TypoX Community

Address UQAojvLRQ9kHLpGbCDLAB78Oi_ZsODsITI9H-deS_IJVub_6

Im invited 4 refferal but not get quota
Telegram username @coretod
ton address: UQCRJgXeYW25UpF7ztrsRLN2FcNUnt2I4VYnzOT2FX5jSY-S

Dear TypoCurator Users,

Last night, due to Telegram's limitations, some valid referrals were not promptly converted into Quota, affecting users' ability to continue answering questions.

We have already replenished the Quota for most of the valid referrals. If there are still a few that haven't taken effect, it may be due to group restrictions after surpassing 10,000 users in a Telegram group. For these cases, please promptly go to our community through the "Help" channel in the group to submit your issue ticket, so our support team can follow up in a timely manner.

We greatly appreciate your patience and understanding and deeply apologize for any inconvenience caused. We are committed to continuously improving our service to ensure a smoother user experience.

親愛的TypoCurator用戶們，

昨晚由於Telegram的限制，部分有效推薦未能及時轉化為Quota，影響了用戶繼續答題。

我們已經補發了大部分有效推薦的Quota。若仍有少數未生效的情況，這可能是由於Telegram群組在超過10000用戶後的群組限制所致。針對這種情況，請您及時在群裡通過頻道「Help」直達我們的Community去提交您的問題票據，這樣我們的後台客服會及時跟進。

我們非常感謝大家的耐心與理解，對由此帶來的不便表示深深的歉意。我們將繼續努力改善服務，確保更流暢的用戶體驗。

谢谢您的建议！收到🫡

Energy Quota Recovery Mechanism Update

Dear users,

We are updating our energy recovery mechanism as follows:

Each day, users will receive a fixed recovery of 1 energy point. Users with more than 1 energy point will not receive additional energy. Existing energy points will not be cleared.

Thank you for your understanding and support.

Best regards,
TypoCurator Team

體力刷新機制調整公告

親愛的用戶們，

我們將對體力更新機制進行調整，具體如下：

每日固定回復 1 點體力。 如果用戶的體力超過 1 點，將不會新增體力。 現有的體力不會被清除。

感謝您的理解與支持。

此致
TypoCurator 團隊

You can easily verify if you have the same address on TON and track the status of your withdrawals using Tonviewer: https://tonviewer.com/

您可以使用Tonviewer輕鬆確認TON上的地址是否一致，並追蹤您的提款狀態：https://tonviewer.com/

支持一下，希望项目越做越好

1. Embracing Randomness

Creating an AI product begins with understanding the essential tasks involved. A primary focus is controlling the randomness of a large language model (LLM). Randomness, a double-edged sword, can either enhance the model’s generalization ability or lead to hallucinations. In creative contexts, where precision is less critical, randomness fuels imaginative outputs. Many users are initially captivated by the generative capabilities of large models. However, in domains requiring strict logic, such as factual data and mathematical code, precision is paramount, and hallucinations are more prevalent.

Early optimizations in mathematics and coding involved integrating external tools (e.g., Wolfram Alpha) and coding environments (e.g., Code Interpreter). Subsequent steps included intensive training and workflow enhancements, enabling the model to solve complex mathematical problems and generate practical code of substantial length.

Regarding objective facts, large models can store some information within their parameters, yet predicting which knowledge points are well-integrated is challenging. Excessive emphasis on training accuracy often leads to overfitting, reducing the model’s generalization capability. Training data inevitably has a cutoff date, limiting the model’s ability to predict future events. Consequently, enhancing generative results with external knowledge bases has gained significant traction.

2. Challenges and Prospects of RAG 2.1 What is RAG?

Retrieval Augmented Generation (RAG), proposed as early as 2020, has seen wider application with the recent development of LLM products. Mainstream RAG processes do not modify the model’s parameters but enhance input to improve output quality. Using a simple formula y = f(x) : f is the large model, a parameter-rich (random nonlinear) function; x is the input; y is the output. RAG focuses on optimizing x to enhance y .

RAG’s cost-efficiency and modularity, allowing model interchangeability, are significant advantages. By retrieving relevant content for the model rather than providing full texts, RAG reduces token usage, lowering costs and response times.

2.2 Long Context Models

Recent models supporting long contexts can process lengthy texts, overcoming the early limitation of short inputs. This development has led some to speculate that RAG may become less relevant. However, in needle-in-a-haystack tests (finding specific content within a long text), long-context models perform better than earlier versions but still struggle with mid-text content. Moreover, current tests are single-threaded, and the complexity of multi-threaded tests remains a challenge. Third-party testers have also raised concerns about potential training biases in some models.

While improvements in long-context models are ongoing, their high training and usage costs remain barriers. Compared to inputting entire texts, RAG maintains advantages in cost and efficiency. Nevertheless, RAG must evolve with model advancements, particularly regarding text processing granularity, to avoid over-processing that diminishes overall efficiency. Product development always involves balancing cost, timeliness, and quality.

2.3 The Vision for RAG

Moving beyond early RAG applications reliant on vector storage and retrieval, a systematic approach to RAG framework construction is necessary. Content, retrieval methods, and generators (LLMs) are the fundamental elements of a RAG system.

Optimizing retrievers has been extensively explored due to their convenience. Beyond vector databases, other data structures like key-value, relational, or graph databases should be considered based on data structure. Adjusting retrievers accordingly can further enhance retrieval aggregation.

LLMs can play a crucial role in content preprocessing. Rather than simple slicing, vectorization, and retrieval, using classification models and LLMs for summarization and reconstruction can store content in more LLM-friendly formats. This approach relies more on semantics than vector distances, significantly improving generation quality. The TypoX AI team has validated this method in product implementation.

