GUI-CC is a benchmark designed to assess the contextual consistency of GUI world models when used as agent environments, rather than just one‑step next‑screen predictors. It includes two tracks: an offline reference‑action track that rolls models along real mobile GUI trajectories, and an online agent‑loop track where fixed probing agents interact with model‑generated UIs. The benchmark evaluates transition fidelity, plausibility, contextual consistency, and task progress across 500 offline trajectory tasks and 200 online tasks spanning 30 mobile apps.
By Lin Fu, Zheyuan Yang, Tianhui Zhang, Jinbiao Wei, Guo Gan, Boxu Liu, Yilun Zhao, Yu Rong
CUDA‑Harness is a framework that enables the generation and optimization of CUDA kernels directly from natural language. It introduces Intermediate‑Structured Generation to bridge high‑level semantics with low‑level kernel code, uses Synthesis‑Based Verification to mitigate reward hacking by providing isolated test data, and employs Feedback‑Adaptive Evolution to prioritize correctness while improving performance. Experiments show the approach generalizes across different large language models, hardware platforms, and even supports C‑to‑CUDA transpilation.
By Qi Fan, An Zou, Yehan Ma
The paper introduces Scientific Agent Skills, an open library of 163 procedural knowledge modules across 16 scientific domains such as genomics, cheminformatics, medical imaging, study design, and scientific communication. Each skill is organized as a directory containing a versioned, human‑readable instruction file, reference material, and runnable scripts, allowing a language‑model agent to load the appropriate procedure only when needed. The library is publicly licensed and hosted on GitHub, but the authors report no task‑level evaluation or host selection rate.
By Timothy Kassis, Vinayak Agarwal, Yuhuan He, Darshil Patel, Aubrey M. Brueckner
The paper investigates how large language models (LLMs) can propose changes to Kubernetes configuration files in a GitOps workflow. It shows that having the model directly generate edited files or diffs is unsafe for unattended automation, as current text‑generation strategies either fail to apply patches or silently misapply them. The authors propose a deterministic, minimal‑diff approach where the model emits only a structured intent for a field change, and a pipeline locates and replaces the exact character span in the raw YAML, preserving formatting and comments while guaranteeing correctness and O(1) generation cost.
By Pruthvi Davineni
The paper introduces CoLT-Drive, a 3,536-sample counterfactual long‑tail benchmark for evaluating decision‑level driving affordance prediction, which tests whether models can infer how rare objects affect an ego vehicle’s high‑level actions. It also proposes KPA, a knowledge‑preserving adaptation framework that combines structured prompting, expert merging, and a regime‑aware LoRA mixture‑of‑experts module to improve small VLMs on driving tasks. Experiments show KPA achieves 60.8% pair accuracy on CoLT‑Drive, outperforming the Qwen3‑VL‑2B baseline and LoRA SFT while keeping competitive in‑domain performance.
By Zhengxu Tang, Guofeng Cui, Ziyu Gong, Xiaozhou Zhang, Ruifeng Deng, Chengzhi Qi, Ke Chen, Sachin Patil, Tianjun Xiao, Langechuan Liu, Pichao Wang
The paper introduces Fed‑LSVI, a federated online reinforcement learning algorithm that uses linear function approximation in episodic Markov decision processes. It achieves a regret bound of ≥O(√{Md^3H^4T}) while only exchanging compressed sufficient statistics, thereby meeting privacy constraints. The method reduces communication cost to logarithmic in the number of episodes, a marked improvement over previous approaches that required linear communication.
By Zihang Liang, Haochen Zhang, Lingzhou Xue
The paper introduces WHALE, a method that alternates between updating a language model’s weights and searching for a better harness (the code that manages context and control flow). By iteratively fine‑tuning the model under the current harness and then optimizing the harness under the updated model, WHALE improves performance across search QA, math reasoning, and chess puzzles, outperforming weight‑only, harness‑only, and Fast‑Slow Training by 4.15–24.38 percentage points in mean@8 accuracy. The approach uses either fixed phase lengths or an adaptive patience rule to decide when to switch phases, and the authors provide code on GitHub.
By Haechan Kim, Yoonho Lee, Gisang Lee, Chelsea Finn, Kangwook Lee
The paper introduces a benchmark for topic matching in real-world ASR transcripts from contact centers, where noisy, punctuation‑free speech data must be classified into predefined topics. It presents a human‑annotated dataset of topic‑utterance judgments and evaluates three matcher types—regex, zero‑shot sentence embeddings, and Gemini‑based LLMs—using two topic representations: keyphrases and natural language descriptions. Experiments show that lightweight LLM matchers outperform the other methods, especially when natural language descriptions are used.
By Saman Rahbar, Xiliang Zhu, Irvin Cardoza, David Rossouw
The paper introduces WM‑RMoE, a World Model‑based Risk‑aware Mixture‑of‑Experts framework for autonomous overtaking. It uses a learned latent dynamics model to perform multi‑step rollouts, evaluating cumulative risk at the trajectory level, and employs a hierarchical gating mechanism to coordinate long‑, short‑horizon, and rule‑based safety experts. A Gaussian Mixture Model preserves multimodal maneuver branches, improving robustness and preventing behavioral averaging, leading to better safety compliance, decision stability, and generalization in experiments.
