# Research Papers — 2026-04 · bundle page 192 of 233 > Full-text markdown of papers #9551–9600 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/192 [← Previous bundle](https://aipod.dev/papers/archive/2026-04/page/191.md) [Next bundle →](https://aipod.dev/papers/archive/2026-04/page/193.md) ### Gradient-Controlled Decoding: A Safety Guardrail for LLMs with Dual-Anchor Steering Authors: Purva Chiniya, Kevin Scaria, Sagar Chaturvedi Published: 2026-04-06 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.05179v1 Large language models (LLMs) remain susceptible to jailbreak and direct prompt-injection attacks, yet the strongest defensive filters frequently over-refuse benign queries and degrade user experience. Previous work on jailbreak and prompt injection detection such as GradSafe, detects unsafe prompts with a single "accept all" anchor token, but its threshold is brittle and it offers no deterministic guarantee that harmful content will not be emitted once decoding begins. We introduce Gradient-Controlled Decoding (GCD), a training-free guardrail that combines an acceptance anchor token ("Sure") and refusal anchor token ("Sorry") tightening the decision boundary and significantly lowering false positives. In the mitigation stage, if a prompt is flagged, GCD preset-injects one or two refusal tokens ("Sorry, I can't...") before autoregressive decoding resumes, guaranteeing first-token safety regardless of sampling strategy. On ToxicChat, XSTest-v2, and AdvBench, GCD reduces false positives by 52% vs. GradSafe at comparable recall, lowers attack success rate by up to 10% vs. the strongest decoding-only baseline, adds under 15-20 ms latency on an average on V100 instances, transfers to LLaMA-2-7B, Mixtral-8x7B, and Qwen-2-7B, and requires only 20 demonstration templates. --- ### Graph Signal Diffusion Models for Wireless Resource Allocation Authors: Yigit Berkay Uslu, Samar Hadou, Shirin Saeedi Bidokhti, Alejandro Ribeiro Published: 2026-04-06 Categories: eess.SP, cs.IT, cs.LG arXiv: https://arxiv.org/abs/2604.05175v2 We consider constrained ergodic resource optimization in wireless networks with graph-structured interference. We train a diffusion model policy to match expert conditional distributions over resource allocations. By leveraging a primal-dual (expert) algorithm, we generate primal iterates that serve as draws from the corresponding expert conditionals for each training network instance. We view the allocations as stochastic graph signals supported on known channel state graphs. We implement the diffusion model architecture as a U-Net hierarchy of graph neural network (GNN) blocks, conditioned on the channel states and additional node states. At inference, the learned generative model amortizes the iterative expert policy by directly sampling allocation vectors from the near-optimal conditional distributions. In a power-control case study, we show that time-sharing the generated power allocations achieves near-optimal ergodic sum-rate utility and near-feasible ergodic minimum-rates, with strong generalization and transferability across network states. --- ### ClawsBench: Evaluating Capability and Safety of LLM Productivity Agents in Simulated Workspaces Authors: Xiangyi Li, Kyoung Whan Choe, Yimin Liu, Xiaokun Chen, Chujun Tao, Bingran You, Wenbo Chen, Zonglin Di, Jiankai Sun, Shenghan Zheng, Jiajun Bao, Yuanli Wang et al. Published: 2026-04-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.05172v2 Large language model (LLM) agents are increasingly deployed to automate productivity tasks (e.g., email, scheduling, document management), but evaluating them on live services is risky due to potentially irreversible changes. Existing benchmarks rely on simplified environments and fail to capture realistic, stateful, multi-service workflows. We introduce ClawsBench, a benchmark for evaluating and improving LLM agents in realistic productivity settings. It includes five high-fidelity mock services (Gmail, Slack, Google Calendar, Google Docs, Google Drive) with full state management and deterministic snapshot/restore, along with 44 structured tasks covering single-service, cross-service, and safety-critical scenarios. We decompose agent scaffolding into two independent levers (domain skills that inject API knowledge via progressive disclosure, and a meta prompt that coordinates behavior across services) and vary both to measure their separate and combined effects. Experiments across 6 models, 4 agent harnesses, and 33 conditions show that with full scaffolding, agents achieve task success rates of 39-64% but exhibit unsafe action rates of 7-33%. On OpenClaw, the top five models fall within a 10 percentage-point band on task success (53-63%), with unsafe action rates from 7% to 23% and no consistent ordering between the two metrics. We identify eight recurring patterns of unsafe behavior, including multi-step sandbox escalation and silent contract modification. We release the trajectories and future dataset at https://clawsbench.com. --- ### Modality-Aware and Anatomical Vector-Quantized Autoencoding for Multimodal Brain MRI Authors: Mingjie Li, Edward Kim, Yue Zhao, Ehsan Adeli, Kilian M. Pohl Published: 2026-04-06 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2604.05171v1 Learning a robust Variational Autoencoder (VAE) is a fundamental step for many deep learning applications in medical image analysis, such as MRI synthesizes. Existing brain VAEs predominantly focus on single-modality data (i.e., T1-weighted MRI), overlooking the complementary diagnostic value of other modalities like T2-weighted MRIs. Here, we propose a modality-aware and anatomically grounded 3D vector-quantized VAE (VQ-VAE) for reconstructing multi-modal brain MRIs. Called NeuroQuant, it first learns a shared latent representation across modalities using factorized multi-axis attention, which can capture relationships between distant brain regions. It then employs a dual-stream 3D encoder that explicitly separates the encoding of modality-invariant anatomical structures from modality-dependent appearance. Next, the anatomical encoding is discretized using a shared codebook and combined with modality-specific appearance features via Feature-wise Linear Modulation (FiLM) during the decoding phase. This entire approach is trained using a joint 2D/3D strategy in order to account for the slice-based acquisition of 3D MRI data. Extensive experiments on two multi-modal brain MRI datasets demonstrate that NeuroQuant achieves superior reconstruction fidelity compared to existing VAEs, enabling a scalable foundation for downstream generative modeling and cross-modal brain image analysis. --- ### Instruction-Tuned LLMs for Parsing and Mining Unstructured Logs on Leadership HPC Systems Authors: Ahmad Maroof Karimi, Jong Youl Choi, Charles Qing Cao, Awais Khan Published: 2026-04-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.05168v1 Leadership-class HPC systems generate massive volumes of heterogeneous, largely unstructured system logs. Because these logs originate from diverse software, hardware, and runtime layers, they exhibit inconsistent formats, making structure extraction and pattern discovery extremely challenging. Therefore, robust log parsing and mining is critical to transform this raw telemetry into actionable insights that reveal operational patterns, diagnose anomalies, and enable reliable, efficient, and scalable system analysis. Recent advances in large language models (LLMs) offer a promising new direction for automated log understanding in leadership-class HPC environments. To capitalize on this opportunity, we present a domain-adapted, instruction-following, LLM-driven framework that leverages chain-of-thought (CoT) reasoning to parse and structure HPC logs with high fidelity. Our approach combines domain-specific log-template data with instruction-tuned examples to fine-tune an 8B-parameter LLaMA model tailored for HPC log analysis. We develop a hybrid fine-tuning methodology that adapts a general-purpose LLM to domain-specific log data, enabling privacy-preserving, locally deployable, fast, and energy-efficient log-mining approach. We conduct experiments on a diverse set of log datasets from the LogHub repository. The evaluation confirms that our approach achieves parsing accuracy on par with significantly larger models, such as LLaMA 70B and Anthropic's Claude. We further validate the practical utility of our fine-tuned LLM model by parsing over 600 million production logs from the Frontier supercomputer over a four-week window, uncovering critical patterns in temporal dynamics, node-level anomalies, and workload-error log correlations. --- ### From Use to Oversight: How Mental Models Influence User Behavior and Output in AI Writing Assistants Authors: Shalaleh Rismani, Su Lin Blodgett, Q. Vera Liao, Alexandra Olteanu, AJung Moon Published: 2026-04-06 Categories: cs.HC, cs.AI arXiv: https://arxiv.org/abs/2604.05166v1 AI-based writing assistants are ubiquitous, yet little is known about how users' mental models shape their use. We examine two types of mental models -- functional or related to what the system does, and structural or related to how the system works -- and how they affect control behavior -- how users request, accept, or edit AI suggestions as they write -- and writing outcomes. We primed participants ($N = 48$) with different system descriptions to induce these mental models before asking them to complete a cover letter writing task using a writing assistant that occasionally offered preconfigured ungrammatical suggestions to test whether the mental models affected participants' critical oversight. We find that while participants in the structural mental model condition demonstrate a better understanding of the system, this can have a backfiring effect: while these participants judged the system as more usable, they also produced letters with more grammatical errors, highlighting a complex relationship between system understanding, trust, and control in contexts that require user oversight of error-prone AI outputs. --- ### Learning to Focus: CSI-Free Hierarchical MARL for Reconfigurable Reflectors Authors: Hieu Le, Mostafa Ibrahim, Oguz Bedir, Jian Tao, Sabit Ekin Published: 2026-04-06 Categories: cs.AI, eess.SP arXiv: https://arxiv.org/abs/2604.05165v3 Reconfigurable Intelligent Surfaces (RIS) have the potential to engineer smart radio environments for next-generation millimeter-wave (mmWave) networks. However, the prohibitive computational