Optimizing multiple elements simultaneously, especially integrating retrieval with generation model training, represents a significant direction for RAG. This integration can enhance both retrieval and generation quality. Some argue that advancements in LLM capabilities and lower-cost fine-tuning will render RAG obsolete. However, enhancing f (the LLM’s parameters) and x (the input) are complementary strategies. Advanced RAG also involves model fine-tuning for specific knowledge bases, extending beyond input enhancement. RAG’s value lies in superior generative results compared to raw LLMs, better timeliness and efficiency than fine-tuned models, and lower configuration costs.

3. Human and AI Interaction 3.1 Positioning of Large Models

The author opposes the notion that large models alone can meet all user needs. While powerful, large models alone do not suffice to create fully functional agents. Custom-trained or fine-tuned smaller models showcase a team’s capability but may not always enhance user experience. Investing resources in other elements (tools, knowledge bases, workflows) may yield better results. In the Web3 industry, lacking proprietary training data and evaluation standards, the focus should be on developing industry-specific databases and standards, not merely ranking universal (small) models.

Underestimating large models is also a mistake. Early TypoX AI product exploration involved low trust in models, leading to overdeveloped processes. Balancing hard logic with LLM involvement, we achieved an optimal balance, addressing issues like increased costs and slower response times due to quality assurance measures (e.g., LLM self-reflection).

3.2 The Value of Humans

Advancements in AI capabilities highlight human value more concretely. In human-computer interactions, humans remain the most efficient agents, knowing their needs best. AI can self-reflect, but human decision-making aligns more closely with user requirements. Prioritizing accuracy over response time may not always be wise, as some decisions are better made by users, who should retain primary decision-making authority.

Ignoring human presence and overemphasizing model performance neglects user experience and the efficiency of human-computer collaboration. Transforming product users into co-creators aligns with the decentralized spirit of the crypto community. From preferences for knowledge bases and tools to prompt and workflow exploration, and the corpus for model training and fine-tuning, all components of DAgent rely on community participation. The TypoX AI team has initiated mechanisms for user collaboration, gradually opening up to the community, ensuring every TypoX AI user, regardless of hardware limitations, can participate in the DAgent ecosystem construction.

Mechanism Update Announcement

To enhance user experience and optimize platform mechanisms, TypoCurator will be implementing the following updates:

1. Invitation Mechanism Update

Return of Quota Rewards for Invitations:

Users can now gain quota by inviting new users to join the platform.

Details of the Mechanism:

For each successful invitation of two new users, if the new users complete one round of puzzles with an accuracy rate of 50% or higher (earning 0.6 $TPX), and join the TypoCurator official Telegram group chat, the inviter will receive one quota.

How to Join the Telegram Group Chat:

Users who have not yet joined the Telegram group can find the link on the welcome page, each puzzle result page, and the Profile page to join the TypoCurator official Telegram group chat.

Stay Updated:

Follow the latest news to ensure you don't miss any important updates and activities. 2. Anti-Cheat Mechanism Update

Changes to Quiz Rewards:

The anti-cheat information pop-up has been removed. Users will now receive a reward for each quiz attempt, with correct answers receiving higher rewards. We still encourage everyone to answer diligently to help improve the AI model's accuracy and capability.

Human Verification Mechanism:

A new verification mechanism has been introduced to prevent cheating and the misuse of automated tools or API operations for scoring.

Advertising Mechanism Update:

After solving a certain number of puzzles, the advertisement mechanism will be activated. 機制更新公告

為了提升用戶體驗並優化平台機制，TypoCurator 即將對現有系統進行以下更新：

1. 邀請機制更新

體力獎勵機制回歸：

用戶現在可以通過邀請新用戶來獲取體力獎勵。

具體規則：

用戶每成功邀請兩位新用戶，且每位新用戶完成一輪 puzzle 並達到 50% 以上的準確率（0.55 $TPX 獎勵），並加入 TypoCurator 官方 Telegram 群聊，邀請者將獲得一點體力。

如何加入 Telegram 群聊：

未加入 Telegram 群聊的用戶可以在歡迎頁、每輪答題結果頁及個人資料頁面中找到加入 TypoCurator 官方 Telegram 群聊的鏈接。

保持更新：

關注最新消息，確保不錯過任何重要更新和活動。 2. 防作弊機制更新

答題獎勵機制調整：

取消防作弊信息的彈出提示。 用戶每次答題均可獲得獎勵，正確答案將獲得更多獎勵。 我們仍然鼓勵用戶認真答題，以幫助提高 AI 模型的準確性和能力。

人機驗證機制：

引入新的驗證機制，以防止用戶通過自動化工具或 API 操作進行作弊和刷分。

廣告機制更新：

用戶在解答一定數量的題後，將觸發廣告展示機制。

Dataset Address

https://huggingface.co/datasets/typox-ai/Typo_Intent_OS

Dataset Description

3383 pairs of questions and answers about the Web3 knowledge base, used for training and testing AI models applied to Web3 scenarios.

Data Format prompt: AI-generated question. completion: The candidate answer selected by the majority of users. Example: { "prompt": "What is a primary advantage of using a decentralized finance (DeFi) platform?", "completion": "Direct peer-to-peer transactions without intermediaries." } Data Source

This dataset contains AI-generated questions and multiple candidate answers. The correct answers are selected by our Web3 product's end-users based on what they consider the most accurate. The answer chosen by the most users is marked as the completion. To ensure the quality of these evaluations, we use an incentive mechanism to encourage sincere responses. Additionally, we include some seed questions with known answers to filter users. Only those who perform well on these seed questions have their choices counted.

