Zhaoliang Chen

Iterative refinement of latent states using a small recurrent reasoner can drastically enhance reasoning performance in frozen LLMs, outperforming traditional methods at a fraction of the compute cost.

Multi-view reasoning can be pretrained and reused, enabling competitive performance without the need for constant retraining on downstream tasks.

Achieve 4.7M ops/sec boundary checks with sub-microsecond latency in human-AI trust delegation by replacing per-operation signing with a single post-quantum signature.