By Yongzhi Liu, Sunan Zhang, Jinchang Xu, Jiawei Wang, Yushu Qiu, Chen Lv, Weichao Zhuang
WiseSpec is a requirements‑driven agent framework designed to improve repository‑level code generation. It automatically builds structured, information‑rich requirements, evaluates their quality via execution‑based tests, and iteratively refines them to better guide code generation. Experiments show WiseSpec outperforms all baselines, achieving an average 13.17% improvement in %Resolved.
By Zhao Tian
HarnessEvolve is a self‑evolving framework that improves agent harnesses—prompts, skills, tools, and execution logic—by learning from reference trajectories. It separates execution, evaluation, optimization, and gating into independent modules, addressing credit assignment failure, shortcut learning, and catastrophic forgetting. The approach uses reference trajectories to extract error signals, applies quality and performance gates to candidate updates, and validates updates on held‑out data, consistently outperforming state‑of‑the‑art baselines across diverse benchmarks.
By Wen Jiang, Mingmin Chu, Yimeng Tian, Qianxin Zhang, Haofei Yang, Rui Yang, Yang Liu, Tao Lv, Fangming Li
DualStake introduces a dual-path confidence calibration for deep research agents, adding step confidence elicitation after each retrieval step. The method shows that evidence confidence (E-Conf) after the final retrieval provides a stronger uncertainty signal than answer confidence (A-Conf), and that A-Conf is largely influenced by E-Conf. By applying margin‑clipped, confidence‑dependent stake rewards, DualStake aligns both E-Conf and A-Conf with answer correctness, improving calibration across multiple QA benchmarks without harming accuracy.
By Yinuo Xu, Yuwei Liang, Jianjie Cheng, Meng Wang, Yongcan Yu, Shuo Lu, Jian Liang
The paper introduces a cross‑modal pseudo‑labeling pipeline for unsupervised domain adaptation in semantic segmentation, particularly for waste sorting. It combines SAM for class‑agnostic region proposals with EVA‑CLIP to assign semantic labels via region‑text similarity, applying confidence filtering to ensure reliable pseudo‑labels for self‑training. An optional BLIP‑based language‑grounded verification further refines ambiguous regions, and the method shows consistent improvements over source‑only baselines on synthetic‑to‑real driving and lab‑to‑factory waste sorting shifts.
By Udo Schlegel, Shubhangi, Gabriel Dax, Sai Rahul Kaminwar, Florian Karl, Thomas Seidl
The paper proposes CANOPY, a minimalist reinforcement learning protocol that addresses two common pitfalls—signal starvation and policy drift—in outcome‑only RL for long‑horizon interactive tasks. By scaling same‑task exploration, keeping updates on‑policy, and anchoring updates with KL divergence, CANOPY enables a Qwen3‑14B agent to achieve top leaderboard results on the AppWorld coding benchmark without auxiliary supervision or elaborate scaffolding. The approach also improves performance on SWE‑bench for a Qwen3.5‑9B model.
By Liming Pu, Xiaoxia Li, Yifu Liu, Teng Cao, Bin Yang
The paper introduces Online Hyperparameter Optimization (OHPO), framing it as an infinitely many‑armed bandit problem over mixed and conditional search spaces. It proposes the IMABO framework, which couples any bandit policy with any oracle for proposing new configurations, and presents IMOSS—a restart‑free anytime policy with provable regret bounds. Experiments show that IMABO, combined with practical oracles such as TPE, an incumbent‑mutation oracle, and a pretrained tabular foundation model, outperforms random search across a range of settings from classical ML models to LLM‑based agents.
By Louis Abraham, Tuan-Anh Nguyen, Nicolas Devatine
arXiv:2609.01588v1 Announce Type: cross
Abstract: Writing involves diverse cognitive activities, from ideation to revision, and writers' needs vary across individuals and moments. Proactive AI promis...
By Chao Zhang, Abe Davis, Chih-Wei Chen, Chin-Chia Hsu
arXiv:2609.01595v1 Announce Type: cross
Abstract: We develop a framework for mechanism design with AI agents whose alignment (preferences) and capabilities (feasible actions and information) are unkn...
By Dirk Bergemann, Andrew Koh, Stephen Morris
arXiv:2609.01600v1 Announce Type: cross
Abstract: Dynamic agent harnesses let language models change the software that shapes their own execution. This flexibility brings a new reasoning burden: a lo...
By Damien Sileo, Dimitri Kachler
arXiv:2609.01603v1 Announce Type: cross
Abstract: Evaluating software engineering agents on realistic benchmarks is costly, since each task may require multi-step code exploration, modification, and...
By Kefeng Duan, Dewu Zheng, Yanlin Wang, Xiwen Wang, Ensheng Shi, Xilin Liu, Yuchi Ma, Jiachi Chen, Mingwei Liu, Zibin Zheng
arXiv:2510.15144v4 Announce Type: replace
Abstract: Simulating human reasoning in open-ended tasks has long been a central aspiration in AI and cognitive science. While large language models now appr...
By Chance Jiajie Li, Zhenze Mo, Yuhan Tang, Ao Qu, Jiayi Wu, Kaiya Ivy Zhao, Yulu Gan, Jie Fan, Jiangbo Yu, Hang Jiang, Paul Pu Liang, Jinhua Zhao, Luis Alberto Alonso Pastor, Kent Larson