Annotation Method (Brief) Generation: Use TypoX to generate questions and candidate answers.
TypoX is a Rag system with a Web3 knowledge base, TypoX. https://www.typox.ai/ Evaluation: User selections are conducted through the TypoCurator telegram Mini-app. https://t.me/typocurator_bot
img.png Participation: Each question is evaluated by at least 300 people. Majority Rule: An option must be selected by more than 75% of participants to be considered the completion. Re-evaluation: If no option reaches 75%, the question is re-evaluated until an option reaches 80%. Invalid Questions: If a question is answered by more than 1000 people without any option reaching 75%, it is marked as invalid. Quality Assurance: We preset 500 seed questions with known answers to filter users. Only users who perform well on these questions have their choices counted. Statistics: On average, each question is evaluated by 453 people, with the completion option having an average selection rate of 78.9%. The dataset has also undergone two rounds of internal review. Chinese version

https://gov.typox.ai/topic/90/typocurator-数据集

一、For Web3 1.1 More 应用(Agent) or 通用模型

通用大模型虽然在不断发展，但是离真正满足用户的需求仍然有不小的距离。

AI应用的最小单元是Agent，构建一个Agent需要给LLM再组装上额外的知识库、工具以及相应的指示。LLM只是决定了Agent的下限，LLM可以使用的组件才决定了其上限。我们需要更多深度定制的Agent（而不是简单的GPTs）来满足用户的需求，进而通过多Agent的协调合作来实现更复杂的应用场景。

通用大模型的训练过程，就是在追求全局最优——多种能力（生成、编程、推理、数学）混合后的最优状态。所以这并不意味着它在局部上就一定优于一些针对性优化的模型。这也是一些MOE（混合专家）模型可以用更小的规模表现出等效于更大规模模型的原理。

训练一个小规模的LLM，固然能展示团队的实力，但是未必就能满足实际的场景。不过这也不是团队意愿的问题，训练数据的匮乏限制了训练/微调一个针对性模型的可能性。在确定聚焦的场景后，积累和挑选数据也不是立竿见影的，需要真实用户的交互和反馈。

1.2 TypoX AI

所以我们打造了TypoX AI这个平台

让用户能够在AI的协助下更好的进行调研，大大降低了调研门槛，让DYOR不再是一句空话。

为了实现这一目标，我们建构了专属的RAG（检索增强生成）框架，为LLM配置了Web3知识库和实时检索工具，可以认为我们打造了一款服务于Web3行业的Perplexity AI。

在RAG的加持下，我们总能得到质量优于原生LLM的输出，这恰好满足了我们进一步训练和微调模型的数据需要（训练/微调模型总是需要质量更高的问答对）。同时这些真实的交互数据也体现出了用户的偏好，使得模型能够更贴近社区的需求。

二、By Web3 2.1 AI发展到底应该怎么去中心化？

如果LLM模式的AI发展到极致，那最后留给人发挥的空间是什么？
至少需求还是由人提出的。实际上人将长期作为需求方存在，因为目前的生成式模型还是没有自主意识的，依然需要由人类来提出问题。

更一般地来说，模型依然是通过从Experience中学习，来提升在具体Task上的Performance。即使当前的LLM的能力相对于最初的生成Task早已溢出，但是对其溢出的能力的优化依然脱离不了这一框架：
提出问题： 确立Task，积累针对于该Task的数据（用于训练和测试）;
解决问题： 在数据上训练，优化模型在具体Task上的表现，从而完成新的工作。

I. 算力去中心化的发展受限

物理规律限制了模型训练和部署（解决问题）的去中心化路径。通信成本是无法克服的瓶颈，相较于集中的算力集群，显卡网络更像是一座座孤岛。 小规模的显卡群基本只能部署生图模型和10B级别的LLM，模型训练上的差距只会更大。这不仅是效率问题，更是能力问题。 AI发展在一定时间内可能都无法摆脱对于集中化高质量算力的依赖。

II. 需求（评价&数据）的去中心化

很多人聚焦于硬件问题，却忽视了数据的重要性，实际上对于训练数据的规模和质量只会逐步提升。

本轮AI的发展是基于语言模型的 Scaling Law，即模型规模越大，能力越强。更大的模型，就需要更多的数据用于训练； 一个模型是无法从其本身生成的内容中学习更多的。 小规模的LLM可以使用GPT4等更大体量模型生成的数据来进行训练/微调，而GPT4级别的头部模型需要的数据质量只会更高。

谁掌握了数据，谁就掌握了AI发展的话语权。目前AI发展话语权是由少部分行业精英掌握的，这是通过树立评估标准、建立训练数据库来实现的。如果不能提出新的需求，那等来的将是更多打榜的通用型模型，用户所真正需要的场景将无法得到解决。

是时候由用户来提出自己对于AI发展的需求了。为了实现这一目标，我们需要建立用户自己的评价体系，进而构建专属的训练数据库，以此来获得对于AI发展方向的话语权。

2.2 TypoCurator

目前对于通用大模型有一些评价手段，但是针对具体AI应用的评价手段还很匮乏。针对于一些有明确答案标准的场景，比如意图识别场景（用户的需求应该调用哪一个DApp/DAgent），需要建构对应的标签数据库，以此来进行针对性的评价和优化。

由此我们最新推出了TypoCurator这个产品，借助Ton生态广泛的用户群体，通过社区分工的形式来构建聚焦于Web3场景的标签数据库，为我们之后训练Web3 AI OS做数据准备。对于普通用户而言，只要完成日常答题/标注任务，即可获得 $TPX 奖励。

2.3 Agent Arena (Soon)

除了标签数据之外，有大量的应用场景是没有明确标准的。比如根据项目信息生成调研报告，在保证事实清晰的基础上，依然会存在关注重点、行文逻辑和风格等问题。这些需要用户来进行对比选择。

受到LMSYS Arena的启发，我们即将推出Agent Arena。用户将在不知道Agent信息的情况下进行对比选择，帮助我们实现对Agent和模型的评估。早期将聚焦于Web3的行业场景，用户的评价将帮助我们更好地优化TypoX AI，从而更好地服务于Web3行业。

