: a surjective map with K-fold ambiguity, resolved by a selector token z, so H(A | B) = log K while H(A | B, z) = 0. The model learns the marginal P(A | B) first, producing a plateau at exactly log K, before acquiring the full conditional in a sharp, collective transition. The plateau has a clean decomposition: height = log K (set by ambiguity), duration = f(D) (set by dataset size D, not K). Gradient noise stabilizes the marginal solution: higher learning rates monotonically slow the transition (3.6* across a 7* η range at fixed throughput), and batch-size reduction delays escape, consistent with an entropic force opposing departure from the low-gradient marginal. Internally, a selector-routing head assembles during the plateau, leading the loss transition by ~50% of the waiting time. This is the Type 2 directional asymmetry of Papadopoulos et al. (2024), measured dynamically: we track the excess risk from log K to zero and characterize what stabilizes it, what triggers its collapse, and how long it takes.

---

### Why LLMs Fail: A Failure Analysis and Partial Success Measurement for Automated Security Patch Generation

Authors: Amir Al-Maamari
Published: 2026-03-10
Categories: cs.CR, cs.AI
arXiv: https://arxiv.org/abs/2603.10072v1

Large Language Models (LLMs) show promise for Automated Program Repair (APR), yet their effectiveness on security vulnerabilities remains poorly characterized. This study analyzes 319 LLM-generated security patchesacross 64 Java vulnerabilities from the Vul4J benchmark. Using tri-axis evaluation (compilation, security via PoV tests, functionality via test suites), the analysis reveals that only 24.8% of patches achieve full correctness, while 51.4% fail both security and functionality. The dominant failure mode is semantic misunderstanding: LLMs produce syntactically valid code but apply incorrect repair strategies. The proposed Security Repair Score (SRS) quantifies this gap, showing LLMs preserve functionality (mean 0.832) but struggle with security (mean 0.251). Vulnerability type strongly predicts difficulty, with fix rates ranging from 0% (input validation) to 45% (infinite loop). These findings demonstrate that LLM security patches require rigorous validation before deployment.

---

### Dissecting Chronos: Sparse Autoencoders Reveal Causal Feature Hierarchies in Time Series Foundation Models

Authors: Anurag Mishra
Published: 2026-03-10
Categories: cs.LG, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2603.10071v1

Time series foundation models (TSFMs) are increasingly deployed in high-stakes domains, yet their internal representations remain opaque. We present the first application of sparse autoencoders (SAEs) to a TSFM, training TopK SAEs on activations of Chronos-T5-Large (710M parameters) across six layers. Through 392 single-feature ablation experiments, we establish that every ablated feature produces a positive CRPS degradation, confirming causal relevance. Our analysis reveals a depth-dependent hierarchy: early encoder layers encode low-level frequency features, the mid-encoder concentrates causally critical change-detection features, and the final encoder compresses a rich but less causally important taxonomy of temporal concepts. The most critical features reside in the mid-encoder (max single-feature Delta CRPS = 38.61), not in the semantically richest final encoder layer, where progressive ablation paradoxically improves forecast quality. These findings demonstrate that mechanistic interpretability transfers effectively to TSFMs and that Chronos-T5 relies on abrupt-dynamics detection rather than periodic pattern recognition.

---