VICO proposes a co‑evolutionary framework where a vision‑language model (actor) and an Environment‑as‑Rewriter (EnvRewriter) are trained together. The EnvRewriter edits visual structures—such as scene graphs, chart tables, or protected region masks—to generate new training samples whose difficulty is calibrated to the actor’s current performance via a pass‑rate reward. Across nine multimodal benchmarks, VICO‑8B outperforms its base model by up to +5.0 % on out‑of‑domain tasks and surpasses other co‑evolution baselines while requiring far fewer labeled samples.
By Meng Lu, Ligeng Zhu, Olivia Xiao, Yuchen Zhuang, Zihan Wang, Kuncheng Wu, Bangya Liu, Yu Wang, Charles Fleming, Wenqi Shi, Xuan Wang
The study evaluates local open‑weight large language model agents for data‑engineering tasks, using a benchmark of fifteen mobility‑workflow tasks that cover discovery, processing, enrichment, and reporting. Ten local model configurations were tested in one‑shot and closed‑loop conditions, producing 1,500 scored attempts on a consumer‑grade GPU. Results show that models larger than two billion parameters achieve up to 85.3% artifact‑level success, with closed‑loop workspaces boosting pass rates by 26.7–52.0 percentage points, and a quantized 9‑billion‑parameter model reaching 69.3% success with a 6.5 GB memory footprint.
By Jorge Garc\'ia-Carrasco, Javier Sanchis, Alejandro Reina-Reina, Alejandro Mat\'e, Juan Trujillo
The paper introduces STI-OPD, a stochastic teacher intervention framework that improves on-policy distillation for multi-turn agentic tasks. By measuring teacher‑student policy discrepancy with KL divergence, STI-OPD probabilistically replaces student actions with teacher actions to provide reliable supervision. An Importance‑Weighted Reverse KL objective then corrects token sampling mismatches, enabling the student to learn from mixed‑policy trajectories and outperform prior OPD baselines across benchmarks.
By Junnan Liu, Linhao Luo, Zhijun Chen, Qianren Mao, Thuy-Trang Vu, Gholamreza Haffari
AdaCast introduces a conditional parameter generation framework that tailors time‑series forecasting models to each input by producing low‑rank updates for a frozen pretrained foundation model. Unlike static adaptation methods that apply a single set of updates to all inputs, AdaCast generates input‑specific updates during both training and inference. Experiments on six public benchmarks show that AdaCast consistently outperforms static baselines in in‑domain forecasting and improves zero‑shot generalization to unseen datasets across domains.
By Darahaas Nallagatla, Darryl Cherian Jacob, Pan He
The paper introduces a benchmark for predicting a program’s final output and checkpoint state using only its source code and input, extending CRUXEval-style tasks with both short‑ and long‑trace inputs and checkpoints inside and after loops. It evaluates 400 cases from 371 Python and C++ programs across seven settings from four model families, achieving up to 93.0% accuracy on short‑trace final output and 77.0% on long‑trace final output, with reasoning‑enabled settings outperforming non‑reasoning ones by 33.1 to 55.2 percentage points. The study also shows that longer‑trace inputs can change correctness outcomes and that the benchmark reveals errors that short‑output scores alone would miss.
By Xiaohong Chen, David Bucur, Chenglong Ma, Yi Zhang, Lingming Zhang, Sriram Vishwanath, Grigore Rosu
EgoBench is an interactive egocentric multimodal benchmark designed to evaluate tool-using AI agents. It contains 1,590 video‑grounded tasks across five daily scenarios and a user‑agent‑tool environment that enforces joint visual perception, multi‑hop reasoning, and dynamic interaction. The benchmark includes a simulated user for realistic feedback and a deterministic validation framework, revealing that current video‑MLLM agents achieve only a 34.95% average Joint Success Rate, highlighting significant capability gaps.
