The paper introduces MIRAGE, a benchmark of 750 multi‑step decision tasks designed to test autonomous web agents’ investigative abilities across Wikipedia Forensics, Shopping Admin adjudication, and Reddit Moderation. Each task contains a misleading visible context and a hidden context that holds decisive evidence, allowing performance to be broken down into Investigation, Reasoning, and Decision Accuracy, with an added Investigative Hallucination Rate. Evaluation of eight LLM agents reveals three consistent patterns: agents often reach relevant pages but fail to extract decisive evidence, procedural hints improve investigation but not decision accuracy on Wikipedia tasks, and 12.6% of trajectories include fabricated facts.
By Syed Nazmus Sakib, Nafiul Haque, Tapodhir Karmakar Taton, Shahrear Bin Amin, Shifat E. Arman
LogicSkills is a benchmark designed to isolate three core logical abilities in large language models: formal symbolization, countermodel construction, and validity assessment. The dataset draws items from the two-variable fragment of first‑order logic without identity, presented in both English and a Carrollian nonce‑word language, and all instances are solver‑verified with Z3. Results show that conventional instruction‑tuned LLMs excel at validity assessment but struggle with symbolization and countermodel construction, whereas recent reasoning‑tuned models perform well across all tasks, indicating a more systematic logical skill profile.
By Brian Rabern, Philipp Mondorf, Barbara Plank
The paper investigates how to apply Reinforcement Learning with Verifiable Rewards (RLVR) to large language models across multiple domains. It compares two training paradigms—mixed multi-task RLVR and separate RLVR followed by model merging—using tasks such as math, coding, science, instruction following, and agent. Experiments show that RLVR across domains causes minimal interference and that reasoning-intensive domains can synergize, with insights drawn from information constraints, prediction behavior, and self-verification.
By Haoqing Wang, Xiang Long, Ziheng Li, Yilong Xu, Tingguang Li, Yehui Tang
AutoFyn is a new agent framework based on the Expert Iteration algorithm that updates a frozen model’s persistent state using verified reward signals instead of changing model weights. Each iteration starts with a fresh model session, and only durable information—such as memory files, reports, and repository state—is carried over through explicit interfaces. The system orchestrates exploration, planning, and specialized agents, while a task‑grounded verifier supplies objective rewards that are distilled back into the persistent state to guide the next round. AutoFyn has been applied to olympiad mathematics, data science, and cybersecurity, outperforming baseline models on the 2026 International Mathematical Olympiad, topping the Spider 2.0 dbt benchmark, and generating sixteen maintainer‑confirmed vulnerability advisories across several popular software projects.
By Adib Hasan, Daniel Schaffield, Akashnil Dutta, Tarik Adnan Moon
PGP-Clinical-TimeKAN is a trajectory-first framework for joint probabilistic forecasting of multivariate physiological data, combining missingness-aware temporal encoders, a soft organ-system prior, patient-specific relations, nonlinear Kolmogorov‑Arnold messages, and a low‑rank multivariate Student‑t head. Evaluated on a MIMIC‑IV cohort of 6,882 patients, it achieves the second‑lowest normalized MAE and the lowest RMSE among 13 models, while providing calibrated probabilistic forecasts with empirical coverage at 50%, 80%, and 95% intervals. Ablation studies show that relational structure is critical for performance, and increasing covariance rank improves likelihood but not point accuracy.
whyItMatters":"The model demonstrates that joint trajectory forecasting can yield highly accurate, calibrated predictions of physiological trajectories, offering a potentially inspectable intermediate task for clinical deterioration prediction."
By Weizhi Nie, Rihao Chang, Weijie Wang, Yuting Su
SciLitBench is a multi-stage benchmark for evaluating large language models (LLMs) in systematic literature reviews, covering title and abstract screening, full-text screening, and schema-guided data extraction across 42,981 records and 888 included papers. The study shows that explicit inclusion/exclusion criteria boost title and abstract screening performance by 28.8% and researcher-authored rationales improve full-text screening by 15%. Data extraction performance varies widely, with high accuracy for publication year but low overlap for computational approaches, and even the best models recover only a fraction of annotated evidence and limitations.
