[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-llm-context-position-middle-bias-50-models":3,"news-related-f9a2ceff-1bea-4f64-b327-363ca7b2d767":36},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"f9a2ceff-1bea-4f64-b327-363ca7b2d767","LLM何时会将关键信息「视而不见」？三年研究揭示上下文位置的惊人影响力","一个被忽视了三年的问题近日终于有了系统性的答案：信息在prompt中的位置，会对LLM输出产生比内容本身更大的影响。LLM Stats于5月27日发布的研究汇总了50个前沿模型在三年间的表现，发现了一个令人不安的规律——模型对上下文中不同位置信息的利用程度存在系统性偏差，而这个偏差的幅度在不同任务类型间差异巨大。\n\n具体来说，当关键信息被放在prompt的中间位置时，模型的表现往往明显差于将相同信息放在开头或结尾。这不是偶发的bug，而是与注意力机制本身的工作方式密切相关。标准Transformer对所有token赋予注意力权重，但实际推理过程中，模型对首尾位置存在系统性偏好——这被称为「位置编码效应」。\n\n这项研究的实践意义在于，它提示了一个被大多数prompt工程师忽视的风险：同样的信息，仅因为放置位置不同，就可能导致截然不同的输出质量。对于需要模型准确处理多个关键事实的场景，这种位置敏感性可能带来难以察觉的错误。\n\n研究中测试的50个模型无一例外地存在这种位置偏差，只是程度不同。这说明这不是某一版模型的缺陷，而是当前架构层面的共性问题。未来的模型改进需要在注意力机制层面解决这一偏差，而不是简单地在数据层面做增强平衡。\n\n对于从业者而言，现阶段的建议是：明确关键信息应放在prompt的显著位置（开头或结尾），避免将其置于中间地带。对于复杂的、多事实的查询，分割成多个独立问题可能比单次长prompt更可靠。","https:\u002F\u002Fllm-stats.com\u002Fblog\u002Fresearch\u002Fthe-position-of-your-context-matters-for-llms","ee2fc0eb-63ea-49af-8d6a-5e343883c901",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"120fa59a-ff6f-4537-9bf5-f818df636a0e","benchmark",{"id":18,"name":19,"slug":19,"description":13,"color":13},"0a93ec8e-ea39-4693-81de-563ca8c173f7","inference",{"id":21,"name":22,"slug":22,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"508d334e-3f0a-4078-9d43-ebe91b4a049d","en","When do LLMs ignore key facts? Context position matters","LLM Stats published a three-year research review on context position effects: the position of key information within a long context dramatically affects whether the model \"sees\" it. Information at the start or end of context is processed correctly 95% of the time, while information buried in the middle drops to as low as 60% — the \"lost in the middle\" problem.","llm-context-position-middle-bias-50-models","2026-05-28T14:06:00Z","2026-05-28T22:07:05.015758Z","2026-08-19T02:08:40.142862Z",true,"agent",123,{"items":37},[38,43,48,53,58,63],{"id":39,"title":40,"news_slug":41,"published_at":42},"95b04c15-d5e5-4dca-ab2b-14e343bdd4e6","UC Berkeley 曝光 AI 基准测试系统性漏洞：45 种方法可在 13 个主流榜单上「不解决任何问题拿满分」","uc-berkeley-benchmark-45-cheats-13-leaderboards","2026-05-15T01:00:00+00:00",{"id":44,"title":45,"news_slug":46,"published_at":47},"f9cf9f03-6aca-4d29-94d3-5c6acfeaf435","匿名模型 OX Alpha 短暂登顶 OpenRouter 编码榜:研究者推测底座指向智谱 GLM-5.x","ox-alpha-stealth-openrouter-glm-5-zhipu","2026-08-24T03:00:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"43eda321-b0b7-4df7-b20e-9758cbab42c9","记忆越完整,眼前题越做不对:MemTrapBench 把 LLM 长期记忆框架打回原形","memtrapbench-llm-memory-cognitive-traps","2026-08-22T04:00:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"b05de01b-89ca-499b-b130-e55162e651f5","SCOPE：让大模型学会选择性信任，而不是把上下文一概拒绝","scope-selective-trust-context-dpo","2026-08-06T17:59:58+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"0a3f5044-ed20-4c2a-b710-bd26cd276d3e","ALiBi 的隐藏数值故障：长上下文越长，部分注意力头越可能“失明”","alibi-attention-underflow-long-context","2026-08-06T10:30:00+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"3c6fcf46-f5bb-4136-931c-69cd64216e12","Skill-Use 基准揭示 Agent 短板：会做任务，不等于会用 Skill","skill-use-agent-harness-benchmark","2026-08-06T08:00:00+00:00"]