我们在LMSYS Arena的基础上，将评估对象从通用模型进一步推广到Agent，同时对于机制进行了进一步优化，降低了作弊和偏见风险。我们希望为AI开发者提供公允的评价体系和参考，以及定制化的AB测试流程，从而促进整个AI应用生态的发展。

一、拥抱随机性

要打造一款AI产品，首先应该知道要做的工作是什么。简单来说，就是控制LLM的随机性。随机性是一把双刃剑，用好了叫能力泛化，用不好就成了幻觉。在创作领域，对于精确性的容忍度高，随机性促成了创作的天马行空，相信大部分用户最初被大模型震撼的都是其创作生成能力。但是在面对事实和数学代码的严密逻辑时，就只有对错，没有模糊空间，所以幻觉大多出现在这种情形。

在数学和代码方面，较早的优化是给LLM配置外部工具（如Wolfram Alpha）、编译环境（Code Interpreter），随后是进一步的加强训练，进一步在一些工作流的加持下，可以解决相对复杂的数学问题，完成一定长度的实用代码。

在客观事实方面，大模型确实是能将部分信息存储在模型参数中，但是很难预知或者预设哪些知识点是必然拟合好的，而一味提高在训练上的准确性往往会导致过拟合，减弱泛化能力；同时大模型的训练数据总是有截止日期的，即使针对所有训练数据的问答都完美匹配，也无法预测出之后发生的事，况且大模型的训练成本决定了模型的更新不会过于频繁。所以通过使用外置的知识库增强生成结果才得到了广泛的重视。

二、RAG的挑战与愿景 2.1 什么是RAG

检索增强生成（Retrieval Augmented Generation, 即RAG) 早在2020年就被提出了，也是随着近期LLM产品的发展需求，才获得了比较广泛的应用。目前主流的RAG流程不涉及对模型本身参数的修改，是通过增强给模型的输入来提高生成内容的质量。用一个简单公式 y=f(x) 来解释一下：f 是大模型，一个参数量很大的（随机非线性）函数；x 就是输入项；y 就是输出项；RAG就是在 x 上下功夫，来提高 y 的质量。

因为不涉及模型本身，所以RAG在成本上有很大优势，组合性也高，可以随意替换模型使用。只需检索出最相关的内容，提供给大模型，相较于直接全文输入，节省了所需Token数量，除了降低使用成本外，也大大压缩了回复用户的反应时间。

2.2 长上下文模型

近阶段出现了很多支持长上下文（输入窗口）模型，可以直接将完整的长篇巨著直接丢给大模型来处理，算是克服了大模型早期输入短的限制，这也是最初RAG的诱因之一，许多人认为RAG也许会就此势微。

在大海捞针测试（让大模型在一段长文本中找到对应的内容）中，长上下文模型确实比之前的模型有更好的表现，但是相对于开头和结束部位，在处理文本中间的针时的表现有些差强人意；同时大海捞针测试本身也是有瑕疵的，目前公开的测试主要还是单针的（即要寻找的线索只有一条），从单针到多针，问题复杂度会急剧提升，对大模型而言依然是挑战，另外也有第三方测试者发现有些模型可能在训练时就已经得知需要找的针是什么了(有作弊嫌疑)。新模型在长上下文的表现上可能确实有提升，但是鲁棒性可能没有预期中的那么高。

当然大模型在长上下文上还会继续优化，目前的瑕疵可能过段时间也许不是问题了，但是成本对于训练者和使用者来说依然是较高的门槛。相较于直接粗暴得将所有文本一股脑甩给模型，RAG在成本和时效性上依然有不可替代的优势。但RAG也需要根据模型的进步进行调整，特别是文本处理的颗粒度，避免出现过度处理反而降低整体效率的情况。在具体的产品中，我们总是需要在成本、时效和质量之间做权衡。

2.3 RAG的愿景

相较于简单使用向量库存储召回的早期RAG应用，我们需要从更系统的角度来审视和构建新的RAG框架，也许我们可以从RAG的持续研究中获得启发。内容、检索方式与检索器、生成器（LLM），是RAG系统的基本元素。

针对检索器的优化相对丰富，同时也因为其便利性，是应用最早也最多的RAG方法。除此之外，应该更多地考虑内容本身，向量数据库未必就是唯一的选择，键值/关系型/图数据库也是可以考虑的方案，一切根据数据的结构来进行选择，同时相应的调整检索器就可以进一步增加检索聚合的效果。

另外LLM也可以在内容预处理中扮演更重要的角色，相较于简单的切片-向量化-检索路径，使用分类模型和LLM进行提炼总结和重构，将内容以对LLM更友好的形式进行存储，同时更加依赖语义而不是向量距离来进行检索，将大幅提高生成的效果，TypoX AI团队在产品实现中也验证了这一点。

除了单独优化其中某一元素，也可以同时优化多个元素，尤其是大部分RAG方法都是锁定（不涉及）生成模型的，将检索与生成模型一起训练，也将是RAG的一个很重要的方向。由此引出了另一个话题：RAG、微调和LLM的关系。每逢LLM本身能力的提升与更低成本的微调技术的出现，都会有人会说要用微调干掉RAG，RAG已死。

回到前文提到的 y=f(x) 上，LLM和微调影响的都是 f 本身的参数，目的是在不加强输入x 的情况下，就能提升生成 y 的质量。而RAG则是通过加强 x 的输入，来提高 y 的质量。两种方案并不是矛盾的，而是相向而行的，模型的能力提升和输入的加强都有助于生成结果的优化。同时进阶的RAG也会涉及到模型对于知识库的针对性微调，将不再局限于对于输入的增强上。RAG的价值就在于，比原始LLM有更好的生成效果，（即使是与微调模型比，RAG的时效性和效果也是更好的，同时配置成本也不会高于微调的成本），以及与把全部知识直接丢给LLM相比，又有更高的效率。

