Self-Improving Language Models with Bidirectional Evolutionary Search 事件

Self-Improving Language Models with Bidirectional Evolutionary Search arXiv:2605.28814v1 Announce Type: new Abstract: Search has been proposed as an effective method for self-improving language models and agentic systems, both for post-training sample generation and for inference. However, widely used methods such as best-of-N sampling and tree search face two fundamental limitations: they are guided by sparse verification signals, and they construct candidates primarily through autoregressive