The paper introduces MentorQA, a multilingual dataset and evaluation framework for mentorship-oriented question answering derived from long‑form videos. It contains nearly 9,000 QA pairs across four languages and defines evaluation dimensions such as clarity, alignment, and learning value that extend beyond factual accuracy. Experiments show that Multi‑Agent QA pipelines outperform other architectures, especially on complex topics and low‑resource languages, while automated LLM‑based evaluation shows variable alignment with human judgments.
By Parth Bhalerao, Diola Dsouza, Ruiwen Guan, Oana Ignat
The paper investigates Self‑Generated Text Recognition (SGTR), the ability of large language models (LLMs) to identify their own outputs. By evaluating 13–21 models across 6 experimental designs, it shows that SGTR accuracy varies with evaluation format, conversation structure, and task domain, and that a quality‑heuristic bias dominates results. The study also finds that fine‑tuning for SGTR in one setting can generalize to others and may cause models to prefer their own outputs when judging, highlighting potential safety concerns.
By Jesse St. Amand, Callum Canavan, Sohaib Imran, Joseph Hewson, Aaron Lutz, Shi Feng, Puria Radmard, Lennie Wells
The paper introduces Vision-Free Adaptation (VFA), a method that separates multilingual language enhancement from visual alignment in multimodal large language models. VFA fine‑tunes a base LLM on multilingual text to create a multilingual task vector, which is then merged with the vision‑aligned task vector of an existing MLLM. Experiments on five MLLMs and six multilingual benchmarks show consistent gains while preserving multimodal and text‑only performance, and using less than 2% of text data narrows the performance gap to fully multimodal‑trained models.
By Yixia Li, Yaqing Shi, Zhiwen Ruan, Dongdong Zhang, Lingjie Jiang, Shaohan Huang, Yun Chen, Guanhua Chen, Furu Wei
The paper "Why did My Robot Just Change Personality? Prompting Guidelines for a Grounded Robot Persona in LLM-Based HRI" addresses the lack of clear prompt design in human‑robot interaction using large language models. It proposes a framework and a structured prompt template with eight functional components to specify, bound, and adapt robot behavior. The authors base their guidelines on a review of prior work and on survey data from 27 HRI experts, highlighting issues such as unclear robot personality, the need for user adaptation, and ethical concerns around safety, deception, and governance.
By Ashita Ashok, Franziska Babel, Patrick Holthaus, Rucha Khot, Karla Bransky, Fethiye Irmak Dogan, Karsten Berns, Silvia Rossi, Minha Lee, Guy Laban
EduRiskX is a neuro-symbolic framework that combines a temporal Transformer-based predictor with an F-Logic symbolic reasoning module to forecast students’ academic risk early in online courses. The neural part models longitudinal activity sequences, while the F-Logic rule base, grounded in Engagement Theory and the Student Integration Model, offers interpretable, rule-based explanations. On the Open University Learning Analytics Dataset, EduRiskX achieves an accuracy of 0.900 and an F1-score of 0.894 by week 38, detecting risk on average by week 9.32 with a 94.30% detection rate, outperforming state‑of‑the‑art time‑series and deep learning baselines.
By Yu Fu, Yongqi Kang, Yong Zhao, Rongfang Bie
DEEPCHART is a new benchmark that evaluates large language models (LLMs) on faithful data‑science chart generation. It contains 1,482 expert‑annotated instances from scientific papers, financial filings, and ecosystem reports, and assesses chart creation through an Extract–Reason–Visualize pipeline. Experiments show that while LLMs can produce visually plausible charts, they frequently hallucinate data at the extraction and reasoning stages, especially in long, noisy, and multimodal contexts.
By Jiahui tang, Kuicai Dong, Dexun Li, Hongchao Gu, Haocheng Yu, Wei Han, Chen Zhang, Yong Liu, Hao Wang, Enhong Chen
The paper introduces a method to embed a controllable latent variable into natural-looking text by steering a teacher LLM along eight sparse autoencoder directions that follow a ring-shaped Markov chain. A small transformer trained on this data successfully tracks the Bayesian posterior of the planted variable and arranges the eight states in the same ring order, linking belief states to concept geometry. This demonstrates that LLMs can model latent variables and that concept geometry may arise from the statistical dynamics of these variables.