By Miguel Zabaleta, Baihan Lin
The paper evaluates the scalability and adversarial generalization of Natural Language Inference (NLI) models trained with Group Relative Policy Optimization (GRPO) for Chain-of-Thought learning. By fine‑tuning 7B, 14B, and 32B language models with LoRA and QLoRA, the authors show strong performance on standard and adversarial NLI benchmarks, with the 32B model outperforming supervised baselines on adversarial sets. Using AWQ quantization, the 32B model fits within 22 GB of CUDA memory, demonstrating a scalable, practical framework for robust NLI without sacrificing inference quality.
By Pablo Miralles-Gonz\'alez, Javier Huertas-Tato, Alejandro Mart\'in, David Camacho
Strategic Doctrine Language Models (sdLM) are a learning‑system framework that integrates multi‑document attention, temporal encoding, and a doctrine‑consistency layer to enable multi‑document strategic reasoning with doctrinal consistency constraints and calibrated uncertainty. The authors evaluate sdLM on expert‑panel scoring of 47 strategic scenarios, doctrine consistency across 336 doctrine publications (12,847 statements), and geopolitical forecasting on 127 historical counterfactuals spanning 1945‑2020 over 12‑60 month horizons, showing higher strategic quality and better calibration than strong general‑purpose LLM baselines and competitiveness with human experts on long‑horizon judgments. Ablation studies, scaling trends, and deployment‑oriented performance/latency characteristics are reported to identify which components drive improvements and how they translate to operational settings.
By Olaf Yunus Laitinen Imanov, Taner Yilmaz, Derya Umut Kulali
AudioFuse is a hybrid architecture that jointly learns from spectrograms and raw waveforms to classify phonocardiograms. It combines a wide-and-shallow Vision Transformer for spectral features with a shallow 1D CNN for temporal waveforms, reducing overfitting while capturing complementary information. On the PhysioNet 2016 dataset, AudioFuse achieves a state‑of‑the‑art ROC‑AUC of 0.8608 and shows superior robustness to domain shift on the PASCAL dataset, outperforming both spectrogram‑only and waveform‑only baselines.
By Md. Saiful Bari Siddiqui, Utsab Saha
The paper introduces LLM-EBG, an evolutionary framework that uses a large language model as a generative operator to automatically create optimization benchmarks. By generating unconstrained single-objective continuous minimization problems expressed as mathematical formulas, the framework can produce benchmarks that consistently favor a target algorithm over a comparison algorithm in over 80% of trials. Landscape analysis shows that these generated problems exhibit distinct geometric traits, such as sensitivity to variable scaling, reflecting the search behaviors of different optimization methods.
By Yuhiro Ono, Tomohiro Harada, Yukiya Miura
The paper introduces DSPA, a dynamic sparse autoencoder (SAE) steering technique that aligns language model outputs with user preferences during inference, avoiding costly weight updates. DSPA constructs a conditional-difference map from preference triples to adjust token-active latents, improving MT‑Bench scores and matching AlpacaEval performance on models like Gemma‑2 and Qwen3 while preserving accuracy. It demonstrates robustness with limited preference data, outperforms the two‑stage RAHF‑SCIT pipeline in FLOPs, and reveals that preference directions are largely driven by discourse and stylistic cues.
By James Wedgwood, Aashiq Muhamed, Mona T. Diab, Virginia Smith
UniQueR is a unified query‑based feedforward framework that reconstructs 3D scenes from unposed images by treating reconstruction as a sparse 3D query inference problem. It learns a compact set of 3D anchor points that serve as explicit geometric queries, allowing the network to infer scene structure—including occluded geometry—in a single forward pass. By encoding spatial and appearance priors directly in global 3D space and using a decoupled cross‑attention design, UniQueR achieves strong geometric expressiveness while reducing memory and computational cost, outperforming state‑of‑the‑art feedforward methods on Mip‑NeRF 360 and VR‑NeRF with far fewer primitives.
By Chensheng Peng, Quentin Herau, Jiezhi Yang, Yichen Xie, Yihan Hu, Wenzhao Zheng, Matthew Strong, Masayoshi Tomizuka, Wei Zhan
The paper introduces Particle MCTS (PMCTS), a parallelized version of Monte Carlo Tree Search designed for GPU acceleration and batch-parallel neural network evaluations. PMCTS preserves policy improvement guarantees of modern MCTS algorithms while scaling efficiently with parallel compute. Empirical results show that PMCTS consistently outperforms or matches heuristic-based baselines across various MCTS and reinforcement learning domains, including chess, Go, and both discrete and continuous control benchmarks.
