SalsaAgent is a multimodal embodied language model that generates expressive, full‑body salsa follower motions in response to a human leader and music. The approach treats partner interaction as nonverbal token passing, extending a large language model’s vocabulary to include discrete motion, pairwise relation, and audio tokens. A two‑stage token‑to‑diffusion pipeline, combined with full‑body and pairwise‑relation tokenizers and alignment with automatically derived text descriptions of skeleton dynamics, yields improved motion quality, spatial coordination, and music‑partner synchrony compared to prior baselines.
By Payam Jome Yazdian, Zoe Stanley, Angelica Lim
The paper introduces AttnSign, a vision‑language model that improves sign language translation by steering spatial‑temporal attention. It first supervises attention on sign‑relevant regions such as faces and hands in each frame, then uses an RL‑based motion‑cadence method to focus on keyframes. Experiments on How2Sign and OpenASL show AttnSign consistently outperforms existing methods.
By Meibo Hu, Guohao Sun, Annemarie D. Ross, Sheng Li, Zhiqiang Tao
HunyuanOCR-1.5 is a lightweight, end‑to‑end OCR‑specialized vision‑language model that unifies document parsing, text spotting, information extraction, text‑image translation, and multi‑image document understanding. It builds on the HunyuanOCR‑1.0 architecture, improving efficiency with DFlash‑based OCR decoding for faster inference (6.37× Transformer speedup, 2.14× under vLLM) and enhancing capability through an Agentic Data Flow system that autonomously constructs high‑quality training data for long‑tail OCR tasks. The model achieves top‑tier performance on OmniDocBench v1.6 and sets new milestones in ancient‑script OCR, chart/table parsing, multilingual parsing, and hallucination evaluation, while remaining lightweight for deployment.
By Gengluo Li, Xingyu Wan, Shangpin Peng, Weinong Wang, Hao Feng, Yongkun Du, Binghong Wu, Zheng Ruan, Zhiqiong Lu, Liang Wu, Pengyuan Lyu, Huawen Shen, Zibin Lin, Shijing Hu, Jieneng Yang, Hongbing Wen, Guanghua Yu, Hong Liu, Bochao Wang, Can Ma, Han Hu, Chengquan Zhang, Yu Zhou
The paper presents a portfolio-based approach to constrained multi-objective Bayesian optimization for materials design, framing acquisition‑function selection as an adaptive policy problem. Two controllers—UCB‑Bandit, a modified UCB multi‑armed bandit, and Agentic‑Switch, a multi‑agent system powered by a large language model—were tested against fixed‑policy baselines on synthetic benchmarks and two real materials design case studies. The adaptive policies achieved competitive results in cumulative feasibility counts and feasible hypervolume improvement, outperforming individual acquisition functions that excelled only in a single metric.
By Sushant Sinha, Christofer Hardcastle, Robert Robinson, Shakti Prasad Padhy, Brent Vela, Douglas Allaire, Raymundo Arroyave
The study examines how conversational LLM agents—specifically ChatGPT, Claude, Grok, and DeepSeek—use Web search, combining real user interactions with controlled API experiments. It finds that agents differ in when they decide to search, how they craft queries, and which domains they favor, and that more frequent searching does not always improve answer quality. While most responses are grounded in search results, some claims are unsupported, raising attribution concerns.
By Mahsa Amani, Seungeon Lee, Abhisek Dash, Asmaa El Fraihi, Yunah Jang, Elisabeth Kirsten, Qinyuan Wu, Krishna P. Gummadi, Manish Gupta, Abhilasha Ravichander, Muhammad Bilal Zafar, Soumi Das
The study examines a two‑agent résumé screening process where both employer‑side and candidate‑side agents exchange evidence before deciding who advances, contrasting it with the traditional one‑call automated screening. Using GPT‑5.5 and Claude Opus 4.7 on 600 constructed résumé‑job pairs, the two‑agent method increased the proportion of applications advanced (up to 39.3% for GPT‑5.5) and raised pass rates for borderline cases from 4.5% to 26.2% (GPT‑5.5) and 6.5% to 16.1% (Opus 4.7). The results show that the screening procedure itself, rather than just the underlying model, determines which candidates reach human review and how consistently that access recurs.