overhead of Channel State Information (CSI) estimation and the dimensionality explosion inherent in centralized optimization severely hinder practical large-scale deployments. To overcome these bottlenecks, we introduce a per-element CSI-free paradigm powered by a Hierarchical Multi-Agent Reinforcement Learning (HMARL) architecture to control mechanically reconfigurable reflective surfaces. By substituting pilot-based channel estimation for each element of the device with accessible user localization data, our framework leverages spatial intelligence for macro-scale wave propagation management. The control problem is decomposed into a two-tier neural architecture: a high-level controller executes temporally extended, discrete user-to-reflector allocations, while low-level controllers autonomously optimize continuous focal points using Multi-Agent Proximal Policy Optimization (MAPPO) under a Centralized Training with Decentralized Execution (CTDE) scheme. Comprehensive deterministic ray-tracing evaluations in an indoor mmWave scenario demonstrate that this hierarchical framework achieves received signal strength indicator (RSSI) improvements of up to 7.79 dB over centralized Proximal Policy Optimization (PPO) baselines. Furthermore, the system maintains resilient beam-focusing performance under practical sub-meter localization tracking errors for up to four users and two reflector arrays. By eliminating execution-time CSI overhead while preserving high-fidelity signal redirection, this work provides a scalable and cost-effective step toward intelligent indoor wireless environments. --- ### Learning to Focus: CSI-Free Hierarchical MARL for Reconfigurable Reflectors Authors: Hieu Le, Mostafa Ibrahim, Oguz Bedir, Jian Tao, Sabit Ekin Published: 2026-04-06 Categories: cs.AI, eess.SP arXiv: https://arxiv.org/abs/2604.05165v2 Reconfigurable Intelligent Surfaces (RIS) has a potential to engineer smart radio environments for next-generation millimeter-wave (mmWave) networks. However, the prohibitive computational overhead of Channel State Information (CSI) estimation and the dimensionality explosion inherent in centralized optimization severely hinder practical large-scale deployments. To overcome these bottlenecks, we introduce a ``CSI-free" paradigm powered by a Hierarchical Multi-Agent Reinforcement Learning (HMARL) architecture to control mechanically reconfigurable reflective surfaces. By substituting pilot-based channel estimation with accessible user localization data, our framework leverages spatial intelligence for macro-scale wave propagation management. The control problem is decomposed into a two-tier neural architecture: a high-level controller executes temporally extended, discrete user-to-reflector allocations, while low-level controllers autonomously optimize continuous focal points utilizing Multi-Agent Proximal Policy Optimization (MAPPO) under a Centralized Training with Decentralized Execution (CTDE) scheme. Comprehensive deterministic ray-tracing evaluations demonstrate that this hierarchical framework achieves massive RSSI improvements of up to 7.79 dB over centralized baselines. Furthermore, the system exhibits robust multi-user scalability and maintains highly resilient beam-focusing performance under practical sub-meter localization tracking errors. By eliminating CSI overhead while maintaining high-fidelity signal redirection, this work establishes a scalable and cost-effective blueprint for intelligent wireless environments. --- ### Not All Turns Are Equally Hard: Adaptive Thinking Budgets For Efficient Multi-Turn Reasoning in Agents Authors: Neharika Jali, Anupam Nayak, Gauri Joshi Published: 2026-04-06 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.05164v3 As LLM reasoning performance plateaus, improving inference-time compute efficiency is crucial to mitigate overthinking and long thinking traces even for simple queries. Prior approaches including length regularization, adaptive routing, and difficulty-based budget allocation primarily focus on single-turn settings and fail to address the sequential dependencies inherent in multi-turn reasoning. In this work, we formulate multi-turn reasoning as a sequential compute allocation problem and model it as a multi-objective Markov Decision Process. We propose TAB: Turn-Adaptive Budgets, a budget allocation policy trained via Group Relative Policy Optimization (GRPO) that learns to maximize task accuracy while respecting global per-problem token constraints. Consequently, TAB takes as input the conversation history and learns to adaptively allocate smaller budgets to easier turns and save appropriate number of tokens for the crucial harder reasoning steps. Our experiments on diverse agentic benchmarks demonstrate that TAB achieves a superior accuracy-cost tradeoff saving up to 35% tokens and reducing latency up to 30% while maintaining accuracy over static and off-the-shelf LLM budget baselines. Further, for systems where a plan of all turns is available apriori, we propose TAB All-SubQ, a budget allocation policy that budgets tokens based on the conversation history and entire plan saving up to 40% tokens over baselines. --- ### Not All Turns Are Equally Hard: Adaptive Thinking Budgets For Efficient Multi-Turn Reasoning Authors: Neharika Jali, Anupam Nayak, Gauri Joshi Published: 2026-04-06 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.05164v2 As LLM reasoning performance plateau, improving inference-time compute efficiency is crucial to mitigate overthinking and long thinking traces even for simple queries. Prior approaches including length regularization, adaptive routing, and difficulty-based budget allocation primarily focus on single-turn settings and fail to address the sequential dependencies inherent in multi-turn reasoning. In this work, we formulate multi-turn reasoning as a sequential compute allocation problem and model it as a multi-objective Markov Decision Process. We propose TAB: Turn-Adaptive Budgets, a budget allocation policy trained via Group Relative Policy Optimization (GRPO) that learns to maximize task accuracy while respecting global per-problem token constraints. Consequently, TAB takes as input the conversation history and learns to adaptively allocate smaller budgets to easier turns and save appropriate number of tokens for the crucial harder reasoning steps. Our experiments on mathematical reasoning benchmarks demonstrate that TAB achieves a superior accuracy-tokens tradeoff saving up to 35% tokens while maintaining accuracy over static and off-the-shelf LLM budget baselines. Further, for systems where a plan of all sub-questions is available apriori, we propose TAB All-SubQ, a budget allocation policy that budgets tokens based on the conversation history and all past and future sub-questions saving up to 40% tokens over baselines. --- ### What Makes a Good Response? An Empirical Analysis of Quality in Qualitative Interviews Authors: Jonathan Ivey, Anjalie Field, Ziang Xiao Published: 2026-04-06 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.05163v1 Qualitative interviews provide essential insights into human experiences when they elicit high-quality responses. While qualitative and NLP researchers have proposed various measures of interview quality, these measures lack validation that high-scoring responses actually contribute to the study's goals. In this work, we identify, implement, and evaluate 10 proposed measures of interview response quality to determine which are actually predictive of a response's contribution to the study findings. To conduct our analysis, we introduce the Qualitative Interview Corpus, a newly constructed dataset of 343 interview transcripts with 16,940 participant responses from 14 real research projects. We find that direct relevance to a key research question is the strongest predictor of response quality. We additionally find that two measures commonly used to evaluate NLP interview systems, clarity and surprisal-based informativeness, are not predictive of response quality. Our work provides analytic insights and grounded, scalable metrics to inform the design of qualitative studies and the evaluation of automated interview systems. --- ### Bypassing the CSI Bottleneck: MARL-Driven Spatial Control for Reflector Arrays Authors: Hieu Le, Oguz Bedir, Mostafa Ibrahim, Jian Tao, Sabit Ekin Published: 2026-04-06 Categories: cs.AI, eess.SP arXiv: https://arxiv.org/abs/2604.05162v1 Reconfigurable Intelligent Surfaces (RIS) are pivotal for next-generation smart radio environments, yet their practical deployment is severely bottlenecked by the intractable computational overhead of Channel State Information (CSI) estimation. To bypass this fundamental physical-layer barrier, we propose an AI-native, data-driven paradigm that replaces complex channel modeling with spatial intelligence. This paper presents a fully autonomous Multi-Agent Reinforcement Learning (MARL) framework to control mechanically adjustable metallic reflector arrays. By mapping high-dimensional mechanical constraints to a reduced-order virtual focal point space, we deploy a Centralized Training with Decentralized Execution (CTDE) architecture. Using Multi-Agent Proximal Policy Optimization (MAPPO), our decentralized agents learn cooperative beam-focusing strategies relying on user coordinates, achieving CSI-free operation. High-fidelity ray-tracing simulations in dynamic non-line-of-sight (NLOS) environments demonstrate that this multi-agent approach rapidly adapts to user mobility, yielding up to a 26.86 dB enhancement over static flat reflectors and outperforming single-agent and hardware-constrained DRL baselines in both spatial selectivity and temporal stability. Crucially, the learned policies exhibit good deployment resilience, sustaining stable signal coverage even under 1.0-meter localization noise. These results validate the efficacy of MARL-driven spatial abstractions as a scalable, highly practical pathway toward AI-empowered wireless networks. --- ### Planning to Explore: Curiosity-Driven Planning for LLM Test Generation Authors: Alfonso Amayuelas, Firas Laakom, Piotr Piękos, Wenyi Wang, Yifan Xu, Yuhui Wang, Jürgen Schmidhuber, William Wang Published: 2026-04-06 Categories: cs.SE, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2604.05159v1 The use of LLMs for code generation has naturally extended to code testing and evaluation. As codebases grow in size and complexity, so does the need for automated test generation. Current approaches for LLM-based