English(EN) Where Does Long-Context Supervision Actually Go? Effective-Context Exposure Balancing

新的EXACT方法提升LLM长上下文理解能力

作者 PulseAugur 编辑部 · [1 个来源] · 2026-05-11 13:23

研究人员开发了一种名为EXACT的新监督目标，以改进语言模型的长上下文适应性。该方法通过为依赖更长有效上下文的目标分配额外权重来解决打包训练中的不匹配问题。在Qwen和LLaMA模型上的实验表明，在NoLiMa和RULER等基准测试中取得了显著改进，特别是在证据位于数千个标记之外时，同时保持了在标准问答和推理任务上的性能。 AI

影响增强了语言模型处理和回忆长文档遥远部分信息的能力。

排序理由该集群包含一篇详细介绍改进语言模型性能新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CL 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍改进语言模型性能新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, model release; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: 111 days old; Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.CL TIER_1 English(EN) · Menglin Yang · 2026-05-11 13:23

长上下文监督到底去向何方？有效上下文暴露平衡

Long-context adaptation is often viewed as window scaling, but this misses a token-level supervision mismatch: in packed training with document masking, each target token's effective context remains short. We introduce EXACT, a supervision-allocation objective that assigns extra …

本文如何被排名

报道来源 [1]

长上下文监督到底去向何方？有效上下文暴露平衡

相关实体

相关话题