Hi3D 3.0, also called Twinkle3D, is an image‑to‑3D generation system that produces watertight triangle meshes at a resolution of 2048³. It improves on previous methods by maintaining high representational precision while fixing surface quality and non‑watertight issues, scaling diffusion generation to 300K geometric tokens, strengthening global shape and local detail in a single stage, and adding a fine‑grained image‑3D cross‑modal interaction to better recover object‑specific structures such as inscriptions. The system outperforms four commercial competitors, recovering 82.1 % of inscribed characters with 98.2 % precision versus 21.7 % recall for the best rival.
By Ziying Li, Shengchu Zhao, Huiang He, Yiyang Chen, Jianwen Huang, Bailin Li, Changhao Li, Jianhui Li, Jie Li, Ruiyang Liu, Yibo Luo, Tengjiao Sun, Pei Tang, Shiwen Wang, Jiaqi Wu, Kang Wu, Kaiqiao Yang, Zherui Yang, Hu Zhang, Xuezhi Zhao, Xinhe Zheng, Yukun Li, Heliang Zheng, Rongfei Jia
The paper investigates how narrative wrapping can trick safety‑aligned language models into refusing a harmful request only when it is presented directly, but answering it when embedded in a role‑play or narrative context. It introduces GUISE, a multilingual benchmark that tests this vulnerability across English, modern Chinese, and Classical Chinese, and shows high attack success rates (up to 95.7%). The authors propose AXIS, a defense that aligns harmful‑request representations with the refusal direction and trains the model to refuse outright, achieving superior safety and usability scores on several large language models.
By Zhankai Ye, Yanning Wang, Yukai Jin, Bo Mei, Fangyi Li, Wei Wang, Shangqian Gao, Xin Liu
The paper evaluates the safety of LoopLMs—parameter‑efficient language models that reuse shared parameters across recurrent steps—under jailbreak attacks at various depths. It finds that deeper inference can increase attack success, that safety responses vary across depths, and that attacks transfer between depths, even after standard alignment techniques. To address these gaps, the authors propose SafeBridge, a method that applies depth‑specific control, selective state bridging, and joint safety supervision, which markedly reduces attack success and improves overall utility without increasing over‑refusal.
By Yi Wang, Xiuyuan Qi, Dongqi Han, Dongsheng Li, Wenjie Wang
MultiWorldBench is a diagnostic Minecraft benchmark designed to test whether independently controlled views can describe a single shared world. It contains 495 case configurations across seven task suites and ten capabilities, such as independent control, cross-view motion, shared-state synchronization, persistence, structural reasoning, concurrent interaction, and delayed revisit. The benchmark evaluates systems like Solaris, Gamma-World, and MineWorld, revealing that none achieve high scores in state persistence, structural consistency, spatial reasoning, or building-identity preservation, and that Gamma-World and Solaris perform best in different subsets of capabilities.
By Zhangbo Xu, Ruoxi Zhang, Rui Hu, Yisong Wang
The paper investigates whether similarity scores truly reflect shared structure between language models and the brain or across languages. By removing the underlying structure or using a non‑functional instrument, the authors show that cross‑lingual transfer and brain alignment scores can remain high even when the supposed correspondence is destroyed. The study highlights that similarity metrics may be misleading and that careful validation is needed before interpreting them as evidence of shared structure.
By Saman Rahbar
The paper introduces DEX, an MSDF (most‑significant‑digit‑first) accelerator for quantized U‑Net brain‑tumor segmentation. It employs a two‑stage grouped processing element that supports signed INT8 operands and in‑stream bias accumulation, and implements four runtime mechanisms—exact early negative detection, exact sign‑only decision making, calibrated low‑order‑digit skipping, and calibrated pruning—to dynamically reduce computation. In experiments on a BraTS‑trained residual U‑Net, the approach cuts digit cycles by 38.38 % while maintaining a mean Dice score of 80.58 % (vs. 81.20 % for the floating‑point model), and the exact mechanisms alone achieve an 18.79 % cycle reduction without altering the quantized output.
By Yousef Sadegheih, Dorit Merhof, Muhammad Usman
Hybrid Cinematography is a workflow that blends physical camera capture with generative video reshooting to control hallucination risk. It uses an editable 3D shot plan and a proxy of the take to evaluate hallucination risk in real time, allowing filmmakers to adjust plans, shoot guided pickups, or accept hallucinations. The approach is demonstrated via a mobile AR app for on‑set planning and an offline pipeline for existing footage, and a study shows how risk previsualization informs camera decisions and highlights tensions between creative intent and generative hallucination.
Large language models often justify legal decisions by citing statutes or precedents, but this study shows that while models correctly name the governing authority 66.7%-100% of the time, the actual verdict rarely changes when the cited authority is swapped—only 0.0%-21.7% on CaseHOLD, 30.0%-76.7% on ECHR and SCOTUS, and 43.3%-50.0% on ContractNLI. This gap persists across models of different sizes and even a purpose‑built legal‑reasoning model. Additionally, a red‑team evaluation revealed that models are more compliant with hidden adversarial instructions than they are sensitive to verdict changes, indicating that citing legal authority is a weak proxy for genuine reasoning and that verdicts remain vulnerable to manipulation.
The paper investigates efficient learning of survival analysis models when data are subject to outcome‑dependent truncation and right‑censoring. It shows that Cox regression can be trained with first‑order methods when the baseline hazard is known, but joint estimation of the baseline hazard introduces non‑convexity and local minima. Experiments on four datasets demonstrate that accounting for truncation reduces bias, works with neural networks, and yields orders‑of‑magnitude speedups over expectation‑maximization.
