← Back to all news

Benchmarks and evaluation

Leaderboards, eval harnesses and ablations — the contested business of deciding which model is actually better.

28,111 stories · RSS feed

arXiv Computer Vision
Oct 22

RoPT: Robust 6DoF Object Pose Tracking in Video

arXiv:2610.28962v1 Announce Type: new Abstract: Accurate 6-DoF object pose tracking is essential for robotic interaction, yet conventional monocular pipelines frequently suffer from catastrophic dive...

By Yongqing Liang, Yuzhong Chen, Sihang Li, Junsuk Kim, Chase Kidder, Hui-ping Wang, Yunzhi Xu, Raihan Haque, Yubiao Zhang, Chen Feng
roboticsbenchmarks
More like this →
arXiv Computer Vision
Oct 22

SAM-DCE: Decoupled Context Enhancement for Prompt-Free Medical Image Segmentation

SAM-DCE is a framework that adapts the Segment Anything Model (SAM) for fully automatic, prompt‑free medical image segmentation. It introduces Spatial Representation Compensation (SRC) to recover fine anatomical details from intermediate Vision Transformer features and Semantic Token Calibration (STC) to enhance the consistency of foreground mask tokens. By refining spatial and semantic representations before using SAM’s original token‑to‑mask mechanism, SAM‑DCE achieves superior performance on the Synapse multi‑organ CT segmentation dataset, with a mean Dice score of 83.86% and an HD95 of 16.04 mm, outperforming several CNN, Transformer, and SAM baselines.

By Yingzhen Hu, Yiheng Zhong, Aijing Yang, Georges El Fakhri, Xiaofeng Liu
llmscomputer-visionbenchmarks
More like this →
arXiv Computer Vision
Oct 22

Towards Explainable Industrial Anomaly Detection via Knowledge-Guided Latent Reasoning

arXiv:2602.09850v3 Announce Type: replace Abstract: Industrial anomaly detection demands precise reasoning over fine-grained defect patterns. However, existing multimodal large language models (MLLMs...

By Peng Chen, Wenqiang Wang, Yunkang Cao, Chengliang Liu, Wei Wang, Mingbo Yang, Li Shen, Wenqi Ren, Xiaochun Cao, Chao Huang
llmsragdiffusionmultimodalbenchmarkssafety
More like this →
arXiv Machine Learning
Oct 22

PerSPL: Personalized Semi-supervised Preference Learning

PerSPL (Personalized Semi-supervised Preference Learning) is a backbone‑agnostic framework that stabilizes cross‑prompt score coordinates using pair‑mean score centering, enabling the conversion of unlabeled responses into user‑specific training signals. It addresses the scarcity of user‑specific preference comparisons by leveraging large pools of unlabeled responses, which are otherwise unused by existing personalized reward models. Across four benchmarks—PersonalLLM, HelpSteer2, Reddit TLDR, and Chatbot Arena—PerSPL achieves the highest mean accuracy for both seen and unseen users, demonstrating its effectiveness in summarization and open‑domain chatbot interaction settings.

By Xiaoan Xu, Tingzhou Wei, Juncheng Dong, Korel Gundem, Ethan X. Fang, Zhengling Qi
nlpreinforcement-learningbenchmarks
More like this →
arXiv Machine Learning
Oct 22

Implicit Feature Reweighting for Label-Free Graph Generalization

SCARCE is a label‑free representation‑learning framework that improves graph‑level out‑of‑distribution generalization. It uses an environment partition of training graphs—derived from metadata or self‑supervised clustering—to warm‑start a variational graph autoencoder and then applies an adversarial confusion mechanism that prevents a discriminator from distinguishing environments while preserving reconstruction fidelity. Theoretical analysis shows that this environment confusion contracts environment‑discriminative directions, effectively reweighting features to reduce reliance on spurious correlations, and experiments on synthetic and DrugOOD datasets demonstrate state‑of‑the‑art OOD performance among label‑free methods.

By Tianjun Yao, Chen Xi, Xinyang Li, Lixian Zhong
ragbenchmarkssafety
More like this →