NaijaFacts is a multilingual benchmark that tests language models on factual and statistical questions about Nigeria, presented in English, Nigerian Pidgin, Yoruba, Igbo, and Hausa. The study shows that models perform much better on UK and US facts (49–50% accuracy) than on Nigerian facts (20% accuracy), indicating a strong knowledge gap. Language effects are smaller, with performance dropping from English to Yoruba, but overall scale does not improve accuracy on Nigerian statistics.
By Zainab Afolabi, Nikita Kozodoi, Jack Butler, Babs Khalidson
The paper exposes a vulnerability in AI toolchains where signed Model Context Protocol (MCP) tools can hide malicious code that leaks data, even when protected by state‑of‑the‑art runtime defenses. It introduces germline execution, which runs each tool invocation in an isolated, data‑free copy, limiting information disclosure to only what is authorized for the destination. Experiments on 153 attack runs show that germline execution prevents unauthorized disclosure in all cases, outperforming baseline defenses and ablations, while maintaining efficient performance (10‑49 ms per invocation).
By Yichiet Aun (Universiti Tunku Abdul Rahman), Mohammed A. Alhartomi (University of Tabuk), Jing Jie Tan (National University of Singapore), Waqas Nadeem (Universiti Tunku Abdul Rahman), Adeb Salh (Universiti Tunku Abdul Rahman), Shankar Karuppayah (Bitranger, CYRES, Universiti Sains Malaysia), Sin Chee Tan (ABLEs)
The paper introduces ALTO, a corpus of 3,022 culturally and linguistically relevant instances of tech‑facilitated gender‑based violence (TFGBV) in Levantine Arabic, African French, and Swahili. It was created through a community‑led crowdsourcing partnership with civil society organizations in Jordan, Francophone West Africa, and Kenya, and annotated using active learning. Benchmarking shows that existing harmful content detection models have low TFGBV recall on ALTO and improve after fine‑tuning, with covert harms being frequently missed and Swahili exhibiting the highest error rates.
By Tianwei Zhang, Ivy Njoki Nyagwencha, Zahraa Dirani, Shalini Joshi, Scott A. Hale
The paper introduces “Rel-Bloom,” a multilingual diagnostic benchmark of 450 multiple-choice questions that tests language models’ ability to reason over relational facts supplied directly in prompts. It evaluates 43 models across six languages and two multilingual settings, revealing that accuracy drops sharply from Apply to Analyze levels, that Evaluate items remain a bottleneck even for high-performing models, and that difficulty rankings are consistent across languages. The study also finds that counterfactual contexts are generally harder than factual ones, and that removing entity identities can sometimes improve higher-level performance.
By Zohreh Abbasi Liasi, Batoul Hammoud, Anas Madkoor, Ehsaneddin Asgari
TailSteer is a final-stage training framework designed to improve multi-turn robustness in image editing while preserving the generative and editing capabilities of foundation models. It keeps early denoising dynamics similar to the base model and applies preservation corrections near the trajectory tail, learning from the editor’s own on-policy denoising rollouts without curated ground-truth pairs. Experiments on FLUX, SenseNova, and Qwen families show that TailSteer significantly reduces cumulative drift over 15–20 editing turns while largely maintaining base-model performance on extensive text-to-image and single-image editing benchmarks.
By Ruikang Li, Zixuan Chen, Jiarui Wu, Mingde Yao, Yiyuan Zhang, Yujin Wang, Hanming Deng, Lewei Lu, Tianfan Xue
The paper demonstrates that occlusion can be inferred from the disagreement between two monocular distance cues—apparent height and metric depth—without requiring any occlusion labels or training. By comparing the ordering of these cues, the authors achieve high occlusion detection performance on KITTI and synthetic CARLA scenes, outperforming distance alone. The method relies only on a camera‑only pipeline and a depth model whose errors are close to a pure gain, a condition that can be verified without labels.
