无需人工标注，通过神经元激活模式筛选数据，实现LLM高效自蒸馏训练。
arXiv:2607.02460v1 Announce Type: cross Abstract: Post-training large language models (LLMs) without real-world interaction feedback or human-labeled …
无需保留集！新方法SHRED用自蒸馏+logit降级实现LLM高效遗忘，拒绝灾难性性能下降。
arXiv:2605.07482v2 Announce Type: replace-cross Abstract: Machine unlearning for large language models (LLMs) aims to selectively remove memorized con…
自蒸馏让大模型在难题上学会专家推理，摆脱依赖更强模型或采样正确解的局限
arXiv:2602.02405v2 Announce Type: replace-cross Abstract: Improving the reasoning capabilities of large language models (LLMs) typically relies either…
一种新颖的反思式策略自蒸馏方法，让语言模型通过自我反思实现跨领域推理能力提升，无需人工标注。
arXiv:2605.28014v1 Announce Type: cross Abstract: On-policy self-distillation (OPSD) improves the reasoning performance of large language models (LLMs…
通过pass-rate加权自蒸馏，恢复LLM推理的“甜蜜点”，破解GRPO归一化带来的学习偏差。
arXiv:2605.27765v1 Announce Type: cross Abstract: Self-Distillation Policy Optimization (SDPO) provides dense token-level credit assignment for reinfo…
视频推理新突破：结构化自蒸馏方法VISD可显著提升模型理解能力
arXiv:2605.06094v4 Announce Type: replace-cross Abstract: Training VideoLLMs for complex reasoning remains challenging due to sparse sequence level re…
一种针对大模型推理能力的自适应蒸馏方法，根据模型当前水平动态调整教学方向，提升效果与效率。
arXiv:2605.22263v1 Announce Type: cross Abstract: On-policy self-distillation (OPSD) is an emerging LLM post-training paradigm in which the model serv…
无需外部信号，自我蒸馏新范式：能力选择子空间投影让LLM自主提升性能
arXiv:2605.22675v1 Announce Type: new Abstract: Self-distillation bootstraps large language models (LLMs) by training on their own generations. Howeve…
探索统一自蒸馏框架UniSD，为大型语言模型的高效优化与性能提升提供新思路。
arXiv:2605.06597v2 Announce Type: replace Abstract: Self-distillation (SD) offers a promising path for adapting large language models (LLMs) without r…
揭秘自蒸馏为何会损害LLM的数学推理能力，并指出抑制关键探索过程是背后原因。
arXiv:2603.24472v3 Announce Type: replace-cross Abstract: Self-distillation has emerged as an effective post-training paradigm for LLMs, often improvi…
互补自蒸馏如何维护大模型上下文完整性？这项研究提出双模型协作新方案，为LLM安全对齐提供创新思路。
arXiv:2605.20258v1 Announce Type: new Abstract: Contextual Integrity (CI) defines privacy not merely as keeping information hidden, but as governing i…
最新研究：后训练MoE模型通过自蒸馏跳过一半专家，无需从头预训练，显著降低计算量。
arXiv:2605.18643v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) scales language models efficiently through sparse expert activation, and its …
扩散LLM无需外部教师，通过“展开回退”策略自我提升推理效率，开辟模型加速新方向。
arXiv:2605.16941v1 Announce Type: new Abstract: Diffusion Large Language Models (DLLMs) promise fast parallel generation, yet open-source DLLMs still …
多模态大模型新突破，通过自蒸馏策略让AI学会捕捉视觉细节，显著提升细粒度理解能力。
arXiv:2605.18740v1 Announce Type: cross Abstract: Multimodal Large Language Models (MLLMs) still struggle with fine-grained visual understanding, wher…
从长时域智能体学习挑战入手，提出目标后见自蒸馏方法，提升复杂任务表现。
arXiv:2605.17873v1 Announce Type: new Abstract: Training long-horizon LLM agents with reinforcement learning is challenging because sparse outcome rew…
提出轨迹自蒸馏方法，让扩散语言模型用少步就能快速并行生成文本，突破推理速度瓶颈。
arXiv:2602.12262v3 Announce Type: replace-cross Abstract: Diffusion large language models (DLLMs) have emerged as powerful generative models with the …
提出粒度自适应优势重加权方法，用自蒸馏实现LLM Agent的细粒度信用分配，改进策略学习效率。
arXiv:2605.11853v2 Announce Type: replace-cross Abstract: Reinforcement learning has become a widely used post-training approach for LLM agents, where…