A3 mechanism dynamically determines execution horizon via self-speculative prefix verification, improving success rate and inference efficiency across tasks.

Feng Chen, Xianghui Wang, Yuxuan Chen et al.

LaSCD method reduces visual hallucinations using Laplacian energy, improving accuracy.

Fanpu Cao, Xin Zou, Xuming Hu et al.

PaperGym enhances LLM research ideation through cross-domain seed retrieval and method synthesis.

Yunju Choi, Min Song

AutoLLMResearch employs multi-fidelity environments and long-horizon MDPs to automate high-cost LLM experiment configuration, guided by low-cost trials.

Taicheng Guo, Nitesh V. Chawla, Olaf Wiest et al.

Proposed Pace-and-Path Correction method improves VLA model success rate by 28.8% in dynamic environments.

Yanyan Zhang, Chaoda Song, Vikash Singh et al.

DOLORES achieves deep reasoning via structured meta-cognition, improving by 24.8% on average.

Dean Light, Michael Theologitis, Kshitish Ghate et al.

Proposed Abstraction Spaces model to analyze abstraction gaps between users and systems.

Bryan Min, Sangho Suh, Jim Hollan et al.

SOMA enables efficient multi-turn LLM serving via small models, reducing costs and latency while maintaining high response quality.

Xueqi Cheng, Qiong Wu, Zhengyi Zhou et al.

Proposes a block-based double decoder architecture combining full supervision training with inference efficiency, reducing memory and computation by over 66%.

Asher Labovich, Benjamin Bradley, Vanessa Alexander et al.

Proposes an adaptive correction scheduling method that balances constraint projection timing and sampling fidelity, improving trajectory consistency with 75% fewer corrections.

Noah Trupin, Yexiang Xue

QUIVER framework quantifies perturbation propagation and bifurcation in heterogeneous LLM pipelines using sensitivity matrices and trajectory divergence metrics.

Prashanti Nilayam, Sankalp Nayak

V4FinBench excels in corporate bankruptcy prediction using TabPFN and Llama-3-8B models.

Marcin Kostrzewa, Sebastian Tomczak, Roman Furman et al.

Proves that the smallest weight norm of a neural network equals the Kolmogorov complexity of its output, explaining weight decay's effectiveness.

Tiberiu Musat

CADBench evaluates 11 models on 18,000 samples across 6 metrics, revealing performance gaps in multimodal 3D CAD generation.

Anna C. Doris, Jacob Thomas Sony, Ghadi Nehme et al.

PathISE leverages transformer-based MIL to estimate path informativeness, boosting KGQA accuracy and efficiency.

Shengxiang Gao, Chao Lei, Jey Han Lau et al.

Proposed ERD framework addresses representation degradation in diffusion model training, significantly improving efficiency and generation quality.

Zhipeng Yao, Dazhou Li, Zitong Zhang et al.

Machine Learning Techniques for parameter inference in astrophysics using simulation-based inference (SBI).

Leander Thiele

CAQ-ZO method eliminates query-time residuals in low-precision evaluations, enhancing NF4 model performance.

Yao Shu, Zilin Zhu

The study explores when teachers should control AI-generated math visuals, finding post-generation control best for accuracy and predictability.

Zhengxu Li, Junling Wang, April Yi Wang

BCJR-QAT uses BCJR algorithm for differentiable quantization training, reducing PPL by 0.084 on WikiText-2.

Venugopalan Iyengar