Xin Li

Adversarial distillation achieves state-of-the-art few-step visual generation by learning log-density ratios directly via classification, eliminating the need for auxiliary score fitting.

Halting LLM debate consensus-drift to stop groupthink can backfire severely: a probe-gated freeze stopped 29 collapses into wrong answers at the direct cost of sacrificing 108 valid self-corrections.

Multi-teacher distillation fails not because of bad routing, but because high-variance instruction feedback quietly drowns out precision domains like mathematics during shared student updates.

TFA is introduced, a reliability-guided, training-free aggregation method requiring neither place labels nor task-specific weight updates that can recover additional retrieval capability from frozen representations, providing a practical baseline for new environments with scarce place supervision.

Forget painstakingly training a single massive model – HeteroFusion lets you Frankenstein a super-model by intelligently merging the strengths of Llama, Qwen, and Mistral.

Achieve both fidelity and plausibility in 3D reconstruction by unifying feed-forward reconstruction with diffusion-based generation in a single cooperative system.