Contrastive Decoding (CD) leverages likelihood differences between large and small models to improve open-ended text generation without extra training.

Xiang Lisa Li, Ari Holtzman, Daniel Fried et al.

Constructed 6.5TB DiffusionDB with 14M images and 1.8M prompts; analyzed prompt features, hyperparameters, and errors, revealing risks of misinformation and model bias.

Zijie J. Wang, Evan Montoya, David Munechika et al.

Proposes ID-unaware deepfake detection, reducing implicit identity leakage, improving cross-dataset generalization with Artifact Detection Module.

Shichao Dong, Jin Wang, Renhe Ji et al.

PlanT employs object-level sparse features with Transformer, achieving expert-level driving scores on CARLA Longest6, 5.3× faster inference.

Katrin Renz, Kashyap Chitta, Otniel-Bogdan Mercea et al.

Algorithm Distillation (AD) uses causal sequence models to convert RL algorithms into neural networks, enhancing data efficiency.

Michael Laskin, Luyu Wang, Junhyuk Oh et al.

Introduced SciFact-Open, leveraging model fusion and retrieval pooling on 500K abstracts, with F1 dropping over 15 points compared to closed-domain benchmarks.

David Wadden, Kyle Lo, Bailey Kuehl et al.

Reinforcement learning (e.g., DQN, UCB) enhances speech and NLP tasks, improving adaptability and efficiency.

Baihan Lin

Introduces EMF metric and new dataset to evaluate monocular DVS, showing 1-2dB performance drop without multi-view cues.

Hang Gao, Ruilong Li, Shubham Tulsiani et al.

Proposes EnCodec, a neural audio codec combining end-to-end training, RVQ, and multi-scale discriminators, achieving 40% higher compression with high fidelity in real-time.

Alexandre Défossez, Jade Copet, Gabriel Synnaeve et al.

Using a GPT variant trained on synthetic Othello sequences, the study uncovers emergent nonlinear internal representations of the board state, validated through intervention experiments.

Kenneth Li, Aspen K. Hopkins, David Bau et al.

L5PC-inspired MCC architecture reduces energy consumption by 62%, excels in multimodal audio-visual noise suppression, leveraging context-sensitive two-point neurons.

Ahsan Adeel, Adewale Adetomi, Khubaib Ahmed et al.

Proposed Code4Struct leverages code generation for few-shot event structure prediction, achieving 29.5% absolute F1 gain over SOTA.

Xingyao Wang, Sha Li, Heng Ji

MALI leverages label-guided diffusion maps and optimal transport for manifold alignment, outperforming state-of-the-art methods.

Andres F. Duque, Myriam Lizotte, Guy Wolf et al.

Utilizing large language models for MCQA, MCP method narrows the gap with SOTA across 20 datasets.

Joshua Robinson, Christopher Michael Rytting, David Wingate

NeuPhysics learns 3D geometry and physics parameters from monocular videos, enabling editable dynamic scene reconstruction.

Yi-Ling Qiao, Alexander Gao, Ming C. Lin

This paper introduces a syntactic surprisal measure based on CCG supertags within neural language models, revealing underestimation of human processing difficulty in garden path sentences.

Suhas Arehalli, Brian Dillon, Tal Linzen

Scaling instruction fine-tuning with 1.8K tasks and chain-of-thought enhances model performance, achieving SOTA on multiple benchmarks.

Hyung Won Chung, Le Hou, Shayne Longpre et al.

ReasonFormer enhances performance across 11 datasets through modular reasoning.

Wanjun Zhong, Tingting Ma, Jiahai Wang et al.

Proposes a microservice-based neuromorphic system integration framework to address heterogeneity and communication challenges.

Mattias Nilsson, Olov Schelén, Anders Lindgren et al.

FRU employs local stored logs and importance filtering to enable fast user forgetting in federated recommenders, reducing retraining time by 7x.

Wei Yuan, Hongzhi Yin, Fangzhao Wu et al.