By Yaniv Oren, Viliam Vadocz, Joery A. de Vries, Wendelin B\"ohmer, Matthijs T. J. Spaan, Hendrik Baier
IdeaAMBIG is a benchmark that evaluates how well research‑idea specifications are ready for implementation. It contains 660 evidence‑grounded instances, including 163 real‑world gaps from reproducibility reports and GitHub issues and 497 synthetic gaps added to codification‑ready references. The benchmark tests three capabilities—codification‑readiness assessment, defect localization, and clarification action generation—across 13 large language models, revealing that defect localization is the main bottleneck while clarification given the defect performs much better.
By Yiling Ma, Yilun Zhao, Sihong Wu, Manasi Patwardhan, Arman Cohan
The paper introduces Many-Tier Instruction Hierarchy (ManyIH), a new framework for resolving conflicts among instructions with arbitrarily many privilege levels in large language model agents. It presents ManyIH-Bench, a benchmark featuring 853 agentic tasks that require navigating up to 12 levels of conflicting instructions across 46 real-world agents. Experiments show current models achieve only about 40% accuracy when instruction conflict scales, highlighting a gap in fine-grained, scalable conflict resolution.
By Jingyu Zhang, Tianjian Li, William Jurayj, Hongyuan Zhan, Benjamin Van Durme, Daniel Khashabi
D3ARC is an asynchronous distributed hierarchical framework designed for time‑critical wildfire detection using multiple robotic agents. It enables cooperative perception, shared situational awareness, and coordinated actions while a remote controller asynchronously directs each robot’s motion. The system incorporates safe navigation, coverage efficiency, and a forward‑looking capability to evaluate candidate strategies before execution, achieving up to 94% mission success and 89.4% detection confidence in realistic simulations.
By Nikolaos Koursioumpas, Lina Magoula, Nancy Alonistioti, Ramin Khalili
RelightFormer is a feed‑forward generative Transformer that performs single‑ and multi‑view image relighting without explicit intrinsic property estimation. It incorporates a latent illumination module that injects target environment maps into spatial features via cross‑attention, and uses permutation‑invariant positional encodings to process unordered multi‑view inputs symmetrically. Trained on the large Laval Objaverse Dataset, the model achieves state‑of‑the‑art visual and photorealistic relighting quality, and demonstrates strong zero‑shot generalization across various relighting tasks.
By Hejun Wang, Jinxi Li, Junwei Jiang, Shiwei Mao, Hu Cheng, Shouwang Huang, Bo Yang
The paper introduces Backward Kernel Herding, an algorithm that iteratively removes data points to create representative subsets for kernel learning, achieving performance comparable to state‑of‑the‑art methods while speeding up subsampling when the reduced size is less than half the original dataset. It also proposes Flexible Kernel Thinning, an extension that allows construction of subsets of any size, not just successive halvings, and demonstrates that this method often yields the best predictive performance. Experiments on Gaussian Processes and Kernel Support Vector Machines show that Backward Kernel Herding excels in training‑time efficiency, while Flexible Kernel Thinning offers superior predictive accuracy and competitive memory usage, emphasizing the need to choose a reduction strategy based on the desired trade‑off between performance, cost, and memory.
By Blanca Cano-Camarero, Yago R. Aguado-Carrillo-de-Albornoz, \'Angela Fern\'andez-Pascual, Jos\'e R. Dorronsoro
TabBench-Bio is a living, interactive benchmark that evaluates machine learning models on 43 high‑dimensional biomedical tables, covering multiple domains. Using a shared cross‑validation protocol, the benchmark compares classical estimators, neural networks, and tabular foundation models across 28 feature‑by‑sample operating points, with RealTabPFN v2.5 achieving the highest performance at the reference cell of 10,000 features and 100 training samples. The benchmark provides reproducible results, fold‑level predictions, and invites community contributions to expand its dataset collection.
By Jules Kreuer, Sofiane Ouaari, Julia Hellmig, Julius Braitinger, Nico Pfeifer
The paper proposes typed federated artifacts—schema‑validated objects with per‑field privacy and dispute resolution—to enable tool‑routing knowledge sharing among frozen, heterogeneous LLM agents. By replacing flat text prompts with typed fields, the authors achieve near‑centralized routing performance on StableToolBench while reducing data size to 20 MB JSON per client. The study also highlights that a simple TF‑IDF classifier can outperform LLM routing on labeled benchmarks, indicating limitations in current evaluation methods.
By Abhijit Chakraborty, Ni Trieu, Vivek Gupta