Competence-Gated Pooling of Language Models and Priors for Event Forecasting

  • 类型：arxiv
  • 标识：2609.12101
  • 链接：https://arxiv.org/abs/2609.12101
  • 主分类：llm-infra
  • 形态：method
  • 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：This work introduces a competence gate that estimates source weights from resolved outcomes, shrinks regime-specific estimates toward a global solution, and chronologically recalibrates the pooled forecast, and shows no statistically detectable performance difference from more flexible contextual aggregators.
  • OpenAlex ID：W7212533095
  • OpenAlex DOI：10.48550/arxiv.2609.12101
  • DOI：10.48550/arxiv.2609.12101
  • DOI来源：OpenAlex
  • 开放获取：green
  • 开放获取链接：https://doi.org/10.48550/arxiv.2609.12101
  • OpenAlex更新：2026-09-16
  • 待LLM分类：否
  • 标题中文：Competence-Gated Pooling of Language Models and Priors for Event Forecasting
  • TLDR中文：提出了一个能力门控（competence gate），根据已解决的结果估计领域级的源权重，将不确定的估计向全局权重收缩，并重新校准合并后的预测，为基于已测边际价值的选择性模型使用提供了实用方案。
  • 成熟度：research
  • 场景：事件预测、模型融合
  • 来源文件：
  • /inbox/tom/_candidates/2026-09-15-rag-retrieval-reranking-candidates.json
  • /inbox/tom/_candidates/2026-09-15-agent-rag-longcontext-candidates.json
  • [OpenAlex backfill]
  • [S2 enrich]