Code generation with large language models (LLMs) is highly sensitive to token selection during decoding, particularly at uncertain decision points that influence program logic. While standard strategies such as greedy decoding treat all tokens uniformly, they overlook code-specific uncertainty patterns, leading to suboptimal performance. This paper presents an empirical study revealing that many generation errors stem from token ranking mistakes at high-uncertainty steps, where the correct token is present but not top-ranked.

Motivated by these findings, we propose AdaDec, a lookahead-based uncertainty-guided adaptive decoding framework that integrates a token-level \textit{pause-then-rerank} mechanism driven by token uncertainty. AdaDec learns model-specific uncertainty thresholds and applies a lookahead-based reranking strategy when uncertainty is high. Experiments on HumanEval+, MBPP+, and DevEval benchmarks show that AdaDec improves Pass@1 accuracy by up to 20.9% in absolute terms over greedy decoding, and consistently outperforms state-of-the-art adaptive decoding methods such as AdapT, while reducing computational cost and latency through efficient, selective pausing. Our results highlight the promise of uncertainty-aware adaptive decoding for improving both the reliability and efficiency of LLM-based code generation.