Spectral Rewiring for Exploration, Purification, and Model Merging

  • 类型:arxiv
  • 标识:2607.03065
  • 链接:https://arxiv.org/abs/2607.03065
  • 主分类:engineering
  • 形态:application
  • 被引:0
  • 被引来源:Semantic Scholar + OpenAlex
  • S2被引:0
  • OpenAlex被引:0
  • 影响力被引:0
  • TLDR:Subspace-Aligned Rewiring (SAR) shows that extracting reasoning-effective updates from parameter geometry can serve as a training-free mechanism to improve reasoning and multi-domain performance.
  • OpenAlex ID:W7167578326
  • OpenAlex DOI:10.48550/arxiv.2607.03065
  • DOI:10.48550/arxiv.2607.03065
  • DOI来源:OpenAlex
  • 开放获取:green
  • 开放获取链接:https://doi.org/10.48550/arxiv.2607.03065
  • OpenAlex更新:2026-07-19
  • 待LLM分类:否
  • 标题中文:面向探索、纯化与模型合并的谱重连
  • TLDR中文:子空间对齐重连(SAR)表明,从参数几何中提取对推理有效的更新,可作为一种无需训练的机制来提升推理与多领域性能。
  • 来源文件
  • /inbox/tom/_candidates/2026-07-17-agent-rag-longcontext-candidates.json
  • [S2 enrich]
  • [OpenAlex backfill]