Self-Reflective Generation at Test Time

Large language models (LLMs) increasingly solve complex reasoning tasks via long chain-of-thought, but their forward-only autoregressive generation process is fragile; early token errors can cascade, which creates a clear need for self-reflection mechanisms. However, existing self-reflection either performs revisions over full drafts or learns self-correction via expensive training, both fundamentally reactive and inefficient. To address this, we propose Self-Reflective Generation at Test Time (SRGen), a lightweight test-time framework that reflects before generating at uncertain points. During token generation, SRGen utilizes dynamic entropy thresholding to identify high-uncertainty tokens. For each identified token, it trains a specific corrective vector, which fully exploits the already generated context for a self-reflective generation to correct the token probability distribution. By retrospectively analyzing the partial output, this self-reflection enables more trustworthy decisions, thereby significantly reducing the probability of errors at highly uncertain points. Evaluated on challenging mathematical reasoning benchmarks and a diverse set of LLMs, SRGen can consistently strengthen model reasoning: improvements in single-pass quality also translate into stronger self-consistency voting. The ability to execute complex multi-step reasoning remains a central frontier in advancing large language models (LLMs). LLMs generate step-by-step reasoning traces, often called chain-of-thought (CoT) (Wei et al., 2022). This capability has enabled substantial progress in mathematics, program synthesis, and other domains (Y ao et al., 2023; Plaat et al., 2024). The fidelity of these traces often determines whether the final answer is correct (Paul et al., 2024; Hammoud et al., 2025). Thus, improving the reliability of the reasoning process is critical to realizing the full potential of LLMs.