The paper introduces SWE-Flux, a repository‑level benchmark designed to test large language models’ ability to reason about runtime behavior. It contains 480 execution‑grounded instances from 12 real Python repositories, with gold answers automatically harvested from instrumented test executions. Evaluation of five LLMs shows the task remains difficult, with the best model achieving only 37% accuracy, and the benchmark can generate challenging variants through input perturbation.
By Hamed Taherkhani, Mohammad Abdollahi, Melika Sepidband, Hridya Dhulipala, Tien N. Nguyen, Hadi Hemmati
The paper introduces a systematic benchmark for evaluating explainable methods that attribute temporal interactions in sequential recommendation systems. Using a dual-model masking metric, it assesses ten XAI techniques across CNN, Transformer, SASRec, and BERT4Rec backbones on KuaiRand and MovieLens datasets, revealing that gradient-based methods like GradientSHAP and Integrated Gradients are the most faithful and robust. It also finds that raw attention weights are unreliable, while gradient-weighted attention works better on short sequences but degrades on longer horizons, and that faithful methods capture genuine task structure rather than recency or popularity bias.
By Akash Pandey, Kanisha Shah, Addrish Roy, Dwipam Katariya, Hongyangyang Shi, Amanda Ding, Kalanand Mishra, Pranab Mohanty
The paper introduces a stability‑aware autoscaling framework for edge serverless workloads that combines an Attention‑Enhanced Double‑Stacked LSTM with Proximal Policy Optimization to address temporal blindness in deep reinforcement learning. By weighting recent historical states non‑uniformly, the method suppresses high‑frequency jitter while preserving demand trends, outperforming single‑layer LSTM, static HPA, and KEDA baselines in latency reduction and stability. Experiments on two Kubernetes clusters with real Azure Functions traces show a ~67% reduction in P90 latency and improved adherence to a 50 ms hard SLO.
By Faraz Shaikh, Gianluca Reali, Mauro Femminella
NADI 2026 is the seventh edition of the Nuanced Arabic Dialect Identification shared task series and the second focused on multidialectal Arabic speech processing. It includes five main tasks—Automatic Speech Recognition, Spoken Dialect Identification, Text-to-Speech, Spoken Language Translation, and Spoken Language Understanding—along with eight subtasks that test realistic scenarios such as low‑bandwidth, mixed dialects, code‑switching, out‑of‑domain, and zero‑shot settings. The event attracted 21 teams from at least 13 countries, with 48 test‑phase submissions and 14 system‑description papers, and the results highlight out‑of‑domain generalization as a major bottleneck while showcasing the strengths of Arabic‑specialized speech models, multimodal dialect identification, and ensemble methods.
By Peter Sullivan, Bashar Talafha, Ahmed Ashraf, Fethi Bougares, Haroun Elleuch, Chiyu Zhang, AbdelRahim Elmadany, Youssef Mohamed, Salima Mdhaffar, Yannick Est\`eve, Mohamed Elhoseiny, Hamzah Luqman, Nizar Habash, Muhammad Abdul-Mageed
The paper introduces DPara, a parallel speculative decoding framework that builds on DSpark-style parallel drafters. DPara eliminates the need for probabilistic guesses by precomputing draft representations for every acceptance boundary and using a lightweight autoregressive head to combine verification outcomes with these representations, enabling full parallelization of the backbone forward pass. Experiments on Qwen3-8B and Qwen3-14B across multiple benchmarks demonstrate average speedups of 3.21× and 3.52× over autoregressive decoding, outperforming existing serial and parallel speculative decoding methods.
By Fuliang Liu, Xue Li, Kun Qian, Zhibin Wang, Wanchun Dou, Wenyuan Yu, Chen Tian
SpeechLLMs used in professional settings often undergo domain customisation through prompts or fine‑tuning, which can inadvertently cause the model to transcribe phonetically similar words from its context or training data, leaking private information. The authors systematically investigate this overlooked privacy risk, creating benchmarks to measure leakage rates for both prompting and fine‑tuning, and find that both mechanisms cause measurable leakage that compounds when combined. They evaluate a prompt‑level mitigation strategy and analyse the accuracy‑leakage trade‑off, concluding that fine‑tuning without context prompts offers the best balance between performance and privacy.