三、人与AI 3.1 大模型的定位

笔者是一贯反对唯模型论和模型极大论的，大模型是很强，但是单独的大模型距离用户真正需要的Agent还是有不少距离的。有独家训练的小模型或微调模型，固然是团队实力的展现，但是落地到产品上，就一定能够给用户提供更好的体验吗？也许将这一部分的资源投入到其他要素（工具、知识库、工作流）上，提升效果可能更明显。而且目前在Web3行业缺乏自己的训练数据和评价标准的情况下，我们行业真的需要更多的打榜通用（小）模型吗？笔者在此呼吁，加密行业需要构建专属的训练数据库和评价标准，进而构建更多更适配行业的专属模型。

当然也不应该低估大模型的能力，在TypoX AI产品的早期探索中，为了使得流程更可控，我们给予了模型较低的信任，有很多流程目前来看都过度开发了。同时也存在为了保证生成质量，而导致成本增加，回复速度过慢的问题（比较典型的就像让LLM进行自反思）。我们在后续的迭代中，逐步优化流程中硬逻辑与LLM的参与度，达到了一个相对较优的平衡。

3.2 人的价值

AI能力的提升，其实反而更突显了人的价值，使得人的价值更具象化了。在人机交互中，人反而是最高效的Agent。只有人最知道自己要什么不要什么，人永远是甲方（需求方），AI固然可以进行自我反思，但是提升质量和准确性，结果也未必符合用户的需求。所以在产品的交互中，过于重视准确性，忽略时间延迟，未必是明智之举。有些决策过程更适合用户自己执行，应该把主要的决策权让渡给用户。

唯模型论的另一方面，就是忽视了人的存在，一味追求模型的表现，忽视了人机交互中人的体验，以及人机协作对于整体流程的效率提升。将产品用户转变为共建者，这也是加密社区去中心化精神所在。 从对知识库、工具的偏好，提示词、工作流的挖掘，到模型训练和微调所需要的语料，DAgent的所有部件都离不开社区的参与。TypoX AI团队已经初步构建了一系列与用户共建的机制，后续将逐步向社区公布，让每一位TypoX AI用户无需考虑硬件门槛，都有机会参与到DAgent生态的建设之中。

设个每日排行榜激励一下，前十名可以获得多少的奖励，然后社交任务系统，分为社交任务和答题任务，连续每天完成登录答题可以获得5次额外答题机会这样子的

Labelbox

Official Website: https://labelbox.com/

Labelbox is currently one of the most mature annotation platforms, supporting various data types and annotation forms, including images, text, point clouds, maps, videos, and medical DICOM images. It offers a wide range of annotation management templates and provides services through AI or human-assisted annotation. Additionally, Labelbox supports operations through its Python SDK.
248105c6-dcb3-4d96-bd16-a13bc129908e-image.png

In terms of team collaboration, Labelbox provides a series of management tools. However, its interface is relatively traditional and somewhat cumbersome compared to the more modern interfaces of newer projects.

3d24c42a-670b-4720-becd-c7c77133b93e-image.png

Amazon SageMaker Ground Truth (AWS GT)

Official Website: https://aws.amazon.com/cn/sagemaker/groundtruth/

AWS GT is one of the most traditional annotation service products, offering both AWS-managed and self-service forms. The self-service option allows users to manage their data and the final testing process using AWS-provided annotation tools, while the managed service can provide AI and human annotation services.
a7179884-46ee-47c5-b747-5aae59b5717c-image.png
8146c33e-ff26-41ac-9487-1c4a4d2f7985-image.png
In terms of team collaboration, AWS GT supports team collaboration and large-scale annotation project management through its tight integration with the AWS ecosystem. Users can use AWS IAM to control access permissions and resource management.

Comparative Analysis

Both Labelbox and AWS Ground Truth have their own advantages, and users can choose the appropriate platform based on their needs.
Labelbox: Suitable for users who need highly customized and flexible annotation processes, particularly those in small to medium-sized enterprises and research institutions with high requirements for team collaboration and data quality.
AWS Ground Truth: Suitable for users already using the AWS ecosystem, especially those large enterprises and projects that require large-scale data annotation and automated annotation features.

DYOR Research Introduction to CrewAI (5).png

市場狀態報告 💡

截至 2024 年 6 月 19 日

宏觀經濟

美聯儲的降息策略

美聯儲官員對通脹接近2%目標保持謹慎樂觀，並考慮今年晚些時候降息。最新數據顯示，從四月到五月，消費者價格沒有上漲，表明通脹壓力正在緩解。然而，強勁的勞動力市場使得降息決策變得複雜。零售銷售在五月僅上升0.1%，反映了高通脹和高利率對消費者支出的影響。 美國預算赤字預計將大幅增加，因為各項計劃支出增加和利息成本上升。

關鍵見解:

Adriana Kugler: 支持在經濟條件改善的情況下放寬政策，指出當前的貨幣政策已經足夠限制價格壓力。 Austan Goolsbee: 最近的通脹數據顯示下降而沒有增加失業率，讓人感到鼓舞。 John Williams: 預計未來幾年內會逐步減少利率，取決於經濟數據。 零售銷售: 溫和的增長突顯了消費者在高通脹和高利率環境中的挑戰。 預算赤字: 預計在2024年上升至1.9萬億美元，公共債務預計到2034年將達到50.7萬億美元。

為什麼重要: 美聯儲的謹慎態度反映了需要在控制通脹和經濟增長之間取得平衡，影響借貸成本、消費支出和整體經濟穩定。預算赤字的上升突顯了影響經濟政策決策的財政挑戰。

替代方案: 快速降息可能刺激經濟增長但會增加通脹風險。保持高利率可以抑制通脹但會放慢經濟增長。減少支出或增加稅收可以管理赤字，但可能面臨政治阻力。

英國通脹與薪酬協議

五月英國通脹首次回到2%的目標，但基本價格壓力仍然強勁。在截至五月的三個月內，英國的基本工資中位數上升了4.6%，高於前三個月的4.5%。

關鍵見解:

