注册并分享邀请链接,可获得视频播放与邀请奖励。

与「pasta」相关的搜索结果

pasta 贴吧
一个关键词就是一个贴吧,路径全站唯一。
创建贴吧
用户
未找到
包含 pasta 的内容
[RAG 论文分享] VikingRAG:匹配 SOTA 准确率、Token 成本降到 5%–32% 现有问题:RAG 高准确率依赖结构上下文与多轮交互,是 token 开销的主要来源 企业问答、法律、财报等场景的语料都是章、节、段落组成的结构化文档,结构本身就是检索线索,它指示事实归属与局部和全局的关联。 现有 RAG 方法陷入两难:不用结构(朴素向量 RAG、图 RAG、SQL-RAG)丢失导航线索,准确率低;用结构(MoDora、BookRAG、DeepRead)准确率高,但 DeepRead 要把候选文档的完整目录塞进 prompt,开销随目录规模线性增长,加上多轮交互历史不断累积,token 成本巨大。 论文地址 # 三个核心设计 1. 层次化语义存储:把结构从 prompt 搬进可查询的外部状态。 文档分块后保留所属结构节点,自底向上生成节点摘要,目录、块、摘要全部物化为 URI 可寻址对象(如 viking://Pasta/Carbonara/),祖先-后代关系编码为 URI 前缀。系统向代理暴露 Search / List / Grep / Read 四个工具,语义与结构路径共享同一 URI 空间。效果:结构 token 与实际访问的目录片段成正比,而非与完整目录成正比——直接消解 DeepRead 的线性开销。 2. 证据缺口驱动的多轮检索。 Agent 每轮判断证据是否充分,不充分则继续调用工具(轮数预算 B=15),充分即作答。设计哲学是粗定位与细验证分离:Search 锚点 → Read 查看后发现缺口 → List 相邻块 → Grep 精确命中 → Read 验证,每步把搜索空间收窄到相关子树内。 3. 经验边 + 自适应升级——让相似查询不必重复探索。 经验边从历史检索轨迹中把“Search 命中的 URI”连向“真正支撑答案的 URI”,边上存历史问题嵌入做查询时过滤;新查询沿边做条件化多跳扩展,复用路径而非重新探索(VikingRAG-E)。经验积累足够后,多数查询一轮检索即可作答——用约束感知的充分性判断器先验证证据是否支撑答案的关键约束,验证不过关才升级为完整多轮代理检索(VikingRAG-E+)。 # 实验结果:token 降至 SOTA 的零头 6 个真实结构化文档数据集(从 0.24M 词元的课程大纲到 8.78M 词元的财报),8 个基线,骨干 LLM 为 DeepSeek-V4-Pro,并在 GPT-5.5、Seed-2.0、GLM-4.7 上验证稳健性。主要结论: · 准确率与所有基线持平或更高(DeepRead 通常是最强基线); · 基础版 VikingRAG 仅消耗 SOTA 方法的 11.6%–51.9% token,完整版 VikingRAG-E+ 降至 5.1%–32.5%,延迟同样显著更低; · 逐层消融:经验边再省 12%–33% token,自适应升级再省 19%–50%; · 可扩展性:LightRAG、HippoRAG-2 在最大数据集上 24 小时内无法完成摄入,BookRAG 在多数数据集上超时;而 VikingRAG 在文档数从“仅相关文档”增至 991 篇时性能基本稳定——因为它是按需定位,不随语料规模膨胀; · 存储方面:插入延迟与 DeepRead/MoDora 相当、远快于图方法;代价是摄入期 token 更高(为每个索引对象生成预览);文档删除零 LLM 成本。
显示更多
Pastacıdan Kylie Jenner pastası istedim, yaptığı pastanın bana attığı görsel bu. Benzemediği için reddettim ama 'iade yok teslim almak zorundasın' diye zorluyor. Benziyor mu gerçekten, ben mi haksızım sizce?
显示更多
0
1.6K
49.6K
798
转发到社区
Classic Italian 🍅🍅 finger licking delicious Lasagna recipe. You can watch the full video now; perhaps you’ll find it inspiring and try making this absolutely magnificent lasagna recipe at home yourself #lasagna# #cooking# #pasta# #italianfood# #traditionalfood#
显示更多
0
10
1.7K
49
转发到社区
Let’s make Nashville mozzarella sticks!! #pizza# #pasta# #italianfood# #dinnertime# #calgary# #entrepreneur# #viral#
0
17
472
24
转发到社区
A cozy weekday dinner – homemade tomato pasta with fresh basil and a glass of iced lemon tea. Perfect fuel to wrap up the workday!
Elon Musk has a relatively low survival threshold and can live on the bare minimum. He once tried living on $1 a day, and for him, it was an important psychological and philosophical experiment. “My threshold for existing is pretty low. I mean, I figured I could be in some dainty apartment with my computer and be okay and not starve. In fact, when I first came to North America, and just to sort of see what it takes to live, I'd try to live on $1 a day, which I was able to do. I went more for the hot dogs. Hot dogs and oranges. But you do get really tired of hot dogs and oranges after a while. And you can also like, you know, pasta and green pepper and a big thing of sauce, and that can go pretty far too. So I was like, oh, okay, you know, if I can live for a dollar a day, then at least from a food cost standpoint, well, it's pretty easy to earn like $30 in a month, you know?”
显示更多
0
11
180
13
转发到社区
@Superpowamario No. The Never-Ending Pasta Pass is only for use by the Passholder whose name is printed on the Pass. Passes are personalized and non-transferable. Passholders must present a valid photo I.D. along with the Pass at the time of ordering.
显示更多
0
1.3K
9.8K
763
转发到社区
No shoes = No problems 💁🏼‍♀️ Anyone else like pasta with a side of toes?
0
79
1.4K
33
转发到社区