By Jian Gao, Hang Jiang
QVAC Genesis III is a 191.43 B‑token synthetic STEM corpus covering 19 domains and multiple difficulty levels, created through a dual generation strategy that uses a weak edge‑scale student model to generate corrective explanations and contrastive reasoning. The authors evaluate the corpus with an LLM‑as‑a‑parser protocol and demonstrate that 1.7 B‑parameter models trained on QVAC Genesis III outperform those trained on Cosmopedia‑v2 and the Cosmo‑1B model on ARC, GPQA Diamond, and MMLU STEM benchmarks, achieving up to +28.57% improvement on ARC‑E and a 99.45% valid answer rate.
By Davide Vitabile, N. Ranjan, Akshay Nambiar, Kamal K. Gupta, Amril Nazir
EconSkills is a skill library and evaluation framework that transforms verified EconWebArena trajectories into parameterized standard operating procedures for retrieving live economic data. Each skill captures scope, navigation steps, site-specific guidance, verification checks, and recovery steps, replacing instance-specific values with placeholders. The framework evaluates skill transfer to new tasks and the effectiveness of library retrieval, showing that matched skills improve success rates and reduce steps compared to no-skill prompting, while abstraction outperforms raw trajectory replay. At scale, retrieval performs competitively overall, excelling on directly covered tasks and compensating for uncovered tasks through approximate matches.
By Yinzhu Quan, Zefang Liu
The paper introduces a hierarchical hybrid architecture combining large language models (LLMs) and multi-agent reinforcement learning (MARL) to manage heterogeneous unmanned aerial systems in low‑altitude wireless networks (LAWNs). An outer LLM‑driven loop interprets service requirements and operator intent to reconfigure objectives and resource priorities, while an inner MARL loop executes decentralized policies under the updated game. A logistics‑monitoring case study demonstrates the framework’s ability to coordinate diverse services and adapt to changing conditions without retraining the MARL policies.
By Nguyen Duc Minh Quang, Chang Liu, Shuangyang Li, Derrick Wing Kwan Ng
AutoData is an agent that autonomously searches for pre‑training data selection algorithms by exploring a program space of scoring, stratification, and stochastic rules. It iteratively refines these algorithms using validation feedback from a proxy model, discovering feature interactions that outperform existing human‑designed curation pipelines. The resulting selection recipe, found in an overnight search, transfers to larger scales and improves the downstream CORE metric.
By Yan Meng, Dhruv Srikanth, Bingchen Zhao, Zhengyao Jiang, Yuxiang Wu
MaSCoD is a multi‑agent framework that explicitly organizes candidate third variables and local structural patterns before making direct‑edge judgments in causal discovery. Using GPT‑5.4 and GPT‑4o on datasets such as Auto‑MPG, DWD, and Sachs, the study shows that providing structural hypotheses in a separate phase (Full) generally improves recall and F1 scores compared to constructing them during judgment, though it can also raise false‑positive rates. Ablation experiments reveal that the benefit of supplying both structural components is not guaranteed, and the advantage varies across datasets and model backbones.
By Yudai Nakada, Yuichiro Nishiura, Jin Michael Splichal
PAPC is a platform-mediated mechanism designed to mitigate privacy-propagation externalities in AI‑mediated workflows. It intercepts information‑moving events—such as memory writes, shared‑workspace updates, inter‑agent messages, or tool events—before they update shared state or external channels, using policy, provenance, topology/fanout, privilege, and content signals to decide whether to release a policy‑safe abstraction, quarantine raw content, block a transition, or narrow onward rights. Experiments on retrieval‑memory and multi‑agent workflow benchmarks show that PAPC preserves deterministic task completion while eliminating measured exact raw‑value and external raw‑value exposure, demonstrating that event‑level mediation can serve as a platform‑governance primitive for agent‑mediated online work.
By Tao Huang, Guosen Wu, Chen Hou, Guolong Zheng
The paper introduces PaRID, a software-directed soft error detection framework that uses parallel instruction duplication and LLM-tuned performance modeling. PaRID is designed to work at compile time for multithreaded programs, handling mixed serial and parallel regions while keeping performance overhead low. Experiments on NPB benchmarks show that PaRID cuts protection overhead from 162.79% to 59.84% on average and can achieve up to a 5× speedup while still detecting all errors.