By Yunqi Liu, Tong Niu, Zitong Wang, Zhenlong Dai, Yuqi Qing, Weiqiang Wang, Jian Liu
The paper introduces a clear definition for the task of converting natural language into First‑Order Logic (FOL), separating ontology extraction from logical translation to improve evaluation consistency. It surveys current datasets, evaluation metrics, and LLM‑based approaches—including fine‑tuning, prompting, and verification‑based refinement. The authors also outline remaining challenges in benchmarking, semantic evaluation, ontology‑aware methods, and practical applications.
By Andrea Brunello, Cristian Curaba, Luca Geatti, Michele Mignani, Angelo Montanari, Nicola Saccomanno
MemTrial is a portfolio‑management agent that learns when to trust its memory of past experiences. It evaluates each experience by drafting decisions with and without that experience, using a fractional factorial design to compute a Banzhaf‑value credit that isolates the experience’s true impact. By pooling noisy credits across dates with a hierarchical Bayesian model and only acting when predictions are reliable, MemTrial outperforms other experience‑learning agents and remains conservative when uncertain, achieving up to 21.2% higher utility and outperforming all LLM baselines on InvestorBench.
By Guanghao Wu, Zhuo Cai, Shoujin Wang
The paper re‑examines the METR time‑horizon metric, which reports the human completion time for software tasks that an AI can solve with 50% probability, by applying splines and item‑response theory to 228 tasks across 26 AIs. The authors find that the relationship between human time and AI difficulty is not strictly linear; the fitted spline is flat between 2–30 minutes and nearly linear outside that range, implying that a tenfold increase from 3 to 30 minutes is easier than from 30 minutes to 5 hours. They provide improved point estimates and diagnostic plots to assess the construct validity of time horizons, recommending that new benchmarks consider these diagnostics.
By Drew T. Nguyen, William Fithian
The paper introduces the Viability‑constrained Behavioral Steering Attack (V‑BSA), a new attack strategy for federated reinforcement learning that targets the timing of interventions rather than just the construction of malicious updates. By selecting sparse intervention states based on local policy uncertainty and applying envelope‑constrained steering, V‑BSA can significantly degrade performance on discrete‑action benchmarks while using far fewer interventions than traditional dense poisoning attacks. The study demonstrates that the specific states at which interventions occur constitute a distinct and critical dimension of robustness in sequential decision‑making settings.
By Shutong Zheng, Sijia Chen
The review examines how machine learning can adapt quantum combinatorial optimization workflows by learning decisions such as formulation, penalties, shot budgets, and backend selection. It categorizes the literature into three paradigms—AI for quantum optimization, quantum for AI-driven optimization, and AI‑quantum co‑optimization—and analyzes 119 papers, finding stronger evidence for AI assisting quantum methods than for quantum enhancing learned optimizers. The authors highlight limited experimental controls, introduce an M0‑M5 evidence hierarchy, and argue that scaling is a systems problem requiring careful allocation of classical and quantum resources.
By Hoong Chuin LAU
The paper introduces a Mixture-of-Experts (MoE) framework for model marketplaces, treating MoE as a market-level coordination mechanism rather than a learning architecture. Brokers use gating networks to combine heterogeneous experts into a composite service, while a cost-aware gating mechanism and a revenue allocation rule distribute residual revenue based on expert participation and execution cost. Experiments on fifteen tabular and image benchmarks show that MoE Market achieves higher welfare and lower expected cost than standard MoE, with competitive predictive performance and reduced computational overhead compared to exact Shapley allocation.
By Yizhou Ma, Wenbo Wu, Xikun Jiang, Zhuoqin Yang, Luis-Daniel Ib\'a\~nez
The paper introduces MIDAPN, a unified spatiotemporal forecasting backbone that integrates media-general graph adaptation and automatic temporal learning for both multivariate and multimodal time series. It employs a Multimedia Identity-Aware Graph to capture static, dynamic, and latent relationships across media, and a Spectral Prism Convolution for hierarchical temporal analysis. Extensive experiments on 13 multivariate and 12 multimodal datasets show MIDAPN consistently outperforms 16 state‑of‑the‑art TSF models and is compatible with various foundation models.
