注册并分享邀请链接,可获得视频播放与邀请奖励。

与「base」相关的搜索结果

base 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 base 的内容
原来这就是赛博活佛的含金量!!免费用啊家人们,这得省多少token!! DeepSeek-V4.1-Flash 最新模型,免费用两周,每天 200 次调用不限速!! 我一开始还以为看错了,反复确认了三遍 作为一个每天都在跑 Agent 任务的人,我最头疼的事不是模型不够强,是管理成本太高 DeepSeek 一个账号、Kimi 一个账号、GLM 一个账号、Qwen 一个账号,API Key 管了五六把,每个月充值要开四五个后台,对完账才知道这个月到底在 AI 上烧了多少钱 更离谱的是,大部分平台充进去的额度根本用不完,但你不充又没法调用,等于在给每个平台交月租 后来发现了 HiLinkup @HiLinkup_AI ——一个 API Key,直接接 DeepSeek、Qwen、Kimi、GLM、Doubao、MiniMax,文本图像视频音频全覆盖 我试了一下,换个 base URL 就行,代码一行不用改,OpenAI SDK 直接兼容 响应速度很快,关键是价格——大概是官方的 1 折左右!!,没有 TPU 上限 说白了就是:以前我每个月在五六个平台上分别充钱,现在一个账户统一计费,按量走,用多少算多少 关键是它现在每天送 1500 万 token 免费额度,日常跑 Agent 任务完全够用 唯一的门槛就是首充 5 块钱激活一下,充完还送 5 块,而且充进去的钱不消耗,每天只走免费额度 对于我们这种每天都在用国产模型跑各种任务的人来说,这个事的意义不只是省钱——是终于不用当五个平台的出纳了 一个 Key 管所有模型,按任务选型,不被任何一家绑死 说实话这种聚合平台的窗口期不会太长,各家模型迟早会收紧 API 授权,能用的时候赶紧先用起来!! 链接:
显示更多
0
30
86
16
转发到社区
The bonus mechanism is out. TL;DR Everyone gets a 25% unlock at TGE. On top of that, participants will receive additional bonus tokens fully unlocked at TGE, based on their original committed amount before pro-rata dilution. How many bonus tokens do I get? Check out the chart below for the exact formula and allocation examples. Here, x is your original committed amount before pro-rata dilution, and 2,472 represents the total number of Axis believers who decided to ape in even with strict unlock terms.
显示更多
0
119
415
64
转发到社区
🚨 HOLY CRAP! UK authorities are RELEASING the 5 suspected terrorists who got close to bombing/attacking RAF Fairford base, which is used by American troops RELEASED ON BAIL. This is UTTER MADNESS! "Following extensive interviews...the 5 men who were arrested are being released on police bail this afternoon." "This does not mean the investigation is over. They are subject to stringent conditions." Why would they do this?! KEEP THEM BEHIND BARS Under NO CONDITIONS should they be released. Even President Trump said we knew about this would-be attack before it happened Now they're going back into society? THE UK IS COOKED unless this soft-on-crime BS stops!
显示更多
0
2.2K
18.4K
5.9K
转发到社区
BNB Chain is already home to one of the largest user bases in crypto with rapid growth across stablecoins and RWAs. Today, we’re welcoming @thomasxchen as our Chief Business Officer as BNB Chain deepens its focus on institutions, capital and onchain market activity. Thomas brings deep experience across traditional finance and digital assets, with a strong track record building institutional businesses across capital markets, custody, trading and growth. At BNB Chain, his focus is clear: bring more institutions and capital onchain, deepen liquidity, grow volume and strengthen long-term value across the ecosystem. Welcome to BNB Chain, Thomas. Let’s keep building.