By Yafan Huang, Guanpeng Li
The paper presents an autonomous AI agent that combines ReAct and Retrieval-Augmented Generation to classify the severity of genetic diseases using 10,211 Human Phenotype Ontology terms. It applies ACMG severity guidelines and ACOG quality-of-life criteria, retrieving PubMed literature to produce interpretable reasoning chains and verify claims. The agent achieved 93.55% accuracy on phenotype-level classification and identified 3,283 autosomal recessive gene pairs with severe presentations, with 95.2% concordance in external validation.
By Tohid Ghasemnejad, Ahmadreza Argha, Mark Grosser, John Wang, Min Yang, Thantrira Porntaveetus, Tony Roscioli, Nigel H. Lovell, Mahmoud Aarabi, Hamid Alinejad-Rokny
The paper introduces Abstract Token Curriculum (ATC), a curriculum learning framework that enables large language models to develop continuous intermediate representations—referred to as abstract thoughts—without explicit supervision or manual scratchpad design. ATC incrementally raises problem difficulty through a sequence of distributions, guiding models to focus attention on the most informative tokens for predicting subsequent tokens. The authors provide theoretical analysis for parity function learning with single‑layer softmax attention and demonstrate ATC’s effectiveness on graph reachability and arithmetic learning tasks.
By Khashayar Gatmiry, Avrajit Ghosh, Parsa Mirtaheri, Jason D. Lee, Nika Haghtalab, Emmanuel Abbe, Peter Bartlett
PetriBench is a compact, fully self‑contained, and scalable benchmark that evaluates large language model (LLM) reasoning over dynamic state spaces using Petri nets. It organizes reasoning into four task families with Easy, Medium, and Hard levels, each generated by increasing structural complexity and evaluated against exact ground truth. Experiments across proprietary and open‑weight models show that accuracy consistently drops with difficulty, revealing distinct task‑specific capability profiles, while test‑time compute and procedural generation affect performance differently across tasks.
By Pyrros Koussios, Benjamin J\"ager, John Hua Yao, Ajay Sridhar, Violet Xiang, Chenhao Li
The paper shows that the wording of prompts in vision‑language models (VLMs) can either improve or worsen robustness to image corruption. Verbose prompts broaden the cross‑modal attention’s frequency filter, making the model less sensitive to corruptions, while semantically complex prompts narrow the filter and increase vulnerability. Experiments on Qwen3‑VL and LLaVA‑OneVision confirm that adding padding or verbose phrasing reduces answer drift by 70–81% on 8B models.
By Farooq Ahmad Wani, Maria Sofia Bucarelli, Mujtaba Hussain Mirza, Oleksandr Pryymak, Aryo Pradipta Gema, Iacopo Masi, Pasquale Minervini, Fabrizio Silvestri
The paper introduces Lens, a training‑free framework that aligns multimodal representations with the semantic perspective required by downstream tasks. Lens uses a task‑specific readout phrase to anchor the perspective and then aggregates token states after the full input, ensuring the extracted representation reflects task‑conditioned evidence integration rather than generic salient content. The method achieves a Precision@1 of 63.9 across 36 MMEB datasets, outperforming the nearest training‑free baseline by 10.2 points.
By Xinran Liu, Shouqian Shi, Yixian Chen, Ruizhi Chen, Xin-Wei Yao, Sheng Zhong
arXiv:2609. 20278v1 Announce Type: cross Abstract: Text, knowledge graphs, and hypergraphs all have elements that play distinct roles within relation instances, structure that is lost when data is flattened into token sequences.
By Mahesh Godavarti
Edustories is a dataset of 1,492 teacher‑written case studies that detail real elementary and high‑school classroom situations involving challenging student behavior, pedagogical interventions, and their outcomes. The collection is designed to enable research on AI assistance in collective teaching contexts, such as evaluating large language models’ ability to predict the success of teacher interventions. Comparative tests show that current models achieve 58% accuracy, below the 64% accuracy of human experts, indicating a gap between AI and human expertise in predicting classroom outcomes.
By Michal \v{S}tef\'anik, Jan Nehyba, Jirina Karasova, Martin Fico, Lucie \v{S}karkov\'a, Mark\'eta Ko\v{s}atkov\'a, David Kosatka