By Jierui Lei, Wenjian Zhang, Qingyi Yang, Yuyang Hong, Fangzheng Chen, Zhengbo Zhang, Haina Tang, Shiming Xiang
arXiv:2610. 11976v1 Announce Type: cross Abstract: We detail scalable methods for approximating the quadratic entropy $p^T d p$ for arbitrary distributions $p$ and common distances $d$ of negative type.
By Steve Huntsman
The paper introduces ReTurn, a benchmark of 7,000 multimodal tasks designed to evaluate how models selectively use conversational history. ReTurn includes pairs that test task-carrying history (Reconfirm/Reground) and evidence-carrying history (Retrieve/Rebind), each preserving the same target question, media, and answer. Experiments on 13 vision‑language and audio‑language models show a drop in open‑ended accuracy from 93.7% to 72.3% when history is involved, revealing that high question recall does not guarantee effective task application and that competing media can misdirect responses.
By Shuoyang Sun, Kerui Gu, Hao Fang, Shaoli Huang, Bin Chen
The paper investigates using raw, unlabeled video as a mid‑training phase for a pretrained language model. By encoding video frames into visual tokens and training the model to predict the next token, the authors mid‑train Qwen3‑1.7B on YT‑Temporal‑1B clips. The resulting model outperforms a non‑mid‑trained counterpart by 2.9 points on video benchmarks and 5.1 points on image benchmarks, while maintaining comparable text performance.
By Jaedong Hwang, Xiaoqian Shen, Ernie Chang, Changsheng Zhao, Chong Zhou, Saksham Suri, Qi Qian, Zechun Liu, Lemeng Wu, Qinsi Wang, Raghuraman Krishnamoorthi, Wei Wen
The paper introduces CUTS-GPR, a method for exact Gaussian Process Regression (GPR) in high-dimensional settings. It achieves near-linear scaling in the number of training points and low-order polynomial scaling in dimensionality by using an additive kernel on an incomplete grid, enabling efficient kernel matrix-vector products. Benchmarks demonstrate scalability to billions of data points and thousands of dimensions, with full GPR runs on up to 4.5 million points in 500 dimensions and practical application to ten high-dimensional potential energy surfaces completed in hours.
By Mads Greisen H{\o}jlund, August Smart Lykke-M{\o}ller, Henry Moss, Ove Christiansen
The paper introduces GAPE (Gated Adaptive Positional Encoding), a drop‑in augmentation for rotary positional encodings that adds content‑aware bias to attention logits while preserving rotary geometry. GAPE uses a query‑dependent gate to contract irrelevant context and a key‑dependent gate to keep salient distant tokens, leading to exponential attenuation of weakly protected context. Experiments show that GAPE improves long‑context robustness, extrapolating up to eight times the training length, and can be retrofitted into a pretrained 7B model without losing benchmark performance.
By Riccardo Ali, Alessio Borgi, Mario Severino, Pietro Li\`o, Christopher Irwin
DVLA-RL++ introduces Dual-Level Vision‑Language Alignment with Reinforcement Learning Gating for few‑shot learning. It adds complementary semantic purification (CSP) to separate intrinsic from nuisance textual descriptions and counterfactual reinforcement‑learning gating (CRG) to learn layer‑wise fusion strengths. Experiments on standard, fine‑grained, and cross‑domain benchmarks demonstrate state‑of‑the‑art accuracy, improving on DVLA‑RL by an average of 1.4%.
By Wenhao Li, Xianjing Meng, Qiangchang Wang, Zhongyi Han, Yilong Yin, Liqiang Nie
LinSlot proposes a joint framework for discovering object and attribute representations from raw images by leveraging the Linear Representation Hypothesis (LRH), which posits that composable concepts can be represented as linearly additive subspaces in slot representations. The method introduces a probabilistic model linking images, slots (objects), and blocks (attributes) and employs block attention to connect attribute representations to slots while incorporating LRH in both representation spaces. Experiments on multiple datasets show that LinSlot achieves effective disentanglement, provides empirical support for LRH in slot space, and enables image editing, outperforming state‑of‑the‑art methods in DCI scores.
By Sanket Gandhi, Utkarsh Giri, Varun Subramanium, Rohan Paul, Parag Singla