GLARE: Generative Learning via Adversarial Reward Estimation For Social Dynamics Forecasting
The Meeting Dynamic Forecasting Benchmark is introduced and GLARE is presented, an adaptation of adversarial imitation learning to conditional language generation and outperforming SFT and SPIN while remaining below the observed human continuation.
Tenghao Huang, Zhao-Xuan Tan, Mu-Hao Chen et al.
· 0 citations