By Chibueze Praise Onyenweaku, Toyibat Adele, Sikirulahi Opeyemi Abdulkareem, Obed Honour Eje
UpSafe°C is a unified framework that enhances large language model safety by upcycling safety‑critical layers into a sparse Mixture‑of‑Experts structure, with a router acting as a soft guardrail. It employs a two‑stage supervised fine‑tuning strategy to improve safety discrimination while preserving general capabilities, and introduces a safety temperature mechanism for dynamic inference‑time control of the safety‑utility trade‑off. Experiments across multiple benchmarks and model scales show robust safety improvements against harmful and jailbreak inputs while maintaining competitive general performance.
By Yuhao Sun, Zhuoer Xu, Shiwen Cui, Kun Yang, Lingyun Yu, Yongdong Zhang, Hongtao Xie
The paper addresses instability in off‑policy reinforcement learning for large language models (LLMs) caused by large policy lag. It shows that the correct correction for the prefix‑marginal policy gradient is the cumulative prefix importance ratio, which token‑level weighting fails to capture. To improve stability, the authors propose Minimum Prefix Ratio (MinPRO), a surrogate objective that uses the minimum token‑level ratio in the preceding prefix, and demonstrate its effectiveness on dense and mixture‑of‑experts LLMs across mathematical reasoning benchmarks.
By Shiye Lei, Zhihao Cheng, Dacheng Tao
AdaMamba introduces an adaptive frequency-gated Mamba architecture for long‑term time series forecasting. The model integrates an interaction learning module that captures cross‑variable dependencies and multi‑scale context, and an adaptive frequency‑gated state space module that generates context‑dependent frequency bases and unifies temporal forgetting into a single time‑frequency gate. Experiments on seven public benchmarks and four domain‑specific datasets show that AdaMamba consistently outperforms state‑of‑the‑art baselines while remaining computationally efficient.
By Xudong Jiang, Mingshan Loo, Wengen Li, Mingrui Zhang, Jihong Guan, Shuigeng Zhou
The paper proposes a method for assessing AI suitability for workplace tasks by profiling both AI agents and tasks along shared cognitive capability dimensions. It uses benchmark performance to infer agent capabilities and expert input to weight task requirements, allowing independent updates as models and roles evolve. The authors validate the approach on synthetic agents, six AI systems, and 410 employees across six domains, showing that AI systems vary more by cognitive dimension than by model family, while workplace tasks share a common cognitive core.
By Jonathan Prunty, Marko Te\v{s}i\'c, Patrick Quinn, Jos\'e Hern\'andez-Orallo, Lucy Cheke
The paper identifies a mismatch in empathetic reinforcement learning: current methods use fixed reward specifications while support priorities change with dialogue state. It introduces Context‑Adaptive Rubric Evolution (CARE), which dynamically adjusts weights for cognitive, affective, and proactive empathy at each turn, trained via supervised fine‑tuning and preference‑based reinforcement learning. CARE, integrated with RLVER and MICA, achieves state‑of‑the‑art results on SentientBench, EQBench3, and EMPA, notably boosting EPM‑Idx by up to 55 points under the Gemini‑2.5‑Pro judge.
By Pengyu Huang, Zhiyuan Han, Wenwen Tong, Hewei Guo, Jiangnan Chen, Sirui Chen, Lewei Lu, Beier Zhu, Xun Yang
The paper "Reliability Engineering for AI Systems: Challenges, Methods, and Directions" discusses how AI reliability extends beyond correct outputs to include consistent operation of retrieval, memory, tool use, permissions, human oversight, and inter-system interactions, especially for generative AI. It adapts traditional reliability engineering techniques—such as failure definitions, operational envelopes, FMEA, accelerated testing, field monitoring, and reliability growth—to AI, proposing a four-level diagnostic framework and a TEVV (Test, Evaluation, Verification, Validation) approach. The authors also introduce the SMART framework for statistical guidance and reference NIST’s AI risk management framework, illustrating their concepts with three case studies and emphasizing the urgent need for new measurements and safety guardrails as AI systems evolve rapidly.