By Maike Z\"ufle, Jan Niehues
The paper introduces a trilingual spoken hallucination detection benchmark covering English, Russian, and Kazakh news, with 12,013 samples that include synthetic alterations and severity levels, as well as 290 fact‑checked misinformation items. Detectors are evaluated in a reference‑free setting on text, ASR transcripts, and audio, revealing that most models underperform baseline classifiers, except Gemma‑3n on transcripts. Synthetic‑trained detectors achieve high macro‑F1 scores on real‑world misinformation, but Russian provenance analysis highlights model‑dependent signals that confound synthetic benchmarks.
By Meruyert Aristombayeva, Jason S. Lucas, Chaewan Chun, Dongwon Lee
ZoomDiff is a high‑fidelity diffusion model designed to improve dual‑camera smooth zooming by producing photo‑realistic transitions. It strengthens dual‑image conditional guidance during denoising, injects flow‑aligned multi‑scale features from the VAE encoder into the decoder to recover high‑frequency details, and uses flow‑guided temporal consistency supervision to ensure smoother transitions. Experiments on synthetic and real‑world datasets show that ZoomDiff outperforms state‑of‑the‑art methods both quantitatively and qualitatively.
By Jiayi Zhang, Renlong Wu, Yukang Ding, Sibin Deng, Wangmeng Zuo
LEGO is a dual‑module framework that combines a Legal Expert GraphRAG system with an expert Chain‑of‑Thought approach to enhance complex legal reasoning. The GraphRAG component uses an expert‑annotated civil code graph and a greedy normative‑coverage retrieval algorithm to extract relevant provision subgraphs, while the Chain‑of‑Thought module structures retrieved provisions and case facts into a Provision‑Fact‑Conclusion reasoning flow. Using a Qwen3‑8B backbone, LEGO achieves 40.53% exact‑match accuracy on LawExamQA_Civil, surpassing baseline RAG and CoT models and matching larger models on multi‑hop and open‑ended benchmarks, with ablation studies confirming the complementary benefits of both modules.
By Qingjing Chen, Junkai Zhang, Shaochun Wang, Jiahao Ding, Siyuan Zheng, Yukun Yan, Zhi Zheng, Antonino Rotolo, Yun Liu, Weixing Shen
MolDesignBench is a new benchmark for evaluating large language model (LLM)-based agents in scenario‑grounded molecular design. It contains 2,000 generation and optimization tasks that blend implicit narrative requirements with explicit property and functional‑group constraints, including infeasible cases, and require the use of 17 specialized chemistry tools. Experiments with leading LLMs show low success rates (best ~43%) and highlight failures in implicit‑constraint reasoning, infeasibility detection, and tool usage, underscoring the benchmark’s role in identifying key bottlenecks for future research.
By Yongjun Jeong, Hanbum Ko, Ye Rin Kim, Chanhui Lee, Rodrigo Hormazabal, Jaewan Lee, Sehui Han, Sungbin Lim, Sungwoong Kim
BiCFlow-MER introduces a conditional-flow framework for audio-text multimodal emotion recognition, treating the task as generative evidence transport within a structured emotion space. It disentangles emotion-oriented evidence from speaker style and lexical content, creating a conflict-aware affective condition that guides bidirectional rectified flow to an explicit emotion-space endpoint. The model verifies candidate emotions via adaptive prototype-cloud scoring and backward class-to-condition consistency, achieving superior performance on IEMOCAP, MELD, and the zero-shot CASE benchmark.
By Yanbing Wang, Shenyue Wang, Chunyang Yu
The paper proposes using category theory to structure reinforcement learning in high‑dimensional, partially observable environments. By partitioning the state space into equivalence classes (symmetry orbits) and treating each class as a groupoid with a canonical representative, the agent can share learning across similar states, reducing redundancy and improving sample efficiency. Experiments on partially observable benchmarks show that this orbit‑based partitioning consistently enhances performance in environments with latent symmetry.
By Ben Opperman, Eduardo Alonso, Esther Mondrag\'on
The paper "Reachable Global Optimization in AI Systems: How Global Is Global?" argues that claims of AI systems optimizing various components (prompts, policies, architectures, etc.) are underspecified unless they define the region actually reachable by the system. It introduces Reachability-Induced Optimization (RIO), a framework where a generator, verifier, controller, memory, tools, and budget determine a reachable candidate region, and proves several theoretical results about reachable-optimality and related concepts. Extensive benchmarks (66,150 trials across 270 landscapes) demonstrate that control can alter reachability, and that optimization quality, reachability quality, and control reliability must be reported separately.
