3

By integrating budget-mode control into the training process, ThinkDial enables seamless switching between different reasoning regimes, balancing speed and accuracy across task requirements.

MemAgent introduces a multi-conv RL-based memory agent that enables language models to handle extremely long documents, extending from 8K to 3.5M tokens with minimal performance degradation.

Investigates LLMs’ ability to develop spatial cognition from local descriptions, demonstrating successful spatial perception and navigation capabilities in simulated urban environments.

We introduce PowerAttention, a novel sparse attention design that facilitates effective and complete context extension through the theoretical analysis.

We introduce TravelAgent, an LLM-powered travel planning system that generates rational, comprehensive, and personalized itineraries through four modules, demonstrating effectiveness in dynamic scenarios.

We propose a novel conversational AI system that enhances researchers’ literature review processes by providing personalized knowledge management, literature recommendations, and query answering through a unified platform.

We propose a simulation to study language’s influence on collective behavior by having agents engage in free chat within various narrative scenarios, with findings suggesting that greater information exchange promotes more orderly and meaningful emergent behaviors.