显示更多
0
65
215
22
转发到社区
NVIDIA 发布 Skill2Env:用“集体技能”强化智能体 NVIDIA 研究者们把社区公开的 Agent Skills 编译成可执行 RL 训练环境的数据流水线:3.4k 个 Skills 变成 8k 个带程序化测试和行为量规的终端任务;仅 300 步 RL 训练就让 Qwen3.8-27B 在 Terminal-Bench 2.1 上提升 4.7 个百分点,且模型行为显著向源 Skills 的方法论对齐。 开源项目: 核心洞察:公开 Agent Skills 是一个被忽视的监督来源 Agent Skills 是“教智能体做某件事”的文件夹:一个 SKILL.md 加上可选的脚本、参考资料和资产。论文指出,把公开 Skill 语料当作数据来读,它同时提供三样东西: · 任务分布的采样:人们真正想让智能体处理的任务分布(有人愿意花时间写下工作流,说明这活儿值得自动化); · 真实世界的锚点:指向真实的仓库、数据集、工具和工件; · 结果测试表达不了的质量标准:领域专长、默认参数、常见坑、“好结果长什么样”。 # 数据流水线:四阶段编译,验证靠构造 1. Plan(分解):容器化的 Codex 规划器读取完整 Skill 包、联网调研相关公共资产,把 Skill 拆解成若干可验证的 workflow,每个附带元计划(场景、初始世界、预埋缺陷、难点来源、解法草案、验证策略)、资产建议和“任务轴池”(任务原型 × 验证器模式 × 人物画像)。 2. Diversify(多样化):宿主从轴池采样一组组合,加上复杂度、指令语气、请求者专业水平。关键设计是轴池以 workflow 为条件:研究型 workflow 配“证据可追溯”验证和研究者画像,而不是从全轴乘积空间乱抽,这让多样化保持 sensible。 3. Create(构造):全新创建者 Codex agent 在 Docker 内工作,尽可能用真实素材(钉在特定 commit 的开源仓库、真实版本化文档、官方 API 规范);需要联网服务的场景改造成本地替身(stub 服务器、录制回放 fixture、PATH 上的假 CLI、种子数据库),求解时绝不依赖网络。创建顺序被严格固定:先建世界 → 写指令 → 写测试 → 写量规 → 最后才写参考解,测试先于解法冻结,保证解法必须迁就评分契约而非反过来。 4. Verify(验证):宿主端无模型参与的接收门:静态检查(布局、符号链接、Dockerfile 安全、基础镜像按内容摘要钉死)+ 两个容器内试跑:Oracle(参考解)必须全指标满分,NOP(什么都不做的 agent)必须全指标零分。任一失败即拒绝。 值得注意的一个反直觉选择:不做 teacher 模型预验证(不像部分工作用强模型试解、解不出就丢弃任务)。理由有二:这会把任务难度上限压到验证器能力,且成本翻倍;而 group-based RL 的在线动态过滤(rollout 无优势的 prompt 自动不产生梯度)天然淘汰过难/过易任务。 # 数据画像:广、贵、且忠实于源 规模与成本:7,971 个任务,用 GPT-5.6 Sol(xhigh 推理档)生成,API 花费超 9 万美元。(脚注:出于法律原因,公开发布的数据集改用 Kimi-K3-max 在同一流水线下生成。) 领域分布:13 个领域中,软件工程仅占 22.5%,AI/ML 10.5%,商业/金融/法律/HR 10.5%,营销 9.3%……论文对比了 TMax-15K、Terminal-Bench、DeepSWE 等,Skill2Env 是唯一全覆盖 13 域、且非技术知识工作占大头的语料。 忠实度探针(很聪明的设计):用任务指令+量规作查询、对 3.4k 个 SKILL.md 做 TF-IDF 检索,73.2% 的任务 top-1 命中真实源 Skill,94.6% 进 top-10(随机 0.03%)。单用量规也有 68.5% top-1,证明量规携带的是 Skill 专属方法论而非泛泛建议。 SFT 数据:用 GLM-5.3 对每个任务 rollout 两次,得到 15,968 条轨迹,平均奖励 0.74,中位轨迹 19 次模型调用 + 23 次工具调用。 S2EBench:考虑到公开基准饱和,从 SkillHub 另外生成、逐条人工审核(指令无歧义、忠实于源 Skill、测试公允)后的 79 任务私有 held-out 基准。 # RL 实验:基础设施 + 极简配方 基础设施(论文明确说“现代 agentic RL 首先是基础设施挑战”):Molt(PyTorch 原生全异步训练,Ray + vLLM + FSDP2)+ Polar(agent rollout 层:rootless Apptainer 沙箱、代理回传 token ID 和采样时 log-prob、prefix merging 把 harness 的多次补全缝合成训练轨迹)。 配方(刻意走“简单路线”):GRPO 组归一优势 + DPPO 的 binary-KL 信任域掩码(δ=0.05,超出阈值的 token 直接丢弃,无需参考模型,还能防训练-推理失配);G=8 rollouts/组,批 64,lr 1e-6 恒定,无 KL 惩罚、无熵奖励、无 SFT 热启动,每任务 65k 上下文。 