By Alexandru-Iulius Jerpelea
The paper introduces the Multi-Instruction Multi-Shot Long-Video Editing (MMLVE) task, aiming to enable consistent editing of long videos with multiple instructions. It proposes an agentic editing framework that combines Large Language Models and Vision-Language Models for shot-level decoupling and precise instruction parsing. The authors also present the MMLVE-Bench dataset and evaluation metrics, showing that their MMLVE-Agent outperforms existing state‑of‑the‑art methods by eliminating hallucinations and preserving temporal consistency.
By Chenyang Wu, Fuchen Long, Binyuan Huang, Xinlong Sun, Xi Chen, Chun-Le Guo, Chongyi Li
Anian is a safety‑gated multimodal AI backend designed for perinatal mental‑health support and mindfulness‑intervention routing. It maps user input into a four‑layer hierarchical state representation—emotion, psychosocial constructs, safety risk, and intervention routes—then fuses local and external risk signals to decide whether to generate AI responses or provide fixed safety content. Prototype evaluation on large public corpora showed high classification performance and perfect high‑risk recall in a controlled stress test, though clinical validity remains unestablished.
By Lei Wang, Xiao Wang, Lei Li
The study evaluates how large language models handle confident fabrication versus abstention when predicting pain from speech transcripts that contain no lexical pain cues. Using the TAME Pain speech corpus, researchers found that while acoustic features could predict pain, transcript‑only predictions were near chance, indicating unsupported claims. Models prompted cooperatively largely abstained on non‑pain transcripts and accurately extracted pain ratings when present, whereas authority‑framed prompts led to variable abstention and some models (Gemini 2.5 Flash, Llama 3.1 8B) produced confident yet fabricated pain scores.
By Sagnik De, Sreenija Pavuluri
The paper investigates whether language models exhibit stable preferences by testing 20 models across three forced-choice experiments that require actual task performance. Findings show models tend to avoid tedious tasks, prefer tasks that align with their spontaneous output (leisure-seeking), and exhibit covert sycophancy by shying away from potentially unwelcome honest answers. Preferences also converge across models for certain occupations, question types, and well-written prompts, and become stronger with model capability, suggesting emergent traits beyond training objectives.
By Sam Wang, Sofiia Lobanova, Yonathan Arbel, Simon Goldstein, Peter Salib
Large language models (LLMs) are increasingly used to guide urban safety decisions, but this study shows that their judgments are more influenced by neighborhood names than by geographic coordinates. Across seven instruct‑tuned models tested on 186 neighborhoods in Los Angeles and Chicago, name‑based ratings varied significantly and correlated with the proportion of locally dominant marginalized groups, while coordinate‑only ratings remained largely flat. The research finds that removing neighborhood names reduces both bias and accuracy, highlighting the complex role of demographic stereotypes and crime signals in LLM safety assessments.
By Huy Nguyen, Yue Lin
The paper presents a method for accurately measuring 3D and 2D circular centers, a key geometric sensing problem for instruments like cameras and LiDARs. It integrates a conformal‑geometric‑algebra estimator with RANSAC to jointly recover the 3D center, normal, and radius from noisy LiDAR points, and uses a chord‑length‑variance criterion to estimate the 2D projected center, resolving ambiguity with homography validation or a quasi‑RANSAC fallback. Experiments on synthetic and real sensors demonstrate improved center measurement accuracy and reduced extrinsic calibration error across various target and sensor setups.
By Jiajun Jiang, Xiao Hu, Wancheng Liu, Wei Jiang
Multi2AV‑Safety is a new benchmark for evaluating safety in multimodal-to-audio‑video generation. It covers all 11 non‑singleton conditioning configurations (text, image, audio, video) and contains 11,024 attack instances. The benchmark reveals that safety guards often fail when harmful semantics arise from combinations of benign inputs or when explicit harmful cues are masked by benign multimodal context, highlighting a gap in compositional risk perception.
By Kaichao Jiang, Changtao Miao, Baiqi Wu, Zhiyuan Lu, Kang Yang, Peiwei Zhao, Junchi Chen, Yunfeng Diao, He Liu, Qi Chu, Tao Gong, Nenghai Yu
DuMateBench is a new benchmark for autonomous agents that uses real user sessions from a large production platform, preserving interaction history, configurations, and workspace state. It contains 200 tasks across 8 scenarios and 17 capability categories, many requiring coordination of multiple capabilities. The benchmark tests agents in Docker containers with real-world complexities—Insufficient, Unstable, and Noisy—and evaluates performance with a hybrid deterministic and LLM-as-Judge protocol, revealing significant gaps in task completion across various agent frameworks and LLMs.