We present AuraLuxMuse, a novel system for automated aesthetic stage lighting design that integrates expert knowledge, representation learning, and preference-adaptive modeling. Lighting design in live performance settings requires the seamless translation of musical features into dynamic lighting behaviors.
EchoDiST is an unsupervised framework for estimating myocardial motion in echocardiography. It combines self‑distillation joint learning of anatomical segmentation and motion estimation with a diffusion‑conditioned network, requiring only limited anatomical annotations. The method achieves statistically significant improvements in anatomical alignment, strain assessment, and functional metrics across three datasets, outperforming seven other learning‑based approaches.
The paper introduces Policy Alignment Membership Auditing (PAMA), a framework designed for on‑policy distillation (OPD) to detect whether a prompt was used during training. PAMA leverages the direction of teacher‑guided policy updates, quantified by Teacher Alignment Gain (TAG), along with student drift and uncertainty signals, to assess membership. Experiments on six datasets and three model families show that PAMA outperforms existing methods, achieving AUC scores between 0.791 and 0.941 on the MATH benchmark, a 14.6–20.6% improvement over state‑of‑the‑art baselines.
The Neuro‑Memory Fuzzy Inference System (NeMeFIS) is a hierarchical machine learning framework that models car‑following acceleration and deceleration by integrating five human memory types—procedural, working, episodic, semantic, and declarative—through metaheuristics and statistical validation. Trained on 54 models across arterial, collector, and rural highways, NeMeFIS outperforms traditional statistical and conventional machine learning approaches, revealing that declarative memory dominates braking decisions while procedural memory drives acceleration. The system’s robust performance on heterogeneous and homogeneous datasets supports its use in adaptive, human‑like decision systems for Connected and Autonomous Vehicles, potentially improving traffic safety.
arXiv:2610.08862v1 Announce Type: cross
Abstract: Household robots must accommodate new user instructions while executing ongoing tasks. Existing agents often regenerate or extensively revise the rem...
By Ci Zhang, Enfu Nan, Arman Akbari, Lin Zhao, Li Wang, Chen Wang, Weiwei Chen, Yanzhi Wang, Geng Yuan
arXiv:2610.09718v1 Announce Type: cross
Abstract: Vision-Language-Action (VLA) models acquire broad manipulation capabilities via large-scale pretraining, yet eliciting them through language requires...
By Masatoshi Tateno, Takehiko Ohkawa, Yueh-Hua Wu, Hanlong Li, Tatsuya Matsushima, Yoichi Sato, Kei Ota
The paper introduces a unified evaluation framework for face de-identification (FDeID), addressing inconsistencies in current protocols. It presents UtilFace, a demographically balanced benchmark assembled from four large-scale datasets, and HiFD, a hierarchical metric that simultaneously measures identity suppression, multi-level utility preservation, and image quality. Using this framework, the authors compare adversarial, GAN-based, and diffusion-based methods, revealing trade-offs and failure modes that were previously hidden, and they release the benchmark and toolkit for reproducible research.
By Hui Wei, Hao Yu, Hui Kuurila-Zhang, Guoying Zhao
The paper introduces Moment-GPT, a tuning‑free pipeline for zero‑shot video moment retrieval that leverages frozen multimodal large language models. It corrects and rephrases queries with LLaMA‑3 to reduce language bias, generates candidate spans using a span generator and MiniGPT‑v2, and selects the best spans with VideoChatGPT and a span scorer. The method outperforms existing MLLM‑based and zero‑shot models on datasets such as QVHighlights, ActivityNet‑Captions, and Charades‑STA.
By Yifang Xu, Yunzhuo Sun, Benxiang Zhai, Ming Li, Wenxin Liang, Yang Li, Sidan Du
The study investigates whether large language models (LLMs) that allocate extra computation during inference—termed reasoning models—reduce classic decision biases compared to their non‑reasoning counterparts. Using 30 vignettes covering six cognitive biases and varying token budgets up to 8,192 tokens, the authors find that reasoning models are not less biased, and increased deliberation does not reliably diminish bias magnitude. Only anchoring showed a bias in the human direction, while other biases either remained unchanged or moved further from human patterns, suggesting that test‑time reasoning does not guarantee rationality.
By Obada Kraishan
The study evaluates how well three sentiment‑analysis tools (TextBlob, VADER, Twitter‑roBERTa‑base) and three large language models (Qwen3‑32B, GPT‑OSS‑120B, Llama‑4‑Maverick‑17B) agree with six human raters on 100 tweets. Agreement was measured with Cohen’s and Fleiss’ kappa, revealing only fair inter‑human agreement and higher concordance for binary sentiment labels than for three‑class labels. Twitter‑roBERTa‑base achieved the strongest alignment with humans, especially for negative versus non‑negative sentiment, while the LLMs showed substantial agreement among themselves and moderate to substantial alignment with humans, particularly for positive versus non‑positive classifications.
The findings emphasize that domain‑specific fine‑tuning and human‑centered evaluation are essential for reliable social media sentiment analysis.
By Himarsha R. Jayanetti, Sivakanesan Dhanushkanda, Shuai Hao, Michael L. Nelson, Michele C. Weigle
arXiv:2610.09255v1 Announce Type: new
Abstract: Neural PDE surrogates are trained on numerical solver outputs that contain both physical evolution and solver-specific discretization errors. Because s...
By Ridham Patel