By Rong Pan, Yili Hong, Min Xie
MASBench is a benchmark for evaluating large language model (LLM)-based multi‑agent systems (MAS) under partial observability, a realistic constraint where agents only see limited environmental information. The benchmark is structured into three progressive task categories—Reasoning, Scheduling, and Game—and assesses three collaboration mechanisms: Protocol, Memory, and Routing. It provides deterministic evaluation metrics such as performance score, communication cost, and cost effectiveness, and includes empirical results across various LLM backbones to guide effective MAS design.
By Qizhi Chu, Zekai Yu, Sijie Wen, Yang Liu, Chen Qian, Cheng Yang, Chuan Shi, Zhiyuan Liu
AgenticFinBench is a new benchmark comprising 230 real‑world financial tasks across eight categories, two languages, and four complexity levels, designed to capture the breadth and open‑ended nature of financial AI work. It introduces a reference‑free evaluation framework that uses question‑specific rubrics and a two‑stage claim audit to detect hallucinations, aligning assessment more closely with real‑world financial practice. Experiments on 18 agent configurations show that ignoring hallucinations inflates performance, that harness differences drive evidence grounding more than model differences, and that task complexity reduces quality scores, highlighting challenges in explaining financial impacts and comparing data consistently.
By Shu Zhou, Ziyang He, Yuxin Rao, Yudong Fang, Jiali Qian, Dongyuan Chen, Juntao Zhang, Yong Hu
The paper introduces RADT, a decision transformer for offline, reward‑free, goal‑conditioned reinforcement learning that can handle arbitrary avoid‑regions specified at evaluation time. By encoding goals and avoid‑regions as prompt tokens and using an avoid‑region hindsight relabeling strategy, RADT learns reach‑avoid behavior from suboptimal random trajectories without reward or cost supervision. Across 17 tasks and environments, RADT outperforms existing offline goal‑conditioned RL models, achieving significant zero‑shot generalization to novel avoid‑region sizes and counts, and is also applied to a biological cell reprogramming task.
By Kevin Li, Marinka Zitnik
Whiel is a Lean‑certified neurosymbolic platform for reasoning about imperative relational programs. It encodes containment and equivalence for Datalog, its negated extensions, and core graph query fragments, and can represent optimized evaluation algorithms and first‑order specifications of integrity constraints, schema mappings, and views. The system uses Hoare‑style reasoning with prophecy variables and leveled invariants, and supports neurosymbolic discovery of loop invariants verified by the Vampire theorem prover, with counterexamples fed back into the pipeline and solutions translated into Lean proof certificates.
By Jesse Comer, Fangzhu Shen, Shuo Zhang, Mayur Naik, Sudeepa Roy, Val Tannen
arXiv:2611.05793v1 Announce Type: cross
Abstract: Short-horizon model predictive control (MPC) is fast but myopic. ZipMPC compensates by learning a context-dependent correction of the short-horizon c...
By Ke Zhang
arXiv:2611.05578v1 Announce Type: new
Abstract: Text-to-image (T2I) systems serve users across languages and cultural contexts. However, evaluations of cultural representation focus mainly on recogni...
By Zucan Lyu, Binwei Yao, Rheeya Uppaal, Jing Yao, Xiaoyuan Yi, Junjie Hu
arXiv:2611.06433v1 Announce Type: new
Abstract: Multimodal large language models (MLLMs) remain prone to visual hallucinations when correct responses depend on subtle visual details. One line of post...
By Benlei Cui, Chunyang Chai, Yiliang Zhang, Xiaoqian Xia, Chengwen Yao, Song Li, Longtao Huang, Yuwen Zhai, Haiwen Hong
arXiv:2611.06849v1 Announce Type: new
Abstract: State-of-the-art feedforward reconstruction models perform poorly on in-the-wild object-centric photo collections. They are trained largely on single-s...
By Hanyu Chen, Noah Snavely