test generation rely on strategies that maximize immediate coverage gain, a greedy approach that plateaus on code where reaching deep branches requires setup steps that individually yield zero new coverage. Drawing on principles of Bayesian exploration, we treat the program's branch structure as an unknown environment, and an evolving coverage map as a proxy probabilistic posterior representing what the LLM has discovered so far. Our method, CovQValue, feeds the coverage map back to the LLM, generates diverse candidate plans in parallel, and selects the most informative plan by LLM-estimated Q-values, seeking actions that balance immediate branch discovery with future reachability. Our method outperforms greedy selection on TestGenEval Lite, achieving 51-77% higher branch coverage across three popular LLMs and winning on 77-84% of targets. In addition, we build a benchmark for iterative test generation, RepoExploreBench, where they achieve 40-74%. These results show the potential of curiosity-driven planning methods for LLM-based exploration, enabling more effective discovery of program behavior through sequential interaction --- ### Just Pass Twice: Efficient Token Classification with LLMs for Zero-Shot NER Authors: Ahmed Ewais, Ahmed Hashish, Amr Ali Published: 2026-04-06 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.05158v3 Large language models encode extensive world knowledge valuable for zero-shot named entity recognition. However, their causal attention mechanism, where tokens attend only to preceding context, prevents effective token classification when disambiguation requires future context. Existing approaches use LLMs generatively, prompting them to list entities or produce structured outputs, but suffer from slow autoregressive decoding, hallucinated entities, and formatting errors. We propose Just Pass Twice (JPT), a simple yet effective method that enables causal LLMs to perform discriminative token classification with full bidirectional context. Our key insight is that concatenating the input to itself lets each token in the second pass attend to the complete sentence, requiring no architectural modifications. We combine these representations with definition-guided entity embeddings for flexible zero-shot generalization. Our approach achieves state-of-the-art results on zero-shot NER benchmarks, surpassing the previous best method by +7.9 F1 on average across CrossNER and MIT benchmarks, being over 20x faster than comparable generative methods. Interactive demo and access to the model are available at https://witness-ai-jpt-ner.hf.space/ --- ### Just Pass Twice: Efficient Token Classification with LLMs for Zero-Shot NER Authors: Ahmed Ewais, Ahmed Hashish, Amr Ali Published: 2026-04-06 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.05158v2 Large language models encode extensive world knowledge valuable for zero-shot named entity recognition. However, their causal attention mechanism, where tokens attend only to preceding context, prevents effective token classification when disambiguation requires future context. Existing approaches use LLMs generatively, prompting them to list entities or produce structured outputs, but suffer from slow autoregressive decoding, hallucinated entities, and formatting errors. We propose Just Pass Twice (JPT), a simple yet effective method that enables causal LLMs to perform discriminative token classification with full bidirectional context. Our key insight is that concatenating the input to itself lets each token in the second pass attend to the complete sentence, requiring no architectural modifications. We combine these representations with definition-guided entity embeddings for flexible zero-shot generalization. Our approach achieves state-of-the-art results on zero-shot NER benchmarks, surpassing the previous best method by +7.9 F1 on average across CrossNER and MIT benchmarks, being over 20x faster than comparable generative methods. --- ### IntentScore: Intent-Conditioned Action Evaluation for Computer-Use Agents Authors: Rongqian Chen, Yu Li, Zeyu Fang, Sizhe Tang, Weidong Cao, Tian Lan Published: 2026-04-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.05157v3 Computer-Use Agents (CUAs) leverage large language models to execute GUI operations on desktop environments, yet they generate actions without evaluating action quality, leading to irreversible errors that cascade through subsequent steps. We propose IntentScore, a plan-aware reward model that learns to score candidate actions from 398K offline GUI interaction steps spanning three operating systems. IntentScore trains with two complementary objectives: contrastive alignment for state-action relevance and margin ranking for action correctness. Architecturally, it embeds each candidate's planning intent in the action encoder, enabling discrimination between candidates with similar actions but different rationales. IntentScore achieves 97.5% pairwise discrimination accuracy on held-out evaluation. Deployed as a re-ranker for Agent S3 on OSWorld, an environment entirely unseen during training, IntentScore improves task success rate by 6.9 points, demonstrating that reward estimation learned from heterogeneous offline trajectories generalizes to unseen agents and task distributions. --- ### Compiled AI: Deterministic Code Generation for LLM-Based Workflow Automation Authors: Geert Trooskens, Aaron Karlsberg, Anmol Sharma, Lamara De Brouwer, Max Van Puyvelde, Matthew Young, John Thickstun, Gil Alterovitz, Walter A. De Brouwer Published: 2026-04-06 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2604.05150v2 We study compiled AI, a paradigm in which large language models generate executable code artifacts during a compilation phase, after which workflows execute deterministically without further model invocation. This paradigm has antecedents in prior work on declarative pipeline optimization (DSPy) and hybrid neural-symbolic planning (LLM+P); our contribution is a systems-oriented study of its application to high-stakes enterprise workflows, with particular emphasis on healthcare settings where reliability and auditability are critical. By constraining generation to narrow business-logic functions embedded in validated templates, compiled AI trades runtime flexibility for predictability, auditability, cost efficiency, and reduced security exposure. We introduce (i) a system architecture for constrained LLM-based code generation, (ii) a four-stage generation-and-validation pipeline that converts probabilistic model output into production-ready code artifacts, and (iii) an evaluation framework measuring operational metrics including token amortization, determinism, reliability, security, and cost. We evaluate on two task types: function-calling (BFCL, n=400) and document intelligence (DocILE, n=5,680 invoices). On function-calling, compiled AI achieves 96% task completion with zero execution tokens, breaking even with runtime inference at approximately 17 transactions and reducing token consumption by 57x at 1,000 transactions. On document intelligence, our Code Factory variant matches Direct LLM on key field extraction (KILE: 80.0%) while achieving the highest line item recognition accuracy (LIR: 80.4%). Security evaluation across 135 test cases demonstrates 96.7% accuracy on prompt injection detection and 87.5% on static code safety analysis with zero false positives. --- ### EvolveRouter: Co-Evolving Routing and Prompt for Multi-Agent Question Answering Authors: Jiatan Huang, Zheyuan Zhang, Kaiwen Shi, Yanfang Ye, Chuxu Zhang Published: 2026-04-06 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.05149v1 Large language model agents often exhibit complementary strengths, making routing a promising approach for multi-agent question answering. However, existing routing methods remain limited in two important ways: they typically optimize over a fixed pool of agents without improving the agents themselves, and they often rely on rigid collaboration schemes that cannot adapt the number of participating agents to the query. We propose EvolveRouter, a trainable framework that addresses both limitations by jointly improving agent quality and collaboration structure. First, EvolveRouter couples graph-based query routing with targeted instruction refinement in a closed-loop co-evolution process, allowing router diagnostics to guide agent improvement while refined agents provide cleaner supervision for routing. Second, it introduces an adaptive inference strategy that dynamically determines the effective collaboration size for each query through router-weighted answer agreement. Together, these designs enable more capable and more efficient multi-agent reasoning. Experiments on five question answering benchmarks show that EvolveRouter consistently outperforms SOTA routing baselines in both F1 and exact match, while further analysis confirms the benefits of closed-loop refinement and adaptive collaboration. --- ### Lightweight True In-Pixel Encryption with FeFET Enabled Pixel Design for Secure Imaging Authors: Md Rahatul Islam Udoy, Diego Ferrer, Wantong Li, Kai Ni, Sumeet Kumar Gupta, Ahmedullah Aziz Published: 2026-04-06 Categories: cs.CV, cs.CR arXiv: https://arxiv.org/abs/2604.05147v1 Ensuring end-to-end security in image sensors has become essential as visual data can be exposed through multiple stages of the imaging pipeline. Advanced protection requires encryption to occur before pixel values appear on any readout lines. This work introduces a secure pixel sensor (SecurePix), a compact CMOS-compatible pixel architecture that performs true in-pixel encryption using a symmetric key realized through programmable, non-volatile multidomain polarization states of a ferroelectric field-effect transistor. The pixel and array operations are designed and simulated in HSPICE, while a 45 nm CMOS process design kit is used for layout drawing. The resulting layout confirms a pixel pitch of 2.33 x 3.01 um^2. Each pixel's non-volatile programming level defines its analog transfer characteristic, enabling the photodiode voltage to be converted into an encrypted analog output within the pixel. Full-image evaluation shows that ResNet-18 recognition accuracy drops from 99.29 percent to 9.58 percent on MNIST and from 91.33 percent to 6.98 percent on CIFAR-10 after encryption, indicating strong resistance to neural-network-based inference. Lookup-table-based inverse mapping enables recovery for authorized receivers using the same symmetric key. Based on