AntifakePrompt leverages prompt tuning on InstructBLIP, boosting deepfake detection accuracy from 71.06% to 92.11% on unseen models.

You-Ming Chang, Chen Yeh, Wei-Chen Chiu et al.

Convex function-based training sharpens output distribution, improving BLEU/ROUGE scores by over 9 points in text generation tasks.

Chenze Shao, Zhengrui Ma, Min Zhang et al.

DejaVu accelerates LLM inference by predicting contextual sparsity, reducing OPT-175B latency by over 2x.

Zichang Liu, Jue Wang, Tri Dao et al.

Zephyr uses dDPO with AI preference data to efficiently train a 7B chat model surpassing Llama2-Chat-70B.

Lewis Tunstall, Edward Beeching, Nathan Lambert et al.

MimicTouch leverages multi-modal tactile demonstrations, low-dimensional representation, non-parametric imitation, and residual RL to improve contact-rich manipulation success to 96%.

Kelin Yu, Yunhai Han, Qixian Wang et al.

DreamCraft3D employs hierarchical generation guided by view-dependent diffusion priors, achieving high-fidelity 3D objects with mutual geometric and texture consistency.

Jingxiang Sun, Bo Zhang, Ruizhi Shao et al.

QMoE framework compresses 1.6 trillion-parameter models to 0.8 bits/parameter, achieving 20× compression with minimal accuracy loss and efficient GPU inference.

Elias Frantar, Dan Alistarh

Proposes MIN-K% PROB, a reference-free method for detecting pretraining data, improving 7.4% AUC on WIKIMIA.

Weijia Shi, Anirudh Ajith, Mengzhou Xia et al.

DP-Prompt combines pretrained LLMs and zero-shot prompting with exponential mechanism to achieve effective local differential privacy in text generation.

Saiteja Utpala, Sara Hooker, Pin Yu Chen

MuSR employs neurosymbolic algorithms to generate complex multi-step reasoning narratives, challenging GPT-4 and similar models.

Zayne Sprague, Xi Ye, Kaj Bostrom et al.

Proposes a multi-conditional diffusion model for text-guided 3D scene synthesis, outperforming state-of-the-art benchmarks.

An Vuong, Minh Nhat Vu, Toan Tien Nguyen et al.

Localizing anomalies in critical infrastructure using model-based drift explanations, focusing on leakages and sensor faults.

Valerie Vaquet, Fabian Hinder, Jonas Vaquet et al.

This paper reviews unsupervised concept drift detection in data streams, comparing detection and localization strategies.

Fabian Hinder, Valerie Vaquet, Barbara Hammer

navlie is a Python package supporting on-manifold state estimation algorithms (EKF, invariant EKF, sigma-point filters) on Lie groups, enabling rapid prototyping.

Charles Champagne Cossette, Mitchell Cohen, Vassili Korotkine et al.

DACOOP-A integrates attention mechanisms and artificial potential fields, boosting multi-robot pursuit success to 57% in simulations.

Zheng Zhang, Dengyu Zhang, Qingrui Zhang et al.

SpecTr leverages optimal transport theory to accelerate autoregressive decoding by 2.13x without quality loss.

Ziteng Sun, Ananda Theertha Suresh, Jae Hun Ro et al.

Zero123++ employs diffusion models with multi-view joint modeling, achieving high-quality, consistent multi-view 3D generation from a single image.

Ruoxi Shi, Hansheng Chen, Zhuoyang Zhang et al.

Wonder3D employs cross-domain diffusion to efficiently generate detailed textured meshes from a single image.

Xiaoxiao Long, Yuan-Chen Guo, Cheng Lin et al.

StenUNet uses a UNet pipeline for pixel-wise coronary stenosis detection, reaching F1 0.5348 on ARCADE.

Hui Lin, Tom Liu, Aggelos Katsaggelos et al.

Proposes a Large Search Model (LSM) that unifies search tasks via prompt-driven LLMs, simplifying architecture and enhancing performance.

Liang Wang, Nan Yang, Xiaolong Huang et al.