让大模型生成更多样又连贯的回复，解决风格单一问题
Semantic Flow Regularization: Teaching LLMs to Generate Diverse Yet Coherent Responses
- 用连续语义流正则化提升生成多样性
- 在9种人物风格下显著提升回复多样性和风格一致性
- 无需额外计算开销，适合工业级对话系统
当大语言模型被微调以生成特定人物或语气的回复时，其输出多样性严重受限——我们称之为跨风格坍缩。我们发现这源于交叉熵目标函数在共享表示下的副作用，会抑制多样化延续。为此提出语义流正则化（SFR），一种轻量级辅助目标，通过条件流匹配监督骨干模型未来片段的连续句子编码嵌入。随机流源天然保持多模态性；推理时丢弃流匹配头，部署成本为零。在大规模工业对话数据集（Qwen3-32B，9种人物风格）上，SFR相比SFT显著提升输出多样性、风格保真度和回复质量。在公开的LiveCodeBench-v5（Qwen2.5-Coder-7B-Instruct）上也持续提升pass@k指标，验证了其通用性。受控对比显示，多标记预测是SFR的一个退化特例。
原文摘要 · Abstract (English)
When large language models are fine-tuned to generate persona- or tone-conditioned responses, their output diversity is severely limited--a failure we term Cross-Style Collapse. We trace this collapse to the cross-entropy objective, which under shared representations tends to suppress diverse continuations. We propose Semantic Flow Regularization (SFR), a lightweight auxiliary objective that supervises the backbone with continuous sentence-encoder embeddings of future segments via conditional flow matching. The stochastic flow source preserves multi-modality by construction; the flow-matching head is discarded at inference, adding zero deployment cost. On a large-scale industrial dialogue dataset (Qwen3-32B, 9 personas), SFR improves output diversity, style fidelity, and response quality over SFT. We further validate on the public LiveCodeBench-v5 (Qwen2.5-Coder-7B-Instruct), where SFR consistently improves pass@k, confirming generality beyond stylized dialogue. A controlled comparison on MBPP reveals Multi-Token Prediction to be a degenerate special case of SFR.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。