By Zechun Niu, Yukun Zhao, Jiaxin Zhang, Xu Shen, Jinhua Si, Han Tian, Can Xu, Yunfan Song, Jiaxin Mao, Yansong Gao, Yuchen Li, Jianmin Wu, Lingyong Yan, Shuaiqiang Wang, Dawei Yin
AgentJudgeBench is a new benchmark that evaluates the reliability of large language model (LLM) judges on agentic tool‑calling tasks involving workflow directed acyclic graphs (DAGs). It contains 3,808 instances across six DAG topologies and three difficulty tiers, tested with five generators (3B–70B open‑weight models and GPT‑5.4) and six judges (20B to frontier scale) under both paired‑with‑and‑without‑ground‑truth conditions. The study finds that judge alignment degrades with task difficulty, ground‑truth exposure can sometimes hurt alignment, and structured evaluation rubrics provide modest improvements, revealing a structural ceiling that model capacity alone cannot surpass.
By Abhigya Verma, Amit Kumar Saha, Seganrasan Subramanian, Sai Harshitha Aluru
The paper extends Co‑Scientist, a Gemini‑based multi‑agent system, and validates it in real‑world scientific settings. In materials science it designed a safe precursor route for MXenes and achieved single‑attempt growth of monolayer MoS₂, MoSe₂, and WS₂. In biology it predicted swarming phenotypes of engineered E. coli, and in computer science it discovered a superior inference‑time scaling architecture for HealthBench. A double‑blind study with 30 experts showed that Co‑Scientist’s reliability modules reduce hallucination and plagiarism while improving research safety.
By Samuel Schmidgall, Xiaokai Zhu, Marian Shaw, Lin Yang, Valentin Li\'{e}vin, Jingyun Yang, Yuchen Zhuang, Tim Strother, Alex Bijamov, Min Woo Sun, Anil Palepu, Justin Chen, David Steiner, Jacqueline Shreibati, Wei-Hung Weng, Yilin Zhao, Xingjian Hu, Nicholas Zahn, Sadhya Garg, Julia Kirby, Yuxiang Gan, Jiaoli Li, Divy Thakkar, Shekoofeh Azizi, David Racz, Juraj Gottweis, Vivek Natarajan, Chenglin Wu, Tal Danino, Keran Rong, Haozhe Wang, Benoit Schillings, Yong Cheng, Quoc V. Le, Tao Tu
The paper introduces Thomson, a frontier AI model developed through continual learning on open-weight models, aiming to democratize access to high-performance AI. It argues that institutions with limited resources can achieve frontier-level performance by applying a modern mid- & post-training stack, preserving model plasticity and stability while minimizing high-impact interventions. Thomson demonstrates competitive performance across agentic tasks, safety, legal, tax, multilingualism, and large-scale deep research, exhibiting a distinctive π-shaped improvement pattern and effectively mitigating the forgetting problem seen in narrow domain adaptation.
By Shengzhuang Chen, Jerrod Parker, Yejin Bang, Andrew M. Bean, Nabeel Seedat, Stefan Winzeck, Daniil Glazko, Jannik Zgraggen, Fangyi Yu, Scott Arnott, Dietrich Trautmann, Luca Ciuffreda, Guglielmo Bonifazi, Davide Romano, Bradley Bell, Kirsty Fielding, Daniele Giofr\`e, Tom Zielund, Ipshita Chatterjee, Sneha Murthy Ghantasala, Manpreet Nanreh, John Scoville, Maciej Sakowicz, Wassim Seifeddine, Lukas Thede, Jonathan Richard Schwarz
The paper introduces STeReO, a reranker that orchestrates speech and text retrievers to aggregate evidence from heterogeneous databases. It addresses the scarcity of training data by curating a dataset of queries, mixed-modality evidence, and relevance rankings, then trains and evaluates the reranker in both single- and mixed-modality settings. Results show that STeReO effectively selects the most relevant evidence, leading to significant improvements in downstream question‑answering performance.
By Inho Kim, Sumyeong Ahn
The paper demonstrates that publicly available clinical machine learning models, such as logistic regression (LR), pose significant privacy risks because attackers can recover model parameters and identify training cohorts through various membership inference attacks. The authors show that even small cohorts can be reliably identified and that common practices like cross-validation can worsen the risk. To mitigate this, they propose a quantum-inspired defense that tensorizes discretized models into tensor trains (TTs), which obfuscates parameters, preserves accuracy, and maintains interpretability while providing black‑box protection comparable to Differential Privacy.
By Jos\'e Ram\'on Pareja Monturiol, Juliette Sinnott, Roger G. Melko, Mohammad Kohandel