SkillForge continuously verifies and evolves skills, improving RL agent performance; achieves +6.3% success rate over SkillRL in ALFWorld.

Shidong Yang, Ziyu Ma, Tongwen Huang et al.

Proposes a contact-rich robot learning method using explicit physical interaction recognition, enabling robust one-shot task reproduction.

A. H. G. Overbeek, H. van der Kooij, M. Vlutters

GaussianWAM uses 3D Gaussian fields to enhance WAM's geometric and semantic representations, boosting success rates from 52.05% to 71.29%.

Zijian Zhang, Yuqing Jiang, Weitao Zhou et al.

OPDVR combines OPD and RLVR, achieving superior performance on six reasoning benchmarks.

Wenze Lin, Jiale Zhao, Xitai Jiang et al.

RED network employs recursive encoder and decoder to separate unknown number of speakers with variable microphones, achieving state-of-the-art results.

Fulin Wu, Zhong-Qiu Wang

Quantization effects on Bangla language: GPTQ-Int8 and GPTQ-Q8 maintain high accuracy on Bangla benchmarks.

Ismail Hossain, Nafi Ullah Shafin, Mohammad Abdullah Al Mumin

Proposes GVLA, integrating multi-gripper encoding and adapter routing, trained on MiGA dataset with 103,000 demonstrations, achieving 7.62% improvement over baselines.

Hanyi Zhang, Zihong Luo, Tianyu Li et al.

Develops a semiparametric efficiency framework for Verma constraints via graphical fixing, characterizing tangent space orthogonal complements for single constraints.

Razieh Nabi, Anna Guo, Lin Liu

X-MULTI leverages pretrained VLM to improve factor disentanglement in image synthesis, boosting novel combination accuracy by 12%.

Sonali Godavarthy, Matthias Neuwirth-Trapp, Tim-Felix Faasch et al.

MHA-CSP uses Mahalanobis distance for complex state propagation, with only 119K parameters.

Xiaohe Li

VideoHarness-RSI recursively optimizes context construction for frozen vision-language models to improve long-video understanding.

Guoyang Xu, Hao Chen

Proposes a trajectory cost-utility alignment framework to analyze LLM agent resource consumption and task contribution.

Dan Liu, Jian Li

ExMesh++ employs topology-adaptive explicit mesh reconstruction and UV-PBR decomposition, achieving high-accuracy geometry and relightable assets from multi-view images.

Chuanjin Fan, Lifan Wu, Wenjie Chang et al.

PhysicsBench provides a unified benchmark for generative and predictive models in engineering design, evaluating geometric and physical accuracy across multiple tasks and scales.

Sang Won Lee, Hyogu Jeong, Namwoo Kang

TAGR introduces a temporally adaptive generative recommender with LSID, IAG, and IOPO, boosting live-stream ad engagement by 8.5%.

Wencai Ye, Guangyi Liu, Chaoyi Wang et al.

AgentSpec enhances LLM agent inference efficiency via structure-isolated drafting and redundancy-aware budget allocation.

Xin Wang, Ziming Miao, Yi Zhu et al.

Luce uses Gaussian clouds for single-image to 3D asset generation, improving FID by 28%.

Mayank Singh, Michele Stoppa, Alvise Memo et al.

GlanceWAM achieves 72.2% success by asynchronously imagining future frames in latent space, breaking speed-success trade-off.

Linhan Wang, Zijian An, Mingyuan Zhang et al.

Wontopos Tablet 2 achieves multilingual and multimodal retrieval without lexical matching, scoring 95.7% on LongMemEval-S.

Sunwoo Kim

This study systematically analyzes how base model priors, reward density, and prompt diversity influence RL post-training success, validated via controlled experiments.

Donovan Clay, Saket Gollapudi, Sankar Harilal et al.