Search papers, labs, and topics across Lattice.
Zhongguancun Academy
4
0
8
Automating the translation of economic intuition into verifiable computational experiments is now possible, thanks to a new system that outperforms generic LLMs in generating grounded and novel research ideas.
Achieve up to 3.1x faster LLM-based recommendations by making speculative decoding aware of token positions within items and speculation depth.
Hallucination mitigation in LVLMs doesn't have to come at the cost of general performance: MPD reduces hallucinations by 23.4% while *improving* overall generative capabilities.
Agent-as-a-Judge can outperform LLM-as-a-Judge in complex environments, but still struggles to reliably verify agent behavior, revealing a critical gap in current LLM-based agent evaluation.
