HallusionBench is a diagnostic benchmark for entangled language hallucination and visual illusion in large vision-language models, evaluating their reasoning robustness.

Tianrui Guan, Fuxiao Liu, Xiyang Wu et al.

Graph neural networks (GNNs) and graph transformers enable holistic modeling of clinical data, enhancing multimodal integration and interpretability.

Ruth Johnson, Michelle M. Li, Ayush Noori et al.

Introduces 'information value', a measure based on neural language models quantifying utterance predictability via multi-dimensional distances, outperforming token surprisal.

Mario Giulianelli, Sarenne Wallbridge, Raquel Fernández

SALMONN integrates Whisper and BEATs encoders with LoRA and Q-Former, enabling multi-task audio understanding and cross-modal emergent abilities.

Changli Tang, Wenyi Yu, Guangzhi Sun et al.

Cutie adds object-level memory reading to VOS, reaching 64.0 J&F on MOSE versus 56.3 for XMem.

Ho Kei Cheng, Seoung Wug Oh, Brian Price et al.

Proposes Luminate, a structured generation framework leveraging LLMs for systematic design space exploration, enhancing human-AI creative collaboration.

Sangho Suh, Meng Chen, Bryan Min et al.

AgentTuning enhances LLMs' generalized agent abilities; AgentLM-70B matches GPT-3.5-turbo on unseen tasks.

Aohan Zeng, Mingdao Liu, Rui Lu et al.

This study analyzes multi-head attention's optimization and generalization, demonstrating theoretical guarantees for training convergence and error bounds.

Puneesh Deora, Rouzbeh Ghaderi, Hossein Taheri et al.

Safety-Gymnasium environment suite and SafePO library enable comprehensive safe RL research with 16 algorithms across diverse tasks.

Jiaming Ji, Borong Zhang, Jiayi Zhou et al.

The CAT framework employs environment augmentation and probabilistic decomposition for efficient closed-loop adversarial training, significantly improving autonomous driving safety.

Linrui Zhang, Zhenghao Peng, Quanyi Li et al.

Empirical analysis of GPT-4’s iterative prompting on graph coloring; reveals limited self-critique and verification abilities, external verifier boosts success to 40%.

Kaya Stechly, Matthew Marquez, Subbarao Kambhampati

This study uses interpretability methods to analyze gender bias in instruction-tuned models, proposing a few-shot bias mitigation approach that significantly improves translation fairness.

Giuseppe Attanasio, Flor Miriam Plaza-del-Arco, Debora Nozza et al.

SPEED accelerates Transformer decoding via speculative execution, significantly reducing latency.

Coleman Hooper, Sehoon Kim, Hiva Mohammadzadeh et al.

DynamiCrafter combines video-diffusion motion priors with dual image conditioning, but the supplied text reports no numerical metrics.

Jinbo Xing, Menghan Xia, Yong Zhang et al.

Proposed PoDS architecture improves POD-Q scores on Cityscapes-OOD and BDD100K-OOD datasets.

Rohit Mohan, Kiran Kumaraswamy, Juana Valeria Hurtado et al.

Proposes Dual Bank Normalization (DBNORM) with DualIS and DualDIS to mitigate hubness, boosting cross-modal retrieval performance.

Yimu Wang, Xiangru Jian, Bo Xue

DIP-3D integrates sparse stereo triangulation and pose estimation to locate diver-pointed objects in 3D, improving multi-object discrimination in underwater scenes.

Chelsey Edge, Demetrious Kutzke, Megdalia Bromhal et al.

Self-RAG employs self-reflection to dynamically retrieve, generate, and critique, significantly improving factual accuracy and controllability.

Akari Asai, Zeqiu Wu, Yizhong Wang et al.

Proposes a novel k-NN based non-parametric conditional independence test for mixed continuous and categorical variables, improving robustness and accuracy.

Oana-Iuliana Popescu, Andreas Gerhardus, Jakob Runge

ReMax, based on REINFORCE, eliminates the value model, reduces GPU memory and training time, achieving SOTA for 7B models.

Ziniu Li, Tian Xu, Yushun Zhang et al.