經濟學家預測: 可能在八月降息，金融市場預計更有可能在九月或十月。 英國央行立場: 強調需要持續的數據改善才能改變貨幣政策。 薪酬協議: 輕微增加反映了雇主在生活成本壓力和經濟不確定性中的謹慎樂觀態度。

為什麼重要: 回到目標通脹是一個積極發展，但強勁的基本價格壓力可能會延遲降息，影響借貸成本和經濟增長。

替代方案: 立即降息可能刺激經濟增長但風險重新點燃通脹。延遲降息保持穩定但可能放慢經濟復甦。

韓國的貨幣穩定努力

韓國的外匯當局正在採取措施防止韓元進一步對美元貶值，這對經濟構成挑戰。

關鍵見解:

干預措施: 如果美元兌韓元匯率超過1,385，將在現貨市場進行干預。 經濟穩定: 穩定韓元對於維持經濟穩定至關重要。 韓元表現: 今年對美元貶值了6.5%，引起政策制定者的擔憂。

為什麼重要: 穩定的貨幣有助於管理通脹並支持經濟增長。持續的疲軟可能會增加進口成本和通脹。

替代方案: 直接干預可以穩定貨幣但會消耗儲備。允許韓元貶值可以促進出口但會增加通脹。

傳統金融市場

美國股市趨勢

市場變動有限，主要指數保持平穩。儘管英偉達上漲了3%，納斯達克仍小幅下跌。商品市場顯示出反彈跡象，油價突破80美元，銅價在達到4.4美元後反轉，黃金接近2330美元。

關鍵見解:

零售數據: 五月零售數據疲軟，四月數據低於預期，反映消費者支出下降和創紀錄的低儲蓄率。 美聯儲立場: 官員對今年只有一次降息保持謹慎樂觀。 美元指數: 由於零售數據疲軟，美元指數下跌，保持在105.2的支撐位。 商品: 商品市場的反彈表明可能出現新的漲勢。

為什麼重要: 市場變動反映了投資者情緒和經濟健康狀況，受零售數據和中央銀行政策影響。商品市場的反彈表明可能出現新的漲勢。

替代方案: 更強勁的經濟數據可以支持市場增長，而較弱的數據可能會促使更加謹慎的投資策略。

奢侈品銷售停滯

全球奢侈品銷售在2024年預計將停滯，在第一季度略有下降之後。

關鍵見解:

創意領導: 主要時裝公司面臨創意領導的變革，缺乏明確性和興奮點。 定價策略: 品牌正在提高價格，專注於最富有的顧客，犧牲中產階級和年輕一代。 經濟不確定性: 美國和中國的政治和經濟不確定性促成了這種放緩。

為什麼重要: 奢侈品市場的表現反映了更廣泛的經濟趨勢和消費者情緒，受定價策略和經濟狀況影響。

替代方案: 創新的營銷和定價策略可以吸引更廣泛的客戶群，專注於體驗式奢侈品可以抵消實物商品銷售的下降。

加密貨幣

去中心化金融和區塊鏈發展

關鍵發展包括RiscZero的生產就緒zkVM、API3的OEV網絡貸款協議，以及加密貨幣領域的重大融資和投資。

重要新聞:

RiscZero的zkVM: 增強區塊鏈的可擴展性和安全性。 API3的OEV網絡: 提供創新的預言機解決方案。 投資: 重大投資表明對加密貨幣市場的信心增強。 以太坊: 美國證券交易委員會將撤銷調查。 比特幣: 比特幣價格上升趨勢“完好無損”，持有者盈利120%

Web3與AI整合的新興趨勢

Web3和AI的整合正在創造新的機會和發展路徑。該領域的主要發展包括：

關鍵見解:

NEAR AI: 計劃分三個階段實施開源用戶擁有的AGI。 Mysten Labs: 推出Walrus，一個去中心化的存儲和數據可用性協議。 Solana's Sonic: 完成由Bitkraft領投的1200萬美元A輪融資。 Renzo協議: 獲得1700萬美元的流動性再質押協議。 Lista DAO: 6月20日開放空投申領。 Starknet: 更新路線圖，預計v0.13.2在八月將交易確認時間縮短至2秒。

為什麼重要: AI與Web3技術的整合正在推動創新，擴展區塊鏈網絡的能力，提供去中心化和用戶賦權的新工具。

替代方案: AI和Web3的持續進步可以進一步增強它們對各行業的聯合影響，促進更高效和安全的數字生態系統。

DYOR Research Introduction to CrewAI (5).png

The State of Markets 💡

As of June 19, 2024

Macroeconomics

Federal Reserve's Approach to Rate Cuts

Federal Reserve officials are cautiously optimistic about inflation nearing the 2% target and are considering rate cuts later this year. Recent data shows no rise in consumer prices from April to May, suggesting easing inflation pressures. However, the strong labor market complicates rate cut decisions. Retail sales rose by just 0.1% in May, reflecting high inflation and interest rates' impact on consumer spending. The U.S. budget deficit is projected to increase significantly due to higher spending on various programs and rising interest costs.

Key Insights:

Adriana Kugler: Supports easing policy if economic conditions improve, noting current monetary policy is restrictive enough to ease price pressures. Austan Goolsbee: Encouraged by recent inflation data showing a drop without increasing unemployment. John Williams: Expects gradual interest rate reductions over the next few years, contingent on economic data. Retail Sales: Modest growth highlights challenges consumers face amid high inflation and interest rates. Budget Deficit: Projected to rise to $1.9 trillion in 2024, with public debt expected to reach $50.7 trillion by 2034.

Why it matters: The Fed's cautious approach reflects the need to balance inflation control with economic growth, influencing borrowing costs, consumer spending, and overall economic stability. Rising budget deficits highlight fiscal challenges impacting economic policy decisions.