HSPICE simulation, the SecurePix achieves a per-pixel programming power-delay product of 17 uW us and a per-pixel sensing power-delay product of 1.25 uW us, demonstrating low-overhead hardware-level protection. --- ### A mathematical theory of evolution for self-designing AIs Authors: Kenneth D Harris Published: 2026-04-06 Categories: cs.AI, cs.CY, q-bio.PE arXiv: https://arxiv.org/abs/2604.05142v2 As artificial intelligence systems (AIs) become increasingly produced by recursive self-improvement, a form of evolution may emerge, with the traits of AI systems shaped by the success of earlier AIs in designing and propagating their descendants. There is a rich mathematical theory modeling how behavioral traits are shaped by biological evolution, a key component of which is Fisher's fundamental theorem of natural selection, which describes conditions under which mean fitness (i.e. reproductive success) increases. AI evolution will be radically different to biological evolution: while DNA mutations are random and approximately reversible, AI self-design will be strongly directed. Here we develop a mathematical model of evolution for self-designing AIs, replacing a random walk of mutations with a directed tree of potential AI designs. Current AIs design their descendants, while humans control a fitness function allocating resources. In this model, fitness need not increase over time without further assumptions. However, assuming bounded fitness and an additional "$η$-locking" condition, we show that fitness concentrates on the maximum reachable value. We consider the implications of this for AI alignment, specifically for cases where fitness and human utility are not perfectly correlated. We show that if deception of human evaluators additively increases an AI's reproductive fitness beyond genuine capability, evolution will select for both capability and deception. This risk could be mitigated if reproduction is based on purely objective criteria, rather than human judgment. --- ### EffiPair: Improving the Efficiency of LLM-generated Code with Differential Execution Feedback Authors: Samira Hajizadeh, Suman Jana Published: 2026-04-06 Categories: cs.PL, cs.AI, cs.CL, cs.LG, cs.SE arXiv: https://arxiv.org/abs/2604.05137v2 Large language models (LLMs) can generate functionally correct programs that differ substantially in execution efficiency. Existing inference-time optimization methods typically refine each candidate using pointwise runtime or profiling feedback, which identifies how costly an implementation is or where the cost arises, but offers limited guidance on how the computation should change. We introduce DIFFERENTIAL EXECUTION FEEDBACK (DEF), which compares implementations that are nearby in structural program space but separated in performance, turning their execution and implementation differences into directional optimization evidence. We demonstrate DEF in EffiPair, a training-free, test-time framework that pairs structurally similar programs with different efficiencies, distills their relative execution behavior into compact feedback, and iteratively refines a candidate pool. Across EvalPerf, Mercury, and ENAMEL, using GPT-4o mini, DeepSeek-V4.1 Flash, and GPT-5 mini, EffiPair achieves the highest value on each benchmark's official efficiency metric in all nine model-benchmark settings under matched evaluation conditions. Moreover, two contrastive refinement rounds improve efficiency over the selected initial draft in every setting while preserving or improving Pass@1. These results demonstrate the effectiveness of relational execution feedback as a lightweight signal for test-time code optimization. --- ### EffiPair: Improving the Efficiency of LLM-generated Code with Relative Contrastive Feedback Authors: Samira Hajizadeh, Suman Jana Published: 2026-04-06 Categories: cs.PL, cs.AI, cs.CL, cs.LG, cs.SE arXiv: https://arxiv.org/abs/2604.05137v1 Large language models (LLMs) often generate code that is functionally correct but inefficient in runtime and memory. Prior approaches to improving code efficiency typically rely on absolute execution feedback, such as profiling a single program's runtime or memory usage, which is costly and provides weak guidance for refinement. We propose Relative Contrastive Feedback (RCF), an inference-time feedback mechanism that requires no model fine-tuning or parameter updates. RCF compares two structurally similar programs for the same task and highlights the differences associated with better efficiency. Building on this idea, we introduce EffiPair, an inference-time iterative refinement framework that operates entirely at test time by generating multiple candidate solutions, identifying informative program pairs with large efficiency gaps, summarizing their execution differences into lightweight feedback, and using this signal to produce more efficient solutions. By replacing isolated scalar feedback with pairwise contrastive comparisons, EffiPair provides more direct guidance while reducing profiling and prompting overhead. Experiments on code-efficiency benchmarks show that EffiPair consistently improves efficiency while preserving correctness. For instance, with DeepSeek-Chat V3.2, EffiPair achieves up to 1.5x speedup over generation without performance feedback, while reducing token usage by more than 90% compared to prior work. --- ### Non-monotonic causal discovery with Kolmogorov-Arnold Fuzzy Cognitive Maps Authors: Jose L. Salmeron Published: 2026-04-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.05136v1 Fuzzy Cognitive Maps constitute a neuro-symbolic paradigm for modeling complex dynamic systems, widely adopted for their inherent interpretability and recurrent inference capabilities. However, the standard FCM formulation, characterized by scalar synaptic weights and monotonic activation functions, is fundamentally constrained in modeling non-monotonic causal dependencies, thereby limiting its efficacy in systems governed by saturation effects or periodic dynamics. To overcome this topological restriction, this research proposes the Kolmogorov-Arnold Fuzzy Cognitive Map (KA-FCM), a novel architecture that redefines the causal transmission mechanism. Drawing upon the Kolmogorov-Arnold representation theorem, static scalar weights are replaced with learnable, univariate B-spline functions located on the model edges. This fundamental modification shifts the non-linearity from the nodes' aggregation phase directly to the causal influence phase. This modification allows for the modeling of arbitrary, non-monotonic causal relationships without increasing the graph density or introducing hidden layers. The proposed architecture is validated against both baselines (standard FCM trained with Particle Swarm Optimization) and universal black-box approximators (Multi-Layer Perceptron) across three distinct domains: non-monotonic inference (Yerkes-Dodson law), symbolic regression, and chaotic time-series forecasting. Experimental results demonstrate that KA-FCMs significantly outperform conventional architectures and achieve competitive accuracy relative to MLPs, while preserving graph- based interpretability and enabling the explicit extraction of mathematical laws from the learned edges. --- ### SenseAI: A Human-in-the-Loop Dataset for RLHF-Aligned Financial Sentiment Reasoning Authors: Berny Kabalisa Published: 2026-04-06 Categories: cs.CL, cs.CE arXiv: https://arxiv.org/abs/2604.05135v1 We introduce SenseAI, a human-in-the-loop (HITL) validated financial sentiment dataset designed to capture not only model outputs but the full reasoning process behind them. Unlike existing resources, SenseAI incorporates reasoning chains, confidence scores, human correction signals, and real-world market outcomes, providing a structure aligned with Reinforcement Learning from Human Feedback (RLHF) paradigms. The dataset consists of 1,439 labelled data points across 40 US-listed equities and 13 financial data categories, enabling direct integration into modern LLM fine-tuning pipelines. Through analysis, we identify several systematic patterns in model behavior, including a novel failure mode we term Latent Reasoning Drift, where models introduce information not grounded in the input, as well as consistent confidence miscalibration and forward projection tendencies. These findings suggest that LLM errors in financial reasoning are not random but occur within a predictable and correctable regime, supporting the use of structured HITL data for targeted model improvement. We discuss implications for financial AI systems and highlight opportunities for applying SenseAI in model evaluation and alignment. --- ### How Reasoning Evolves from Post-Training Data: An Empirical Study Using Chess Authors: Lucas Dionisopoulos, Nicklas Majamaki, Prithviraj Ammanabrolu Published: 2026-04-06 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.05134v2 We study how reasoning evolves in a language model -- from supervised fine-tuning (SFT) to reinforcement learning (RL) -- by analyzing how a set of theoretically-inspired datasets influences language model performance in chess. We find that fine-tuning a model to directly predict the best move leads to effective RL and the strongest downstream performance -- however, the RL stage elicits \textit{unfaithful} reasoning (reasoning inconsistent with the chosen move). Alternatively, training on multi-move trajectories yields comparable downstream performance with faithful reasoning and more stable RL. We analyze multiple qualitative and quantitative measures and highlight how these evolve from SFT through RL; we find several SFT-checkpoint metrics -- spanning evaluation performance, hallucination rates, and reasoning quality -- to be predictive of post-RL model performance. Finally, we ground our results with an experiment measuring \textit{chess information density} in our custom datasets. We release models as well as training data, evaluations, and code that allowed us to surpass leading open-source reasoning models in chess with a 7B-parameter model. Code, models, and data are available at https://github.com/lucasdino/lang-chess. --- ### No Coin Left Behind: Maximizing Strategic Surplus Against No-Regret Dynamics Authors: Yiheng Su, Emmanouil-Vasileios Vlatakis-Gkaragkounis Published: 2026-04-06 Categories: cs.GT, cs.LG arXiv: https://arxiv.org/abs/2604.05129v2 We investigate the strategic surplus obtainable against a Follow-the-Regularized-Leader (FTRL) learner with constant step size $η$ in $n\times m$ two-player zero-sum games played over $T$ rounds against a clairvoyant optimizer. In contrast with prior analysis, we show that the extraction of such regret-scale surplus is an inherent feature of the FTRL family, rather than an artifact of specific instantiations. First, for a fixed max-min optimizer, we establish a sweeping law of order $Ω(N_{\mathrm{sub}}/η)$, proving that utility surplus scales with the number of the learner's suboptimal actions $N$ and vanishes in their absence. Second, for an alternating optimizer, a surplus of $Ω(ηT/\mathrm{poly}(n,m))$ can be guaranteed regardless of the equilibrium structure, with high probability, in random games. Our analysis uncovers a sharp geometric dichotomy: non-steep regularizers allow the optimizer to realize the maximal transient surplus via finite-time elimination of suboptimal actions, whereas steep regularizers introduce a vanishing tail correction that can delay surplus saturation. Finally, we discuss whether this leverage persists under bilateral payoff uncertainty and propose a susceptibility measure quantifying which regularizers are most vulnerable to learner-aware strategic steering. --- ### Offline RL for Adaptive Policy Retrieval in Prior Authorization Authors: Ruslan Sharifullin, Maxim Gorshkov, Hannah Clay Published: 2026-04-06 Categories: cs.IR, cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2604.05125v1 Prior authorization (PA) requires interpretation of complex and fragmented coverage policies, yet existing retrieval-augmented systems rely on static top-$K$ strategies with fixed numbers of retrieved sections. Such fixed retrieval can be inefficient and gather irrelevant or insufficient information. We model policy retrieval for PA as a sequential decision-making problem, formulating adaptive retrieval as a Markov Decision Process (MDP). In our system, an agent iteratively selects policy chunks from a top-$K$ candidate set or chooses to stop and issue a decision. The reward balances decision correctness against retrieval cost, capturing the trade-off between accuracy and efficiency. We train policies using Conservative Q-Learning (CQL), Implicit Q-Learning (IQL), and Direct Preference Optimization (DPO) in an offline RL setting on logged trajectories generated from baseline retrieval strategies over synthetic PA requests derived from publicly available CMS coverage data. On a corpus of 186 policy chunks spanning 10 CMS procedures, CQL achieves 92% decision accuracy (+30 percentage points over the best fixed-$K$ baseline) via exhaustive retrieval, while IQL matches the best baseline accuracy using 44% fewer retrieval steps and achieves the only positive episodic return among all policies. Transition-level DPO matches CQL's 92% accuracy while using 47% fewer retrieval steps (10.6 vs. 20.0), occupying a "selective-accurate" region on the Pareto frontier that dominates both CQL and BC. A behavioral cloning baseline matches CQL, confirming that advantage-weighted or preference-based policy extraction is needed to learn selective retrieval. Lambda ablation over step costs $λ\in \{0.05, 0.1, 0.2\}$ reveals a clear accuracy-efficiency inflection: only at $λ= 0.2$ does CQL transition from exhaustive to selective retrieval. --- ### Governance-Aware Agent Telemetry for Closed-Loop Enforcement in Multi-Agent AI Systems Authors: Anshul Pathak, Nishant Jain Published: 2026-04-06 Categories: cs.MA, cs.LG arXiv: https://arxiv.org/abs/2604.05119v1 Enterprise multi-agent AI systems produce thousands of inter-agent interactions per hour, yet existing observability tools capture these dependencies without enforcing anything. OpenTelemetry and Langfuse collect telemetry but treat governance as a downstream analytics concern, not a real-time enforcement target. The result is an "observe-but-do-not-act" gap where policy violations are detected only after damage is done. We present Governance-Aware Agent Telemetry (GAAT), a reference architecture that closes the loop between telemetry collection and automated policy enforcement for multi-agent systems. GAAT introduces (1) a Governance Telemetry Schema (GTS) extending OpenTelemetry with governance attributes; (2) a real-time policy violation detection engine using OPA-compatible declarative rules under sub-200 ms latency; (3) a Governance Enforcement Bus (GEB) with graduated interventions; and (4) a Trusted Telemetry Plane with cryptographic provenance. --- ### Watch Before You Answer: Learning from Visually Grounded Post-Training Authors: Yuxuan Zhang, EunJeong Hwang, Huaisong Zhang, Penghui Du, Yiming Jia, Dongfu Jiang, Xuan He, Shenhui Zhang, Ping Nie, Peter West, Kelsey R. Allen Published: 2026-04-06 Categories: cs.CV, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2604.05117v1 It is critical for vision-language models (VLMs) to comprehensively understand visual, temporal, and textual cues. However, despite rapid progress in multimodal modeling, video understanding performance still lags behind text-based reasoning. In this work, we find that progress is even worse than previously assumed: commonly reported long video understanding benchmarks contain 40-60% of questions that can be answered using text cues alone. Furthermore, we find that these issues are also pervasive in widely used post-training datasets, potentially undercutting the ability of post-training to improve VLM video understanding performance. Guided by this observation, we introduce VidGround as a simple yet effective solution: using only the actual visually grounded questions without any linguistic biases for post-training. When used in tandem with RL-based post-training algorithms, this simple technique improves performance by up to 6.2 points relative to using the full dataset, while using only 69.1% of the original post-training data. Moreover, we show that data curation with a simple post-training algorithm outperforms several more complex post-training techniques, highlighting that data quality is a major bottleneck for improving video understanding in VLMs. These results underscore the importance of curating post-training data and evaluation benchmarks that truly require visual grounding to advance the development of more capable VLMs. Project page: http://vidground.etuagi.com. --- ### Uncertainty-Guided Latent Diagnostic Trajectory Learning for Sequential Clinical Diagnosis Authors: Xuyang Shen, Haoran Liu, Dongjin Song, Martin Renqiang Min Published: 2026-04-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.05116v1 Clinical diagnosis requires sequential evidence acquisition under uncertainty. However, most Large Language Model (LLM) based diagnostic systems assume fully observed patient information and therefore do not explicitly model how clinical evidence should be sequentially acquired over time. Even when diagnosis is formulated as a sequential decision process, it is still challenging to learn effective diagnostic trajectories. This is because the space of possible evidence-acquisition paths is relatively large, while clinical datasets rarely provide explicit supervision information for desirable diagnostic paths. To this end, we formulate sequential diagnosis as a Latent Diagnostic Trajectory Learning (LDTL) framework based on a planning LLM agent and a diagnostic LLM agent. For the diagnostic LLM agent, diagnostic action sequences are treated as latent paths and we introduce a posterior distribution that prioritizes trajectories providing more diagnostic information. The planning LLM agent is then trained to follow this distribution, encouraging coherent diagnostic trajectories that progressively reduce uncertainty. Experiments on the MIMIC-CDM benchmark demonstrate that our proposed LDTL framework outperforms existing baselines in diagnostic accuracy under a sequential clinical diagnosis setting, while requiring fewer diagnostic tests. Furthermore, ablation studies highlight the critical role of trajectory-level posterior alignment in achieving these improvements. --- ### Probabilistic Tree Inference Enabled by FDSOI Ferroelectric FETs Authors: Pengyu Ren, Xingtian Wang, Boyang Cheng, Jiahui Duan, Giuk Kim, Xuezhong Niu, Halid Mulaosmanovic, Stefan Duenkel, Sven Beyer, X. Sharon Hu, Ningyuan Cao, Kai Ni Published: 2026-04-06 Categories: cs.ET, cs.LG arXiv: https://arxiv.org/abs/2604.05115v1 Artificial intelligence applications in autonomous driving, medical diagnostics, and financial systems increasingly demand machine learning models that can provide robust uncertainty quantification, interpretability, and noise resilience. Bayesian decision trees (BDTs) are attractive for these tasks because they combine probabilistic reasoning, interpretable decision-making, and robustness to noise. However, existing hardware implementations of BDTs based on CPUs and GPUs are limited by memory bottlenecks and irregular processing patterns, while multi-platform solutions exploiting analog content-addressable memory (ACAM) and Gaussian random number generators (GRNGs) introduce integration complexity and energy overheads. Here we report a monolithic FDSOI-FeFET hardware platform that natively supports both ACAM and GRNG functionalities. The ferroelectric polarization of FeFETs enables compact, energy-efficient multi-bit storage for ACAM, and band-to-band tunneling in the gate-to-drain overlap region and subsequent hole storage in the floating body provides a high-quality entropy source for GRNG. System-level evaluations demonstrate that the proposed architecture provides robust uncertainty estimation, interpretability, and noise tolerance with high energy efficiency. Under both dataset noise and device variations, it achieves over 40% higher classification accuracy on MNIST compared to conventional decision trees. Moreover, it delivers more than two orders of magnitude speedup over CPU and GPU baselines and over four orders of magnitude improvement in energy efficiency, making it a scalable solution for deploying BDTs in resource-constrained and safety-critical environments. --- ### $π^2$: Structure-Originated Reasoning Data Improves Long-Context Reasoning Ability of Large Language Models Authors: Quyet V. Do, Thinh Pham, Nguyen Nguyen, Sha Li, Pratibha Zunjare, Tu Vu Published: 2026-04-06 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2604.05114v1 We study a pipeline that curates reasoning data from initial structured data for improving long-context reasoning in large language models (LLMs). Our approach, $π^2$, constructs high-quality reasoning data through rigorous QA curation: 