By Wesley Shu
The paper presents DT4X+, an improved diagnosis algorithm that builds on DT4X by refining training set construction and symbolic‑regression loss to better separate target classes while maintaining coherence among non‑target classes. This results in relations that fully align with analytical redundancy relation properties, yielding more informative decision‑tree splits, enhanced robustness, and superior performance on dynamic‑system datasets. Experiments on benchmark systems confirm these advantages.
By Igor Bezmaternykh (INSA Toulouse), Louise Trav\'e-Massuy\`es (LAAS-DISCO, Comue de Toulouse, ANITI), Elodie Chanthery (LAAS)
The paper introduces the Systemic Risk Index, an open pipeline and dashboard that aggregates evidence from 19 public AI benchmarks into four systemic‑risk categories defined by the EU GPAI Code of Practice. It evaluates 18 models using harm‑preserving perturbations and simulated deployment contexts, offering users the ability to switch between average and worst‑case aggregation and to trace each risk rating back to its benchmark evidence. The study finds that worst‑case scores can be 14 to 37 points lower than average scores, and that LLM judges agree with human graders at a level comparable to human‑human agreement.
By Jacob T. Emmerson, Phuong-Anh Nguyen-Le, Ronan Romano, Wilber Sean V. Anterola, Yann Billeter, Zhijing Jin
The paper investigates why some tasks in the Terminal‑Bench/Frontier‑Bench datasets fail for all agents, distinguishing genuine difficulty from artifacts such as missing context, broken solutions, infrastructure failures, or verifier bypasses. Analyzing 125 all‑fail tasks, only 78 are certified as genuinely unsolved after applying a validity screen; the rest are attributable to broken oracles, infrastructure issues, bypassable verifiers, or insufficient evidence. The study concludes that a zero pass rate does not automatically indicate a hard task and recommends that frontier benchmarks provide evidence for all‑fail tasks before claiming capability gaps.
By Edward Lue Chee Lip, Boden Moraski, Tim Knappe, Lang Xiong, Sarvesh Gharat, Antonio Mari, Ivan Bercovich
The paper introduces four distinct evaluation levels—schema validity, topological validity, backend executability, and component‑set agreement—to assess large language model‑generated electrical circuits. Using a 150‑circuit trilingual benchmark and a typed circuit interchange pipeline, the authors show that each level captures errors missed by the others, with significant discrepancies observed between validator rejections and ngspice execution outcomes. A repair study further demonstrates that targeted model adjustments can markedly improve topological validity while having mixed effects on executability and component agreement.
By Ali Hedayati Pirouzan
The paper introduces Spec2COBOLRot, an agentic AI pipeline that generates realistic COBOL programs by combining specification-driven creation with iterative degradation guided by real production code patterns and complexity targets. The authors evaluate the pipeline on three programs from different business domains, showing that it reliably produces syntactically valid code and increases structural complexity, but it does not consistently preserve business behavior. They discuss the limitations of targeting structural metrics alone and propose future work that would generate legacy programs from scratch along a simulated development history.
By Jean-Baptiste Espinasse (DiverSe), Djamel Eddine Khelladi (DiverSe, CNRS, IRISA, KHORA), Mathieu Acher (INSA Rennes, IRISA, DiverSe)
COPE (Continual Optimization with Personalized embedding and self-Evaluation) is a new framework that continually personalizes large language models using learnable user embeddings and self‑evaluation to generate proxy rewards. It integrates preference capture, self‑evaluation calibration, and personalized response optimization into a single update step, allowing continuous model updates even when explicit user feedback is sparse. Experiments demonstrate that COPE outperforms both training‑free and training‑based baselines, remains complementary to Retrieval‑Augmented Prompting, and shows reliable self‑evaluation, meaningful preference patterns, stable general capabilities, and robustness to shifting preferences and alternative evaluators.
By Ruike Cao, Fugen Yao, Liang Dong, Jian Xu, Guanjun Jiang, Li Xiao
The paper benchmarks static embedding models—Word2Vec, FastText, and Doc2Vec—for detecting anomalous HTTP requests using a single‑class classification framework. It introduces HEDA, a modular pipeline that trains both embeddings and detectors solely on benign traffic in an unsupervised setting. Experiments on synthetic and real datasets show that FastText embeddings consistently yield high detection rates with controlled false positives.
By Amanda Riverol, Gustavo Betarte, Rodrigo Mart\'inez, \'Alvaro Pardo