post-training alignment

large language model (13587) sentiment distribution (3) bias mitigation (500) inference-time alignment (16) best-of-n decoding (3) speculative rejection (1) calibration error (61) model debiasing (58) multi-perspective fusion (2) reward model (276)