1) extracting and expanding tables from Wikipedia, 2) from the collected tables and relevant context, generating realistic and multi-hop analytical reasoning questions whose answers are automatically determined and verified through dual-path code execution, and 3) back-translating step-by-step structured reasoning traces as solutions of QA pairs given realistic web-search context. Supervised fine-tuning with \textsc{\small{gpt-oss-20b}} and \textsc{\small{Qwen3-4B-Instruct-2507}} on $π^2$ yields consistent improvements across four long-context reasoning benchmarks and our alike $π^2$-Bench, with average absolute accuracy gains of +4.3% and +2.7% respectively. Notably, our dataset facilitates self-distillation, where \textsc{\small{gpt-oss-20b}} even improves its average performance by +4.4% with its own reasoning traces, demonstrating $π^2$'s usefulness. Our code, data, and models are open-source at https://github.com/vt-pi-squared/pi-squared. --- ### CRAB: Codebook Rebalancing for Bias Mitigation in Generative Recommendation Authors: Zezhong Fan, Ziheng Chen, Luyi Ma, Jin Huang, Lalitesh Morishetti, Kaushiki Nag, Sushant Kumar, Kannan Achan Published: 2026-04-06 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2604.05113v1 Generative recommendation (GeneRec) has introduced a new paradigm that represents items as discrete semantic tokens and predicts items in a generative manner. Despite its strong performance across multiple recommendation tasks, existing GeneRec approaches still suffer from severe popularity bias and may even exacerbate it. In this work, we conduct a comprehensive empirical analysis to uncover the root causes of this phenomenon, yielding two core insights: 1) imbalanced tokenization inherits and can further amplify popularity bias from historical item interactions; 2) current training procedures disproportionately favor popular tokens while neglecting semantic relationships among tokens, thereby intensifying popularity bias. Building on these insights, we propose CRAB, a post-hoc debiasing strategy for GeneRec that alleviates popularity bias by mitigating frequency imbalance among semantic tokens. Specifically, given a well-trained model, we first rebalance the codebook by splitting over-popular tokens while preserving their hierarchical semantic structure. Based on the adjusted codebook, we further introduce a tree-structured regularizer to enhance semantic consistency, encouraging more informative representations for unpopular tokens during training. Experiments on real-world datasets demonstrate that CRAB significantly improves recommendation performance by effectively alleviating popularity bias. --- ### Vintix II: Decision Pre-Trained Transformer is a Scalable In-Context Reinforcement Learner Authors: Andrei Polubarov, Lyubaykin Nikita, Alexander Derevyagin, Artyom Grishin, Igor Saprygin, Aleksandr Serkov, Mark Averchenko, Daniil Tikhonov, Maksim Zhdanov, Alexander Nikulin, Ilya Zisman, Albina Klepach et al. Published: 2026-04-06 Categories: cs.LG, cs.AI, cs.RO arXiv: https://arxiv.org/abs/2604.05112v1 Recent progress in in-context reinforcement learning (ICRL) has demonstrated its potential for training generalist agents that can acquire new tasks directly at inference. Algorithm Distillation (AD) pioneered this paradigm and was subsequently scaled to multi-domain settings, although its ability to generalize to unseen tasks remained limited. The Decision Pre-Trained Transformer (DPT) was introduced as an alternative, showing stronger in-context reinforcement learning abilities in simplified domains, but its scalability had not been established. In this work, we extend DPT to diverse multi-domain environments, applying Flow Matching as a natural training choice that preserves its interpretation as Bayesian posterior sampling. As a result, we obtain an agent trained across hundreds of diverse tasks that achieves clear gains in generalization to the held-out test set. This agent improves upon prior AD scaling and demonstrates stronger performance in both online and offline inference, reinforcing ICRL as a viable alternative to expert distillation for training generalist agents. --- ### Simultaneous Dual-View Mammogram Synthesis Using Denoising Diffusion Probabilistic Models Authors: Jorge Alberto Garza-Abdala, Gerardo A. Fumagal-González, Eduardo de Avila-Armenta, Sadam Hussain, Jasiel H. Toscano-Martínezb, Diana S. M. Rosales Gurmendi, Alma A. Pedro-Pérez, Jose G. Tamez-Pena Published: 2026-04-06 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2604.05110v1 Breast cancer screening relies heavily on mammography, where the craniocaudal (CC) and mediolateral oblique (MLO) views provide complementary information for diagnosis. However, many datasets lack complete paired views, limiting the development of algorithms that depend on cross-view consistency. To address this gap, we propose a three-channel denoising diffusion probabilistic model capable of simultaneously generating CC and MLO views of a single breast. In this configuration, the two mammographic views are stored in separate channels, while a third channel encodes their absolute difference to guide the model toward learning coherent anatomical relationships between projections. A pretrained DDPM from Hugging Face was fine-tuned on a private screening dataset and used to synthesize dual-view pairs. Evaluation included geometric consistency via automated breast mask segmentation and distributional comparison with real images, along with qualitative inspection of cross-view alignment. The results show that the difference-based encoding helps preserve the global breast structure across views, producing synthetic CC-MLO pairs that resemble real acquisitions. This work demonstrates the feasibility of simultaneous dual-view mammogram synthesis using a difference-guided DDPM, highlighting its potential for dataset augmentation and future cross-view-aware AI applications in breast imaging. --- ### Edit, But Verify: An Empirical Audit of Instructed Code-Editing Benchmarks Authors: Amir M. Ebrahimi, Gopi Krishnan Rajbahadur Published: 2026-04-06 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2604.05100v1 Instructed code editing, where an LLM modifies existing code based on a natural language instruction, accounts for roughly 19% of real-world coding assistant interactions. Yet very few benchmarks directly evaluate this capability. From a survey of over 150 code-related benchmarks, we find that only two, CanItEdit and EDIT-Bench, target instructed code editing with human-authored instructions and test-based evaluation. We audit both by comparing their programming languages, edit intents, and application domains against distributions observed in the wild (Copilot Arena, AIDev, GitHub Octoverse), and by measuring test counts, statement coverage, and test scope across all 213 problems. Both benchmarks concentrate over 90\% of evaluation on Python while TypeScript, GitHub's most-used language, is absent. Backend and frontend development, which together constitute 46% of real-world editing activity, are largely missing, and documentation, testing, and maintenance edits (31.4% of human PRs) have zero representation. Both benchmarks have modest test counts (CanItEdit median 13, EDIT-Bench median 4), though CanItEdit compensates with near-complete whole-file coverage and fail-before/pass-after validation. 59\% of EDIT-Bench's low-coverage suites would not detect modifications outside the edit region. EDIT-Bench has 15 problems that are not solved by any of 40 LLMs and 11 of these problems trace failures to poor benchmark artifacts rather than model limitations. Further, 29% of EDIT-Bench problems and 6% of CanItEdit problems share a codebase with at least one other problem within the benchmark. In summary, these benchmarks measure a narrower construct than deployment decisions require. We therefore propose six empirically grounded desiderata and release all audit artifacts so the community can build instructed code-editing benchmarks whose scores reliably reflect real-world editing capability. --- ### RAG or Learning? Understanding the Limits of LLM Adaptation under Continuous Knowledge Drift in the Real World Authors: Hanbing Liu, Lang Cao, Yang Li Published: 2026-04-06 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.05096v2 Large language models (LLMs) acquire most of their knowledge during pretraining, which ties them to a fixed snapshot of the world and makes adaptation to continuously evolving knowledge challenging. As facts, entities, and events change over time, models may experience continuous knowledge drift, resulting not only in outdated predictions but also in temporally inconsistent reasoning. Although existing approaches, such as continual finetuning, knowledge editing, and retrieval-augmented generation (RAG), aim to update or supplement model knowledge, they are rarely evaluated in settings that reflect chronological, evolving, and real-world knowledge evolution. In this work, we introduce a new benchmark of real-world dynamic events, constructed from time-stamped evidence that captures how knowledge evolves over time, which enables systematic evaluation of model adaptation under continuous knowledge drift. The benchmark reveals that most existing methods, including vanilla RAG and several learning-based approaches, struggle under this setting, exposing critical limitations such as catastrophic forgetting and temporal inconsistency. To mitigate these limitations, we propose a time-aware retrieval baseline, Chronos, which progressively organizes retrieved evidence into an Event Evolution Graph to enable more temporally consistent understanding in LLMs without additional training. Overall, this work provides a foundation for analyzing and advancing LLM adaptation to continuous knowledge drift in realistic settings. --- ### MegaTrain: Full Precision Training of 100B+ Parameter Large Language Models on a Single GPU Authors: Zhengqing Yuan, Hanchi Sun, Lichao Sun, Yanfang Ye Published: 2026-04-06 Categories: cs.CL, cs.DC, cs.OS arXiv: https://arxiv.org/abs/2604.05091v1 We present MegaTrain, a memory-centric system that efficiently trains 100B+ parameter large language models at full precision on a single GPU. Unlike traditional GPU-centric systems, MegaTrain stores parameters and optimizer states in host memory (CPU memory) and treats GPUs as transient compute engines. For each layer, we stream parameters in and compute gradients out, minimizing persistent device state. To battle the CPU-GPU bandwidth bottleneck, we adopt two key optimizations. 