Alternatives: Rapid rate cuts could stimulate economic growth but risk increasing inflation. Maintaining higher rates could curb inflation but slow economic growth. Reducing spending or increasing taxes could manage the deficit but may face political resistance.

UK Inflation and Pay Deals

UK inflation returned to its 2% target in May for the first time since 2021, but underlying price pressures remain strong. Median basic pay settlements in the UK rose by 4.6% in the three months to May, up from 4.5% in the previous three months.

Key Insights:

Economists' Prediction: A rate cut may occur in August, with financial markets anticipating it more likely in September or October. BoE's Stance: Emphasizes the need for sustained data improvement before altering monetary policy. Pay Deals: Slight increase reflects cautious optimism among employers amid ongoing cost-of-living pressures and economic uncertainties.

Why it matters: Returning to target inflation is a positive development, but strong underlying price pressures could delay rate cuts, impacting borrowing costs and economic growth.

Alternatives: Immediate rate cuts could stimulate economic growth but risk reigniting inflation. Delaying cuts maintains stability but may slow economic recovery.

South Korea's Currency Stabilization Efforts

South Korea's foreign exchange authorities are taking steps to prevent further weakening of the won against the dollar, which poses economic challenges.

Key Insights:

Intervention Measures: To intervene in the USD/KRW spot market if the 1,385 level is breached. Economic Stability: Stabilizing the won is crucial for maintaining economic stability. Won's Performance: Weakened by 6.5% against the dollar this year, causing concern among policymakers.

Why it matters: A stable currency helps manage inflation and supports economic growth. Continued weakness could increase import costs and inflation.

Alternatives: Direct intervention can stabilize the currency but deplete reserves. Allowing the won to weaken could boost exports but increase inflation.

TradFi Markets

U.S. Stock Market Trends

Limited market movement with major indices remaining flat. Nvidia's 3% rise couldn’t prevent a slight Nasdaq dip. Commodities showed signs of rebound with oil crossing the $80 mark, copper reversing after hitting $4.4, and gold nearing $2330.

Key Insights:

Retail Data: Poor May retail data and lower-than-expected April figures reflect consumer spending decline and record low savings rates. Fed's Stance: Officials express cautious optimism about only one rate cut this year. Dollar Index: Fell due to weak retail data, maintaining support at 105.2. Commodities: Rebounds in commodities indicate potential new rally phases.

Why it matters: Market movements reflect investor sentiment and economic health, influenced by retail data and central bank policies. Rebounds in commodities indicate potential new rally phases.

Alternatives: Stronger economic data could support market growth, while weaker data could prompt more cautious investment strategies.

Luxury Sales Stagnation

Global luxury sales of handbags, shoes, and apparel are expected to stall in 2024, following a slight dip in the first quarter.

Key Insights:

Creative Leadership: Major fashion houses face transitions in creative leadership, leading to a lack of clarity and excitement. Pricing Strategies: Brands are increasing prices, focusing on wealthiest customers at the expense of the middle class and younger generations. Economic Uncertainties: Political and economic uncertainties in the US and China contribute to the slowdown.

Why it matters: The luxury market's performance reflects broader economic trends and consumer sentiment, influenced by pricing strategies and economic conditions.

Alternatives: Innovative marketing and pricing strategies could attract a broader customer base, while focusing on experiential luxury could offset tangible goods' sales decline.

Crypto

DeFi and Blockchain Developments

Key developments include the launch of RiscZero's production-ready zkVM, API3's OEV Network lending protocol, and significant fundraising and investments in the crypto space.

Key News:

RiscZero's zkVM: Enhances blockchain scalability and security. API3's OEV Network: Offers innovative oracle solutions. Investments: Significant investments indicate growing confidence in the crypto market. Ethereum: SEC to drop investigation. Bitcoin: Bitcoin price uptrend ‘intact’ with hodlers 120% in profit

Emerging Trends in Web3 and AI Integration

The integration of Web3 and AI is creating new opportunities and development paths in the digital landscape. Key developments in this area include:

Key Insights:

NEAR AI: Plans to implement open-source user-owned AGI in three phases. Mysten Labs: Launches Walrus, a decentralized storage and data availability protocol. Solana's Sonic: Completes $12 million Series A funding led by Bitkraft. Renzo Protocol: Secures $17 million for its liquidity re-pledge protocol. Lista DAO: Opens airdrop claim on June 20. Starknet: Updates roadmap, with v0.13.2 to reduce transaction confirmation time to 2 seconds by August.

Why it matters: The integration of AI with Web3 technologies is driving innovation and expanding the capabilities of blockchain networks, providing new tools for decentralization and user empowerment.

Alternatives: Continued advancements in AI and Web3 could further enhance their combined impact on various industries, promoting more efficient and secure digital ecosystems.

Quota Mechanism Adjustment

Dear Users,

We have received your feedback regarding the quota mechanism and have made the following adjustments:

Daily quota recovery is now set to 2 points: A maximum of 2 points of quota will be recovered daily. Users with more than 2 points will not receive additional recovery, but the excess will not be reset.

Changes to the quota purchase prices:

New prices: Amount ($TPX) quota Gained Reward Multiplier 10 $TPX 12 1.2 20 $TPX 23 1.15 50 $TPX 55 1.1 100 $TPX 105 1.05 200 $TPX 210 1.05 Old prices: Recharge Amount Quota Gained Reward Multiplier 10 $TPX 12 Quota x1.2 20 $TPX 24 Quota x1.2 50 $TPX 65 Quota x1.3 100 $TPX 140 Quota x1.4 200 $TPX 280 Quota x1.4 500 $TPX 800 Quota x1.6

Note 1: For fractional amounts, the recharge will be rounded down to the nearest whole number.
Note 2: Specified amounts will receive a recharge bonus. Non-specified amounts will be exchanged 1:1 for quota.

