Multi-Q learning with dual agents optimizes UAV path and connectivity in challenging terrains, achieving over 90% success rate with low communication outages.

Mohammed M. H. Qazzaz, Syed A. R. Zaidi, Desmond C. McLernon et al.

Chameleon uses early fusion to generate and understand images and text, outperforming Llama-2.

Chameleon Team

Proposes Scientific Generative Agent (SGA), integrating LLMs and differentiable simulation for physical scientific discovery, outperforming baselines with significant improvements.

Pingchuan Ma, Tsun-Hsuan Wang, Minghao Guo et al.

STAR benchmark evaluates real-world situated reasoning via hypergraph abstraction and programmatic QA, revealing current models' limitations in dynamic videos.

Bo Wu, Shoubin Yu, Zhenfang Chen et al.

Proposes Bergman kernel-based determinantal point process for unbiased Monte Carlo integration on compact complex manifolds, achieving optimal convergence rate of N^{-1-2/d_{R}}.

Thibaut Lemoine, Rémi Bardenet

COAST is a sampling-based algorithm combining stream motion planning with constrained task planning, achieving an order-of-magnitude speedup in complex robotic tasks.

Brandon Vu, Toki Migimatsu, Jeannette Bohg

Proposes DML-IV, a double machine learning-based nonlinear IV regression method, reducing bias and enabling high-quality policy learning.

Daqian Shao, Ashkan Soleymani, Francesco Quinzan et al.

Proposed an active sampling flow using sensitivity analysis and Bayesian surrogate modeling to enhance analog circuit simulation efficiency.

Reda Chhaibi, Fabrice Gamboa, Christophe Oger et al.

MetaReflection enhances language agents via offline reinforcement learning, improving performance by 4%-16.82%.

Priyanshu Gupta, Shashank Kirtania, Ananya Singha et al.

Proposed Consistency Policy accelerates visuomotor control by 10x via self-consistency distillation, maintaining high success rates.

Aaditya Prasad, Kevin Lin, Jimmy Wu et al.

Proposes a controllable, scalable, human-involved user simulator framework CSHI, significantly improving realism and reducing data leakage in CRS evaluation.

Lixi Zhu, Xiaowen Huang, Jitao Sang

This comprehensive survey classifies machine unlearning into exact and approximate methods, emphasizes verification and security, and outlines future research directions.

Weiqi Wang, Zhiyi Tian, Chenhan Zhang et al.

BoQ uses learnable global queries with cross-attention, outperforming SOTA in visual place recognition with high speed and efficiency.

Amar Ali-Bey, Brahim Chaib-draa, Philippe Giguère

Proposed CoRE system uses large language models as interpreters, unifying natural language, pseudo-code, and flow programming for AI agents.

Shuyuan Xu, Zelong Li, Kai Mei et al.

Proposes Woods-Saxon scalar field model constrained by Bayesian analysis, showing stable attractor for late-time cosmic acceleration.

Sreerag Radhakrishnan, Sarath Nelleri, Navaneeth Poonthottathil

This survey organizes video representation learning around RGB, optical flow, CNNs, RNNs, and GCNs, without reporting unified benchmark scores.

Elham Ravanbakhsh, Yongqing Liang, J. Ramanujam et al.

Proposes a variational regularization using cylindrical shearlets for dynamic imaging, achieving near-optimal sparse approximation and robust reconstruction.

Tatiana A. Bubba, Tommi Heikkilä, Demetrio Labate et al.

Automated model and data card generation using LLMs enhances documentation completeness and accountability.

Jiarui Liu, Wenkai Li, Zhijing Jin et al.

This study shows that fine-tuning LLMs on new knowledge slows learning and increases hallucinations, supporting pretraining as the main knowledge source.

Zorik Gekhman, Gal Yona, Roee Aharoni et al.

Introduces multi-level sketch preconditioning with Nyström low-rank approximation, achieving faster linear system solutions.

Michał Dereziński, Christopher Musco, Jiaming Yang