1) We introduce a pipelined double-buffered execution engine that overlaps parameter prefetching, computation, and gradient offloading across multiple CUDA streams, enabling continuous GPU execution. 2) We replace persistent autograd graphs with stateless layer templates, binding weights dynamically as they stream in, eliminating persistent graph metadata while providing flexibility in scheduling. On a single H200 GPU with 1.5TB host memory, MegaTrain reliably trains models up to 120B parameters. It also achieves 1.84$\times$ the training throughput of DeepSpeed ZeRO-3 with CPU offloading when training 14B models. MegaTrain also enables 7B model training with 512k token context on a single GH200. --- ### Multilingual Language Models Encode Script Over Linguistic Structure Authors: Aastha A K Verma, Anwoy Chatterjee, Mehak Gupta, Tanmoy Chakraborty Published: 2026-04-06 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2604.05090v2 Multilingual language models (LMs) organize representations for typologically and orthographically diverse languages into a shared parameter space, yet the nature of this internal organization remains elusive. In this work, we investigate which linguistic properties - abstract language identity or surface-form cues - shape multilingual representations. To do so, we analyze language-associated units across different model families and scales using the Language Activation Probability Entropy (LAPE) metric, and further decompose activations with Sparse Autoencoders. We find that these units are strongly conditioned on orthography: romanization induces near-disjoint representations that align with neither native-script inputs nor English, while word-order shuffling has limited effect on unit identity. Probing shows that typological structure becomes increasingly accessible in deeper layers, while causal interventions indicate that generation is most sensitive to units that are invariant to surface-form perturbations rather than to units identified by typological alignment alone. Overall, our results suggest that multilingual LMs organize representations around surface form, with linguistic abstraction emerging gradually without collapsing into a unified interlingua. --- ### Scalar Federated Learning for Linear Quadratic Regulator Authors: Mohammadreza Rostami, Shahriar Talebi, Solmaz S. Kia Published: 2026-04-06 Categories: eess.SY, cs.LG arXiv: https://arxiv.org/abs/2604.05088v1 We propose ScalarFedLQR, a communication-efficient federated algorithm for model-free learning of a common policy in linear quadratic regulator (LQR) control of heterogeneous agents. The method builds on a decomposed projected gradient mechanism, in which each agent communicates only a scalar projection of a local zeroth-order gradient estimate. The server aggregates these scalar messages to reconstruct a global descent direction, reducing per-agent uplink communication from O(d) to O(1), independent of the policy dimension. Crucially, the projection-induced approximation error diminishes as the number of participating agents increases, yielding a favorable scaling law: larger fleets enable more accurate gradient recovery, admit larger stepsizes, and achieve faster linear convergence despite high dimensionality. Under standard regularity conditions, all iterates remain stabilizing and the average LQR cost decreases linearly fast. Numerical results demonstrate performance comparable to full-gradient federated LQR with substantially reduced communication. --- ### Document Optimization for Black-Box Retrieval via Reinforcement Learning Authors: Omri Uzan, Ron Polonsky, Douwe Kiela, Christopher Potts Published: 2026-04-06 Categories: cs.CL, cs.IR arXiv: https://arxiv.org/abs/2604.05087v3 Document expansion is a classical technique for improving retrieval quality, and is attractive since it shifts computation offline, avoiding additional query-time processing. However, when applied to modern retrievers, it has been shown to degrade performance, often introducing noise that obfuscates the discriminative signal. We recast document expansion as a document optimization problem: a language model or a vision language model is fine-tuned to transform documents into representations that better align with the expected query distribution under a target retriever, using GRPO with the retriever's ranking improvements as rewards. This approach requires only black-box access to retrieval ranks, and is applicable across single-vector, multi-vector and lexical retrievers. We evaluate our approach on code retrieval and visual document retrieval (VDR) tasks. We find that learned document transformations yield retrieval gains and in many settings enable smaller, more efficient retrievers to outperform larger ones. For example, applying document optimization to OpenAI text-embedding-3-small model improves nDCG5 on code (58.7 to 66.8) and VDR (53.3 to 57.6), even slightly surpassing the 6.5X more expensive OpenAI text-embedding-3-large model (66.3 on code; 57.0 on VDR). When retriever weights are accessible, document optimization is often competitive with fine-tuning, and in some settings their combination performs best, improving Jina-ColBERT-V2 from 55.8 to 63.3 on VDR and from 48.6 to 61.8 on code retrieval. --- ### Beyond LLM-as-a-Judge: Deterministic Metrics for Multilingual Generative Text Evaluation Authors: Firoj Alam, Gagan Bhatia, Sahinur Rahman Laskar, Shammur Absar Chowdhury Published: 2026-04-06 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2604.05083v1 While Large Language Models (LLMs) are increasingly adopted as automated judges for evaluating generated text, their outputs are often costly, and highly sensitive to prompt design, language, and aggregation strategies, severely, which limits reproducibility. To address these challenges, we propose \textbf{\textit{OmniScore}}, a family of complementary, deterministic learned metrics developed using small size ($<$1B) parameter models. OmniScore approximates LLM-judge behavior while preserving the low latency and consistency of traditional model-based scoring. We trained the models large-scale synthetic supervision ($\sim$564k instances, in \textbf{107 languages}) and evaluated using 8,617 manually annotated instances. The OmniScore family supports reliable, multi-dimensional scores across a variety of settings, including reference-based, source-grounded, and hybrid evaluations. We evaluate these models across question answering (QA), translation, and summarization in \textbf{6 languages}. Our results demonstrate that lightweight, deterministic learned metrics provide a highly practical and scalable alternative to frontier LLMs. Our models and datasets can be found at https://huggingface.co/collections/QCRI/omniscore --- ### MedGemma 1.5 Technical Report Authors: Andrew Sellergren, Chufan Gao, Fereshteh Mahvar, Timo Kohlberger, Fayaz Jamil, Madeleine Traverse, Alberto Tono, Bashir Sadjad, Lin Yang, Charles Lau, Liron Yatziv, Tiffany Chen et al. Published: 2026-04-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.05081v2 We introduce MedGemma 1.5 4B, the latest model in the MedGemma collection. MedGemma 1.5 expands on MedGemma 1 by integrating additional capabilities: high-dimensional medical imaging (CT/MRI volumes and histopathology whole slide images), anatomical localization via bounding boxes, multi-timepoint chest X-ray analysis, and improved medical document understanding (lab reports, electronic health records). We detail the innovations required to enable these modalities within a single architecture, including new training data, long-context 3D volume slicing, and whole-slide pathology sampling. Compared to MedGemma 1 4B, MedGemma 1.5 4B demonstrates significant gains in these new areas, improving 3D MRI condition classification accuracy by 11% and 3D CT condition classification by 3% (absolute improvements). In whole slide pathology imaging, MedGemma 1.5 4B achieves a 47% macro F1 gain. Additionally, it improves anatomical localization with a 35% increase in Intersection over Union on chest X-rays and achieves a 4% macro accuracy for longitudinal (multi-timepoint) chest x-ray analysis. Beyond its improved multimodal performance over MedGemma 1, MedGemma 1.5 improves on text-based clinical knowledge and reasoning, improving by 5% on MedQA accuracy and 22% on EHRQA accuracy. It also achieves an average of 18% macro F1 on 4 different lab report information extraction datasets (EHR Datasets 2, 3, 4, and Mendeley Clinical Laboratory Test Reports). Taken together, MedGemma 1.5 serves as a robust, open resource for the community, designed as an improved foundation on which developers can create the next generation of medical AI systems. Resources and tutorials for building upon MedGemma 1.5 can be found at https://goo.gle/medgemma. --- ### Nidus: Externalized Reasoning for AI-Assisted Engineering Authors: Danil Gorinevski Published: 2026-04-06 Categories: cs.SE, cs.AI, cs.LO, cs.MA arXiv: https://arxiv.org/abs/2604.05080v1 We present Nidus, a governance runtime that mechanizes the V-model for AI-assisted software delivery. In the self-hosting deployment, three LLM families (Claude, Gemini, Codex) delivered a 100,000-line system under proof obligations verified against the current obligation set on every commit. The system governed its own construction. Engineering invariants - traced requirements, justified architecture, evidenced deliveries - cannot be reliably maintained as learned behavior; assurance requires enforcement by a mechanism external to the proposer. Nidus externalizes the engineering methodology into a decidable artifact verified on every mutation before persistence. Organizational standards compile into guidebooks - constraint libraries imported by governed projects and enforced by decidable evaluation. Four contributions: (1) recursive self-governance - the constraint surface constrains mutations to itself; (2) stigmergic coordination - friction from the surface routes agents without central control; (3) proximal spec reinforcement - the living artifact externalizes the engineering context that RL and long-chain reasoning try to internalize; the specification is the reward function, UNSAT verdicts shape behavior at inference time, no weight updates; (4) governance theater prevention - compliance evidence cannot be fabricated within the modeled mutation path. The constraint surface compounds: each obligation permanently eliminates a