electron

[electron @electro86153277]

dsh-ATN 让 DeepSeek Harness 的智能体合作更精准稳定，减低长程工作时的幻觉。
Apache-2.0 开源。
https://t.co/jADcSRZma8
#MultiAgent #DeepSeekHarness #DeepSeek #plugin #opensource

[electron @electro86153277]

白驹过隙，忽然而已。

[Max For AI @MaxForAI]

你醒了？

现在是2022年11月30日，你在刷抖音的过程中发现有一个叫@ChatGPT 的聊天机器人很火。

于是你跃跃欲试去做了尝试 它的强大十分令人激动，你与他的交谈使你非常兴奋。

过了几个月后，你发现ChatGPT推出了plus套餐 而免费用户的额度完全不足够你去使用。

你开始去寻找各种各样的镜像站与免费，APIKey 每当发现一个能用的 你都如获珍宝般的珍藏。

可惜时间总是过得很快。

如今，你用着强大的Astra与Fable系列模型 享受着AI给你带来的无限乐趣

2026年10月23日，最初让你感到非常兴奋GPT-3.5模型在那天下线了。

而那天你和平常以���似乎没什么区别，OpenAI的@thsottiaux 在玩雷霆大重置、A社@claudeai 在玩开心消消乐、谷歌@GeminiApp 还在没完没了的造屎。

你依旧每天都要打开Codex去输入你的idea。

AI给你带来的那头一份喜悦，就像你的青春 不经意间回头看了你一眼，好像在和你告别。

[electron @electro86153277]

@snwiki238337 deepseek harness有一个插件叫dsh-ATN，解决思路也是隔绝agent间的公共广播，引入拓扑网络让agent通过拓扑结构去通信。

[electron @electro86153277]

@Pyuyi233 学姐勾引我🥰🥰

[electron @electro86153277]

@haoooo_zhang @SiferTimya 来我宿舍

[electron @electro86153277]

@SirHumphrey1908 我哭了

[electron @electro86153277]

@waylybaye 喜欢miuix

[electron @electro86153277]

@Pyuyi233 opus 5.5带来了新的范式。
也许以后可以先用image模型生成几张精美的设定图，具体实操时先把图片转为SVG，然后让模型通过修改SVG制作动漫？
btw，astra通过图片编写SVG时确实很还原。

[Zane @zane_os]

对于最近降速之后推出GPT-6.1 Sol 的 Ultrafast 模式和把算力全部借走研究数学题的行为

今天我们宣布推出新的Tibo

她在这里：

[electron @electro86153277]

@Pyuyi233

[electron @electro86153277]

@Pyuyi233 来了

[electron @electro86153277]

@SeeleVollerei_A 百团来抓你

[electron @electro86153277]

@0xfufu_d667 生活是��幻泡影吗😭

[electron @electro86153277]

@WuGuangYao 没有人类了

[electron @electro86153277]

@letdarky Hey bro u need opencode go subscription, I think China models like DeepSeek is perfect for 10 dollars user

[electron @electro86153277]

@Pyuyi233 😭

[electron @electro86153277]

@SirHumphrey1908 电子签名可以用WPS / MS office制作PDF扫描件，自动去背景

[electron @electro86153277]

@Chen_LingShui ohno妈咪

[洛天依(Luo Tianyi) @hn_luotianyi712]

Happy Mid-Autumn Festival 🌕🩵🌸🥮

[Fuli Luo @_LuoFuli]

MiMo-V3 is getting a new architecture. The core of it, HySparse2, is out today.

Less prefill, a smaller KV cache, better long-context retrieval—and we got all three at once.

Compared with MiMo-V2.6's Hybrid SWA architecture:
• 5.02× lower prefill FLOPs at 1M tokens
• 4.5× smaller KV cache at 1M tokens
• Better MRCRv2 and RULER-v2 scores, plus lower AgentPPL and LongPPL

Why build a new architecture?
Agentic inference is a very different workload. Each round, a short action can return a long observation that needs to be prefilled, while the context keeps growing. That puts prefill cost, KV-cache size, and retrieval accuracy on the critical path at the same time.

HySparse2 tackles all three with two levels of KV sharing:
• KV Bridging: Following YOCO, full-attention layers in the cross-decoder build their K/V from self-decoder hidden states.
• KV Reuse: Within each hybrid block, sparse layers reuse the preceding full-attention layer's KV cache and selection indices.

Two more changes: token-level selection replaces block-level selection, and a forced window of recent tokens replaces the separate SWA branch, so local and global tokens share one KV cache. Since all cross-decoder KV caches now come from the self-decoder, prefill can stop once the self-decoder finishes.

Paper: https://t.co/REeEdd5pL7