Zhengzhong Tu

From Self-Confidence to Statistical Accuracy: Scaling Human-in-the-Loop Evaluation with Direct and LLM-Assisted Auditing

Harine Choi, Chenchen Kuai, Zihao Li, et al.

The Missing Primitive: Diagnosing and Repairing Mathematical Reasoning in Large Language Models

Shuo Xing, Zilin Dai, Chengyuan Qian, et al.

ViTeX-Bench: Benchmarking High-Fidelity Video Scene Text Editing

Xinghao Chen, Xiangbo Gao, Jiongze Yu, et al.

VEFX-Bench: A Holistic Benchmark for Generic Video Editing and Visual Effects

Xiangbo Gao, Sicong Jiang, Bangya Liu, et al.

NavTrust: Benchmarking Trustworthiness for Embodied Navigation

Huaide Jiang, Yash Chaudhary, Yuping Wang, et al.

NavTrust is a new benchmark that tests how well AI navigation systems handle real-world problems like blurry images, sensor noise, and unclear instructions. The researchers found that current state-of-the-art navigation systems perform much worse when exposed to these corruptions, and they tested four strategies to make these systems more reliable. They also validated their findings on an actual robot to confirm the improvements work in practice.