This work fine-tunes GPT-4.1 and Qwen2.5-Coder for text-to-CAD sequence generation, achieving significant improvements in automation and structural accuracy.

Prashant Govindarajan, Davide Baldelli, Jay Pathak et al.

This study evaluates whether non-equivariant CNNs trained with rotation augmentation can learn equivariance, showing comparable performance to equivariant models in molecule tasks.

Ewa M. Nowara, Joshua Rackers, Patricia Suriana et al.

Overtone employs cyclic patch size modulation via CSM and CKM modules, reducing spectral harmonic errors by up to 40%, enabling flexible, compute-adaptive PDE surrogates.

Payel Mukhopadhyay, Michael McCabe, Ruben Ohana et al.

RCT study shows 2025 AI tools unexpectedly increased developer task time by 19%, contradicting expectations.

Joel Becker, Nate Rush, Elizabeth Barnes et al.

Proposes CoPart, a multi-part latent diffusion framework for 3D generation, trained on Partverse dataset, achieving fine-grained control and part consistency.

Shaocong Dong, Lihe Ding, Xiao Chen et al.

Introduces MisAttribution Framework and AttriData dataset; trains MisAttributionLLM for multi-task error evaluation with high correlation (0.935).

Zishan Xu, Shuyi Xie, Qingsong Lv et al.

VFMTok leverages frozen vision foundation models with region-adaptive quantization, achieving gFID 1.36 on ImageNet, 3x faster convergence, and high-fidelity class-conditional synthesis.

Anlin Zheng, Xin Wen, Xuanyang Zhang et al.

Linear optical diffractive processors enable parallel approximation of millions of nonlinear functions, validated experimentally with 35 functions in a single shot.

Md Sadman Sakib Rahman, Yuhang Li, Xilin Yang et al.

Controlled study compares autonomous coding agent OpenHands with GitHub Copilot, showing 35% productivity increase and 50% effort reduction in real tasks.

Valerie Chen, Ameet Talwalkar, Robert Brennan et al.

RegGS addresses sparse view reconstruction with 3D Gaussian registration, excelling on the RE10K dataset.

Chong Cheng, Yu Hu, Sicheng Yu et al.

Proposes STTM, a training-free spatio-temporal token merging method, achieving 2× speed-up with 0.5% accuracy drop at 50% token budget.

Jeongseok Hyun, Sukjun Hwang, Su Ho Han et al.

Introduces LongVideo-Reason dataset and LongVILA-R1 framework, combining multimodal RL for long video reasoning, achieving 65.1% accuracy.

Yukang Chen, Wei Huang, Baifeng Shi et al.

MIRIX is a multi-agent, multimodal memory system that improves LLM accuracy by 35% and reduces storage by 99.9%.

Yu Wang, Xi Chen

PUMA combines layer pruning self-distillation with modality-adaptive contrastive loss, reducing parameters by 50% and doubling inference speed with minimal performance loss.

Yibo Lyu, Rui Shao, Gongwei Chen et al.

Proposes IPSsim, integrating content similarity to improve bias correction, boosting NDCG by ~3%.

Zeyan Liang, Graham McDonald, Iadh Ounis

IPPRO introduces a scale-invariant pruning criterion based on projective geometry, outperforming magnitude-based methods in neural network compression.

Jaeheun Jung, Jaehyuk Lee, Yeajin Lee et al.

SynthEHR-Eviction combines LLM-based synthetic data and prompt optimization to detect eviction statuses, achieving 88.8% Macro-F1.

Zonghai Yao, Youxia Zhao, Avijit Mitra et al.

PruneRec combines three-stage structured pruning and distillation, retaining 88% performance with over 95% of non-embedding parameters removed.

Shanle Zheng, Keqin Bao, Jizhi Zhang et al.

FlexGaussian combines mixed-precision quantization and attribute-discriminative pruning for training-free 3D Gaussian compression, achieving up to 96.4% reduction with <1dB PSNR loss.

Boyuan Tian, Qizhe Gao, Siran Xianyu et al.

Expanded MorphScore for 70 languages shows negligible correlation between morphological alignment and language model performance.

Catherine Arnett, Marisa Hudspeth, Brendan O'Connor