Thank you for your support and understanding.

Enjoy your puzzles!

體力機制調整公告

尊敬的用戶：

我們收到了您對於體力機制的反饋，並對此進行了如下調整：

每日恢復體力值更改為 2 點：每日只會恢復最多 2 點體力，超過 2 點的用戶將不會繼續恢復體力，但超出的部分不會清零。

體力購買價目變更：

新價目： 充值金額（$TPX） 獲得體力 獎勵倍率 10 $TPX 12 1.2 20 $TPX 23 1.15 50 $TPX 55 1.1 100 $TPX 105 1.05 200 $TPX 210 1.05 舊價目： 充值金額 獲得體力 獎勵倍率 10 $TPX 12 x1.2 20 $TPX 24 x1.2 50 $TPX 65 x1.3 100 $TPX 140 x1.4 200 $TPX 280 x1.4 500 $TPX 800 x1.6

注1：對於小數數額，充值將按面值向下取整計算。
注2：指定金額將有充值優惠，非指定金額會1比1將 $TPX 等額兌換為體力

感謝您的支持和理解。

祝您解谜愉快！

The Open Network (TON): Sector Performance Overview

The Open Network (TON) is a high-performance blockchain platform originally developed by Telegram. It supports over 650 decentralized applications (dApps) across various sectors such as DeFi, NFTs, gaming, and memecoins with a TVL exceeding $590 million. TON's integration with Telegram enhances its usability, providing seamless access to blockchain functionalities for Telegram's extensive user base.

Top Performing Sectors in TON.png

1. Decentralized Finance (DeFi) - 25%

DeFi is the leading sector in TON, contributing to 25% of the ecosystem's performance. DeFi applications on TON provide a wide range of financial services, from lending and borrowing to staking and trading.

Top Projects:

StonFi: A leading decentralized exchange on Telegram, offering a user-friendly platform for swapping cryptocurrencies with significant liquidity and staking options.
Screen Shot 2024-06-18 at 16.37.46.png Evaa: The first lending and borrowing protocol on TON, allowing users to earn interest by lending their assets and using borrowed funds within the ecosystem.
Screen Shot 2024-06-18 at 16.39.05.png DeDust: A decentralized exchange similar to Uniswap, integrated with Telegram, enabling users to swap and manage their TON-based portfolios.
Screen Shot 2024-06-18 at 16.39.55.png Bemo: The first liquid staking solution for the TON token, allowing users to earn staking rewards while using staked assets in other DeFi dApps.
Screen Shot 2024-06-18 at 16.40.44.png 2. Non-Fungible Tokens (NFTs) - 20%

NFTs are a significant part of the TON ecosystem, contributing to 20% of its performance. These tokens represent unique digital assets and are used in various applications such as gaming, digital art, and virtual real estate.

Top Projects:

PunkCity (PUNK): Combines an NFT collection with a play-to-earn game and a metaverse for crypto-based gaming.
Screen Shot 2024-06-18 at 16.41.47.png TON Diamonds NFT: Offers unique animated diamonds with exclusive privileges and a robust trading platform.
Screen Shot 2024-06-18 at 16.42.32.png 3. Gaming - 15%

The gaming sector accounts for 15% of TON's performance. Gaming on TON leverages blockchain technology to offer play-to-earn models, creating opportunities for users to earn rewards through gameplay.

Top Projects:

NOTcoin: A click-to-earn game on Telegram with over 30 million players, known for its rapid growth and social interaction features.

Jetton Games: A gaming platform with over 500 games integrated into Telegram, offering financial rewards and its own $JETTON token.
Screen Shot 2024-06-18 at 16.46.39.png

GAMEE: A decentralized gaming platform on TON, allowing users to play, earn, and trade in-game assets with the native $GMEE token.
Screen Shot 2024-06-18 at 16.46.54.png

Hamster Kombat (HMSTR): A tap-to-earn game on Telegram, popular for its engaging gameplay and substantial user base.

4. Social Media - 10%

Social media applications on TON enhance user engagement and content monetization. This sector contributes to 10% of the ecosystem's performance.

Top Projects:

Resistance Dog (REDO): A memetoken project aimed at growing the community and fighting censorship through the popularity of memes.
Screen Shot 2024-06-18 at 16.49.58.png Resistance Cat ($RECA): Another memetoken focused on community engagement and humor.
Screen Shot 2024-06-18 at 16.50.25.png 5. Infrastructure - 18%

Infrastructure projects on TON provide the backbone for various applications, ensuring scalability, security, and efficient operation of the blockchain.

Top Projects:

Tonkeeper: A non-custodial Web3 wallet for secure asset management, integrated with various services and supporting staking and trading.
Screen Shot 2024-06-18 at 16.51.19.png TonStake: A staking platform offering Staking-as-a-Service, promoting network security through a proof-of-stake mechanism.
Screen Shot 2024-06-18 at 16.52.03.png 6. Payments - 12%

Payment solutions on TON facilitate seamless and secure transactions, leveraging the blockchain's speed and efficiency. This sector accounts for 12% of the ecosystem's performance.

Top Projects:

Toncoin (TON): The primary cryptocurrency of the TON ecosystem, designed to support the integration with Telegram for various crypto activities including DeFi and NFTs.
Screen Shot 2024-06-18 at 16.54.44.png HyperGPT: A Web3 AI marketplace that uses TON to offer decentralized access to various AI applications, streamlining management and reducing costs.
Screen Shot 2024-06-18 at 16.54.58.png Conclusion

The TON ecosystem showcases a diverse range of projects and sectors, each contributing to the platform's growth and adoption. The integration with Telegram amplifies TON's reach, making blockchain technology accessible to a broader audience. With sectors like DeFi, NFTs, gaming, social media, infrastructure, and payments leading the way, TON is well-positioned to become a major player in the blockchain space.
This list is subject to future updates.