🏢 Gaoling School of Artificial Intelligence, Renmin University of China

StreamingDialogue revolutionizes prolonged dialogue learning by compressing long contexts into conversational attention sinks, minimizing information loss and achieving a 4x speedup with 18x less memo…

COPPER enhances LLM-based multi-agent collaboration via a self-reflection mechanism and counterfactual PPO. It improves reflection quality, alleviates credit assignment issues, and shows strong perfo…

P2C2Net: A physics-encoded neural network efficiently predicts complex spatiotemporal dynamics using coarse grids and limited training data, achieving state-of-the-art results.

Over-parameterized Distillation Framework (OPDF) boosts knowledge distillation by efficiently over-parameterizing student models via tensor decomposition, significantly improving performance without i…

Autoregressively trained transformers surprisingly learn algorithms during pretraining, enabling in-context learning; this paper reveals when and why this ‘mesa-optimization’ happens.

MoICE, a novel plug-in, significantly enhances LLMs’ long context awareness by dynamically routing attention using multiple RoPE angles, achieving superior performance with high inference efficiency.

This paper establishes tight lower bounds for the uniform stability of gradient-based bilevel programming algorithms used for hyperparameter optimization, resolving a key open problem regarding the ti…

FineCLIP boosts fine-grained image understanding by combining real-time self-distillation with semantically rich regional contrastive learning, significantly outperforming existing methods.

Researchers extended large language models’ context windows by training-free methods via analyzing and manipulating positional vectors, improving long-text processing.

High-degree representations significantly boost the expressiveness of E(3)-equivariant GNNs, overcoming limitations of lower-degree models on symmetric structures, as demonstrated theoretically and em…