REFLECT diagnoses error steps via controlled replay and outcome flip verification, achieving top localization accuracy in LLM traces.

Xiaofeng Lin, Yingxu Wang, Tung Sum Thomas Kwok et al.

Proposes ATM: an action-consistency transfer matrix for diagnosing latent world models, enabling over 100× speedup in evaluation.

Jiaheng Chen

RadOT-Eval employs structured evidence optimal transport for radiology report evaluation, achieving a Spearman correlation of 0.715 with total errors.

Weixin Liu, Juming Xiong, Yang Li et al.

SAPO framework optimizes policy by validating skill marginal utility, improving skill quality and performance.

Zhiwei Zhang, Yudi Lin, Nikki Lijing Kuang et al.

Imagine-OPD enhances efficiency and accuracy in visual reasoning through a self-distillation framework.

Yishuo Cai, Jiahui Liu, Yuanxin Liu et al.

PhysAgent employs multi-agent reasoning and visual feedback to automate physics-based 4D scene synthesis, surpassing traditional gradient-based optimization.

Chunji Lv, Jiaxi Ye, Yuchen Jiang et al.

BLUE method enhances efficiency in vision-language-action models for autonomous driving with a lightweight gating mechanism, achieving an 8.9% success rate improvement.

George Ling, Lijin Yang, Hao Yang et al.

Proposes Streaming-Train-248K dataset and Streaming Harness system, enhancing vision-language models for proactive, long-term, real-time streaming understanding.

Dingyu Yao, Shuhuan Gu, Qingyi Si et al.

RLDT, a density transport-based reinforcement learning algorithm, significantly improves reward quality and convergence speed in continuous control tasks.

Boshu Lei, Kostas Daniilidis, Antonio Loquercio

EinSort leverages index sorting to reveal low-rank structures in LLM weights, achieving superior compression with minimal performance loss.

Toshiaki Koike-Akino, Jing Liu, Ye Wang

LOTTERY method leverages reference samples for efficient detection in asymmetric two-sample testing.

Xunye Tian, Zhijian Zhou, Liuhua Peng et al.

GraspFoM leverages SAM3D-based 3D priors for reconstruction-driven robotic grasping, achieving state-of-the-art AP and reconstruction metrics.

Dongli Wu, Xiaobao Wei, Hao Wang et al.

Proposes distributional divergence metrics (e.g., MAUVE) to address gen-PPL's failure in evaluating text quality.

Antonio Franca, Alexander Tong

TrustMargin is a training-free arbitration method combining parametric priors and evidence support, improving answer selection in large language models by 4-6 points F1.

Jingyan Xu, Hong Shi, Yi Shan et al.

STAR-KV employs soft-thresholding for adaptive low-rank KV cache compression, achieving up to 75% compression, 6.9× speedup, and 20× overall reduction.

Priyansh Bhatnagar, Ashkan Moradifirouzabadi, Se-Hyun Yang et al.

CATPO uses tree informativeness scoring and critique-guided repair to improve mathematical reasoning accuracy to 37.5%.

Ayush Singh, Umang Goyal, Ankur Dahiya

EmbodimentSemantic introduces a spatial scene-graph dataset to enhance spatial understanding in robotic vision-language models.

Hassan Jaber, Refinath S N, Luca Cagliero et al.

Test-Time Scaling enhances multimodal foundation models in generation and reasoning, categorized into sampling, feedback, and search strategies.

Cong Wan, Ying He, Zhongzhan Huang et al.

LH-NeF adds hierarchical locality to neural-field tokens, cutting memory 42× and enabling 133× larger batches.

Alonso Urbano, David W. Romero, Max Zimmer et al.

Proposed BAND algorithm for user association in mmWave V2X networks, reducing regret by 40% and improving communication rate by 33.1%.

Weiqi Chi, Manabu Tsukada