量规校准奖励:开量规时,额外由 GPT-6 Astra 做 LLM-as-Judge(带“宪法”:惩罚无脑循环、reward hacking、答非所问;hacking 实证 = -5 分),总奖励 r = r_V + λs/5(λ=0.2),即 judge 最多把程序化奖励拉动 ±0.2。量规是校准可执行结果奖励,而非取代它,这是与“Rubrics as Rewards”一系的定位差异。 # 四项发现(论文最有信息量的部分) 发现 1:小规模 RL 即有跨域迁移。 仅 300 步、只用 2,400 任务子集训一个 epoch:S2EBench pass@1 +4.3(均分 +18.5),Terminal-Bench 2.1 +4.7(49.4→54.1)。训练集与 TB 无重叠(13-gram Jaccard < 0.8),且训练集从未针对 TB 调过,论文将其解读为规划、工具使用、收尾能力的通用提升而非任务族记忆。这让 27B 本地模型显著缩小了与云端前沿模型的差距。 发现 2:量规校准 RL 在基准上落后于纯结果 RL,一个诚实的负结果。 量规版在 TB 2.1 只有 50.1(纯结果版 54.1);训练中量规版的程序化奖励长期停在 0.5–0.6,judge 分项从头到尾无上升趋势,两个奖励在训练分布上互相拉扯。论文不把它当作对量规奖励的终审判决(两者优化不同目标,而基准只考结果那一半),并给出两个疑因:λ=0.2 的加性形式让失败任务仍能拿正奖励、judge 看不到文件系统等设定均未调优;以及更本质的,Skill 写下的方法论可能本来就不是最大化基准通过率的分布。 发现 3:行为确实向 Skill 对齐,量规的价值所在。 200 个任务的成对偏好测试(judge 拿源 SKILL.md 当标准,比较匿名化的 base 与 RL 轨迹):纯结果 RL 已被偏好 54.5% vs 33.5%;量规版被偏好 73.0% vs 24.0%。这说明量规奖励买到的东西在结果基准上看不见,但对“怎么做事”影响实质,对网页开发、报告综合、开放研究这类难验证任务尤其重要。 发现 4:GLM-5.3 蒸馏 SFT 反而伤害 Qwen。 在 GLM-5.3 轨迹上做 SFT:27B 上 TB 2.1 掉到 45.8;4B 上直接崩塌(TB 18.7→3.4,出现思维/工具调用死循环)。归因:教师的 interleaved-thinking + 工具调用风格与学生自身 post-training 不兼容,模仿覆盖了学生依赖的行为模式却带不来教师的能力。与 TMax 报告的“SFT 混合数据劣化已后训练的 Qwen”互相印证。因此论文所有 RL 结果都从未修改的原始 checkpoint 出发。
显示更多
Feeling the AGI/ASI, few observations 1. Opus 5.5 really gives strong vibes of AGI (maybe not fully there, but we're very close, like 80%-90%). For example, few breakthoughs in training humanoids are needed. We need to move this intelligence into physical world. 2. I don't think Anthropic has discovered any magical formula. 3. Which means, SpaceXAI, Google, and others will soon follow. This prediction is based on the amount of compute they already have + additional compute being added all the time. 4. Never was more confident that Anthropic, SpaceXAI, Google, OpenAI, likely Meta, basically all big US AGI labs will have proper AGI in 2027. 5. And from there we'll move quickly to ASI territory (likely also in 2027). It looks like few GW of compute are enough for strong AGI, now imagine what we will able to do with 10-20GW of compute which are rapidly coming online. And then with 100GW-200GW... 2027 will be an utterly insane year.
