Zhen Wang

This work proposes a generational genetic algorithm to coordinate specialized large language model agents that integrate mechanistic arguments, reconsider assumptions, and assess evidence and testability that advances a vision of autonomous science in which AI research teams achieve a capacity for discovery beyond that of individual models.

The Mind2Dialogue framework proposes a psychology-guided simulator that preserves personal characteristics while updating mental states through interaction to generate coherent conversations, and trains models on the Oracle's well-informed responses to assist users without direct access to their mental states at deployment.

UGC-enhanced images can harbor subtle anomalies that existing quality assessment methods completely overlook, but our new framework identifies these issues with unprecedented precision.

PFGS can reduce word error rates by up to 19.3% compared to random selection, highlighting the critical role of phoneme frequency in TTS augmentation for ASR.

Conditional memory can either enhance or hinder scientific reasoning, and knowing when to activate it is crucial for optimal performance.

Achieve high-fidelity, training-free visual generation from coarse inputs by cleverly steering pretrained diffusion models with a noise-adaptive h-transform.

Forget gradients: this new sampler learns complex distributions, even with discrete parameters, by enforcing time-reversibility and comparing forward and backward Markov trajectories.

Forget discrete search and trial-and-error prompting: nabla-Reasoner uses gradient descent in the LLM's latent space to boost reasoning accuracy by 20% while cutting model calls by up to 40%.

Key contribution not extracted.