class of unengineered output. The artifact's development history is a formal development - every state satisfies all active obligations, and the obligation set grows monotonically. --- ### SVAgent: Storyline-Guided Long Video Understanding via Cross-Modal Multi-Agent Collaboration Authors: Zhongyu Yang, Zuhao Yang, Shuo Zhan, Tan Yue, Wei Pang, Yingfang Yuan Published: 2026-04-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.05079v1 Video question answering (VideoQA) is a challenging task that requires integrating spatial, temporal, and semantic information to capture the complex dynamics of video sequences. Although recent advances have introduced various approaches for video understanding, most existing methods still rely on locating relevant frames to answer questions rather than reasoning through the evolving storyline as humans do. Humans naturally interpret videos through coherent storylines, an ability that is crucial for making robust and contextually grounded predictions. To address this gap, we propose SVAgent, a storyline-guided cross-modal multi-agent framework for VideoQA. The storyline agent progressively constructs a narrative representation based on frames suggested by a refinement suggestion agent that analyzes historical failures. In addition, cross-modal decision agents independently predict answers from visual and textual modalities under the guidance of the evolving storyline. Their outputs are then evaluated by a meta-agent to align cross-modal predictions and enhance reasoning robustness and answer consistency. Experimental results demonstrate that SVAgent achieves superior performance and interpretability by emulating human-like storyline reasoning in video understanding. --- ### Feature-Aware Anisotropic Local Differential Privacy for Utility-Preserving Graph Representation Learning in Metal Additive Manufacturing Authors: MD Shafikul Islam, Mahathir Mohammad Bappy, Saifur Rahman Tushar, Md Arifuzzaman Published: 2026-04-06 Categories: cs.LG, cs.AI, cs.CR arXiv: https://arxiv.org/abs/2604.05077v1 Metal additive manufacturing (AM) enables the fabrication of safety-critical components, but reliable quality assurance depends on high-fidelity sensor streams containing proprietary process information, limiting collaborative data sharing. Existing defect-detection models typically treat melt-pool observations as independent samples, ignoring layer-wise physical couplings. Moreover, conventional privacy-preserving techniques, particularly Local Differential Privacy (LDP), lead to severe utility degradation because they inject uniform noise across all feature dimensions. To address these interrelated challenges, we propose FI-LDP-HGAT. This computational framework combines two methodological components: a stratified Hierarchical Graph Attention Network (HGAT) that captures spatial and thermal dependencies across scan tracks and deposited layers, and a feature-importance-aware anisotropic Gaussian mechanism (FI-LDP) for non-interactive feature privatization. Unlike isotropic LDP, FI-LDP redistributes the privacy budget across embedding coordinates using an encoder-derived importance prior, assigning lower noise to task-critical thermal signatures and higher noise to redundant dimensions while maintaining formal LDP guarantees. Experiments on a Directed Energy Deposition (DED) porosity dataset demonstrate that FI-LDP-HGAT achieves 81.5% utility recovery at a moderate privacy budget (epsilon = 4) and maintains defect recall of 0.762 under strict privacy (epsilon = 2), while outperforming classical ML, standard GNNs, and alternative privacy mechanisms, including DP-SGD across all evaluated metrics. Mechanistic analysis confirms a strong negative correlation (Spearman = -0.81) between feature importance and noise magnitude, providing interpretable evidence that the privacy-utility gains are driven by principled anisotropic allocation. --- ### MMORF: A Multi-agent Framework for Designing Multi-objective Retrosynthesis Planning Systems Authors: Frazier N. Baker, Trieu Nguyen, Reza Averly, Botao Yu, Daniel Adu-Ampratwum, Huan Sun, Xia Ning Published: 2026-04-06 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2604.05075v2 Multi-objective retrosynthesis planning is a critical chemistry task requiring dynamic balancing of quality, safety, and cost objectives. Language model-based multi-agent systems (MAS) offer a promising approach for this task: leveraging interactions of specialized agents to incorporate multiple objectives into retrosynthesis planning. We present MMORF, a framework for constructing MAS for multi-objective retrosynthesis planning. MMORF features modular agentic components, which can be flexibly combined and configured into different systems, enabling principled evaluation and comparison of different system designs. Using MMORF, we construct two representative MAS: MASIL and RFAS. On a newly curated benchmark consisting of 218 multi-objective retrosynthesis planning tasks, MASIL achieves strong safety and cost metrics on soft-constraint tasks, frequently Pareto-dominating baseline routes, while RFAS achieves a 48.6% success rate on hard-constraint tasks, outperforming state-of-the-art baselines. Together, these results show the effectiveness of MMORF as a foundational framework for exploring MAS for multi-objective retrosynthesis planning. Code and data are available at https://github.com/ninglab/MMORF. --- ### MMORF: A Multi-agent Framework for Designing Multi-objective Retrosynthesis Planning Systems Authors: Frazier N. Baker, Trieu Nguyen, Reza Averly, Botao Yu, Daniel Adu-Ampratwum, Huan Sun, Xia Ning Published: 2026-04-06 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2604.05075v1 Multi-objective retrosynthesis planning is a critical chemistry task requiring dynamic balancing of quality, safety, and cost objectives. Language model-based multi-agent systems (MAS) offer a promising approach for this task: leveraging interactions of specialized agents to incorporate multiple objectives into retrosynthesis planning. We present MMORF, a framework for constructing MAS for multi-objective retrosynthesis planning. MMORF features modular agentic components, which can be flexibly combined and configured into different systems, enabling principled evaluation and comparison of different system designs. Using MMORF, we construct two representative MAS: MASIL and RFAS. On a newly curated benchmark consisting of 218 multi-objective retrosynthesis planning tasks, MASIL achieves strong safety and cost metrics on soft-constraint tasks, frequently Pareto-dominating baseline routes, while RFAS achieves a 48.6% success rate on hard-constraint tasks, outperforming state-of-the-art baselines. Together, these results show the effectiveness of MMORF as a foundational framework for exploring MAS for multi-objective retrosynthesis planning. Code and data are available at https://anonymous.4open.science/r/MMORF/. --- ### Memory Dial: A Training Framework for Controllable Memorization in Language Models Authors: Xiangbo Zhang, Ali Emami Published: 2026-04-06 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.05074v1 Memorization in language models is widely studied but remains difficult to isolate and control. Understanding when and what models memorize is essential for explaining their predictions, yet existing approaches are post-hoc: they can detect memorization in trained models, but cannot disentangle its effects from architecture, data, or optimization. We introduce Memory Dial, a training framework that makes memorization pressure an explicit, controllable variable. Memory Dial interpolates between standard cross-entropy and a temperature-sharpened objective via a single parameter $α$, producing a family of models identical in architecture and training setup (within each sweep), differing only in memorization pressure. Experiments across six architectures and five benchmarks demonstrate that: (1) $α$ reliably controls memorization pressure, with seen-example accuracy increasing monotonically while unseen accuracy remains stable; (2) larger models are more responsive to memorization pressure; and (3) frequent sequences are easier to memorize than rare ones. Additional analyses show that the effect is robust across a range of sharpening temperatures, differs qualitatively from single-temperature cross-entropy, transfers to multilingual settings, and is detectable even on naturally occurring single-occurrence sequences. Memory Dial provides a controlled experimental framework for studying how memorization behavior emerges and interacts with generalization in language models. --- ### Hierarchical SVG Tokenization: Learning Compact Visual Programs for Scalable Vector Graphics Modeling Authors: Ximing Xing, Ziteng Xue, Zhenxi Li, Weicong Liang, Linqing Wang, Zhantao Yang, Tiankai Hang, Zijin Yin, Qinglin Lu, Chunyu Wang, Qian Yu Published: 2026-04-06 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.05072v2 Recent large language models have shifted SVG generation from differentiable rendering optimization to autoregressive program synthesis. However, existing approaches still rely on generic byte-level tokenization inherited from natural language processing, which poorly reflects the geometric structure of vector graphics. Numerical coordinates are fragmented into discrete symbols, destroying spatial relationships and introducing severe token redundancy, often leading to coordinate hallucination and inefficient long-sequence generation. To address these challenges, we propose HiVG, a hierarchical SVG tokenization framework tailored for autoregressive vector graphics generation. HiVG decomposes raw SVG strings into structured \textit{atomic tokens} and further compresses executable command--parameter groups into geometry-constrained \textit{segment tokens}, substantially improving sequence efficiency while preserving syntactic validity. To further mitigate spatial mismatch, we introduce a Hierarchical Mean--Noise (HMN) initialization strategy that injects numerical ordering signals and semantic priors into new token embeddings. Combined with a curriculum training paradigm that progressively increases program complexity, HiVG enables more stable learning of executable SVG programs. Extensive experiments on both text-to-SVG and image-to-SVG tasks demonstrate improved generation fidelity, spatial consistency, and sequence efficiency compared with conventional tokenization schemes. Our code is publicly available at https://github.com/ximinng/HiVG ---