显示更多
0
350
5.4K
452
转发到社区
finds crypto you bridged but never claimed. $135M+ sits unclaimed across 84 bridges on over 500k wallets (Arbitrum, Optimism, Base, zkSync, Wormhole, CCTP...).
0
142
1.6K
169
转发到社区
继9月24日发现安全事件后,Bitget 将开始分阶段有序恢复提现功能。 此次事件所涉及的漏洞已被查明并完成修复。 Bitget 的安全与技术团队一直在对提现基础设施开展进一步的验证与安全核查,独立网络安全专家 Mandiant 与慢雾(SlowMist)也持续为调查工作提供支持。 提现功能的暂时暂停仍属于安全防范措施,与用户资产的可用性无关。用户账户余额未受任何影响,Bitget 用户保护基金(Protection Fund)将用于覆盖本次平台级事件所造成的资金影响。 安全核查完成后,提现功能将有序恢复。 当前的提现恢复时间安排如下: 9月28日08:00(UTC):BTC(Bitcoin 网络) 9月29日08:00(UTC):ETH(Ethereum、BSC、Arbitrum、Base、Optimism) 9月30日08:00(UTC):USDT(Ethereum、BSC、Solana、Tron) 10月2日08:00(UTC):其他代币/法币/C2C 我们的目标是尽快、安全地恢复所有支持资产及网络的提现服务。 该事件持续处于受控状态,不会再发生新的未经授权转账。在此过程中,用户资金始终不受影响。交易与充值功能持续正常运行。 用户在提现功能逐步开放前无需采取任何操作。提现功能的开放情况将直接在 Bitget 平台上显示,建议用户关注 Bitget 官方渠道以获取最新进展。
显示更多
0
37
38
6
转发到社区
Bitget will begin resuming withdrawals in orderly phases following the security incident identified on September 24. The vulnerability involved in the incident has been identified and remediated. Bitget's security and technical teams have since been conducting additional validation and security checks across the withdrawal infrastructure, while independent cybersecurity experts Mandiant and SlowMist continue to support the investigation. The temporary withdrawal pause remains a security measure and is not related to the availability of user assets. User account balances remain unaffected, and Bitget's Protection Fund covers the financial impact of this platform-wide incident. Withdrawals will resume in an orderly manner once these security checks are completed. The current withdrawal resumption schedule is as follows: > Sep 28, 8:00 (UTC): BTC (Bitcoin Network) > Sep 29, 8:00 (UTC): ETH (Ethereum, BSC, Arbitrum, Base, Optimism) > Sep 30, 8:00 (UTC): USDT (Ethereum, BSC, Solana, Tron) > Oct 2, 8:00 (UTC): Other Tokens / Fiat / P2P Our objective is to restore withdrawal services across all supported assets and networks as quickly and safely as possible. The incident remains contained, and no further unauthorized transfers are possible. User funds are unaffected throughout this process. Trading and deposits continue to operate. Users do not need to take any action ahead of the rollout. Withdrawal availability will be reflected directly on the Bitget platform, and users are advised to follow Bitget's official channels for updates.
显示更多
0
253
679
113
转发到社区
SpaceXAI has just introduced Grok @Bot Creator Rewards! Rewards will go out every two weeks. Rewards are based on usage: • How many people use your Bots. Are your Bots useful and popular? • How often your Bots are used. Are your Bots providing consistent value over time? For your templates to be eligible for rewards, you must: • Post about your public Bots (and specifically link to them) on X ≥ 2x per pay period • Have X Money set up • Payouts will be sent as part of X’s Original Content Rewards for now.   In the last two weeks, I’ve earned $500 in rewards due to people using my Home Robots template 🔥 More info:
显示更多
0
116
1.3K
110
转发到社区