Get Started
Home
Topics
Search
Library
Research questionHow can text generation retain consensus-decoding quality without repeated candidate sampling and utility scoring at inference time?Consensus-based decoding can improve generated text by comparing several sampled candidates under a utility function. Repeating both candidate generation and utility scoring at inference time is costly, while amortizing that work may require gold references or preference-labeled data.
AI
Inference Optimization
LLM Pretraining & Post-training
Machine Learning
Natural Language Processing
Reinforcement Learning
Research Paper
Statistical Machine Learning
Technology
Latest papersRecent research connected to this question, newest first.Consensus Group Relative Policy Optimization for Text GenerationThe source studies machine translation on WMT 2024 and summarization on XSum. Its approach uses a utility function and policy samples without gold references or explicit preference labels, with evidence comparing performance against Minimum Bayes Risk decoding and reference-free baselines.research paper · Sep 4, 2026
Related questions
How can continuous diffusion language models reduce denoising steps without sacrificing text-generation quality?How can unified multimodal models reduce redundant inference computation across understanding and generation without sacrificing quality?How can autoregressive LLM decoding generate multiple tokens in parallel at large batch sizes without sacrificing quality?How can lossy speculative decoding accelerate LLM inference without distorting token distributions or degrading generation quality?