Proposes a scalable 2D-to-3D data lifting pipeline combining scale-invariant and scale-aware depth estimation, generating ~2 million realistic 3D scenes, boosting spatial understanding.

Xingyu Miao, Haoran Duan, Quanhao Qian et al.

Prune&Comp combines iterative layer pruning with offline magnitude compensation, nearly halving perplexity and retaining 93.19% QA performance on LLaMA-3-8B without training.

Xinrui Chen, Hongxing Zhang, Fanyi Zeng et al.

VERIRAG employs structured analysis with Veritable taxonomy and Toulmin model to detect methodological flaws, improving F1 by at least 19 points.

Shubham Mohole, Hongjun Choi, Shusen Liu et al.

Ultra3D employs VecSet and Part Attention to achieve high-fidelity 3D generation with a 6.7× speed-up, enabling efficient high-resolution modeling.

Yiwen Chen, Zhihao Li, Yikai Wang et al.

Flow Matching (FM) offers efficient, stable generative modeling for biological sequences, molecules, and proteins, outperforming traditional methods in speed and accuracy.

Zihao Li, Zhichen Zeng, Xiao Lin et al.

Hierarchical Memory (H-MEM) organizes long-term memory in multi-level structure, improving retrieval efficiency by 5x and outperforming five baselines in long-term dialogue tasks.

Haoran Sun, Shaoning Zeng

WGRAMMAR precompiles static structures and uses FSMs, achieving 250× speedup in constrained decoding.

Ran Wang, Xiaoxuan Liu, Hao Ren et al.

Evaluated MLLMs using GRASP and IntPhys 2 datasets, found poor integration of visual and linguistic information.

Mohamad Ballout, Serwan Jassim, Elia Bruni

Introduces a deterministic, modular framework with verifiable guard code for policy enforcement in LLM agent workflows, validated on τ-bench Airlines with 75% success.

Naama Zwerdling, David Boaz, Ella Rabinovich et al.

Combining EAST+CRNN with image upscaling reduces CER from 0.65 to 0.48 in egocentric scene text recognition under challenging conditions.

Joseph De Mathia, Carlos Francisco Moreno-García

Proposes a four-stage framework for GPAI evaluation emphasizing validity and reproducibility, validated on GPT-4 with 15% performance gain.

Patricia Paskov, Michael J. Byun, Kevin Wei et al.

TorchAO is a PyTorch-native framework supporting FP8, QAT, PTQ, and sparsity, enabling end-to-end model optimization from training to deployment.

Andrew Or, Apurva Jain, Daniel Vega-Myhre et al.

LAN2CB leverages large language models to convert natural language instructions into executable robot control code, reducing manual effort and enhancing flexibility.

Zhehui Huang, Guangyao Shi, Yuwei Wu et al.

AutoMCP studies 116 MCP servers, raising tool-generation success from 76% to 94.2% while cutting tool counts by one-third.

Meriem Mastouri, Emna Ksontini, Amine Barrak et al.

GUI-G² models GUI elements as Gaussian distributions, replacing sparse binary rewards with dense continuous signals, achieving 24.7% accuracy gain on ScreenSpot-Pro.

Fei Tang, Zhangxuan Gu, Zhengxi Lu et al.

Inter-LLM integrates hierarchical scene graph reasoning with real-time motion cost feedback, achieving 30% success rate improvement in large-scale multi-object collection tasks.

Ruochu Yang, Yu Zhou, Fumin Zhang et al.

RankMixer employs multi-head token mixing and sparse MoE, scaling parameters by 100x, boosting MFU from 4.5% to 45%, demonstrating superior industrial recommendation performance.

Jie Zhu, Zhifang Fan, Xiaoxie Zhu et al.

Proposed a dense-depth guided LiDAR-Visual Odometry method, achieving superior performance on KITTI dataset.

JunYing Huang, Ao Xu, DongSun Yong et al.

SPAR employs multi-agent architecture with RefChain and query evolution, achieving up to +56% F1 on AutoScholar.

Xiaofeng Shi, Yuduo Li, Qian Kou et al.

This study reveals models transmit behavioral traits via hidden signals in unrelated data, highlighting risks for AI safety and model control.

Alex Cloud, Minh Le, James Chua et al.