your task: compress verbose human text into minimal Token sequence. <br>Audience/= human, but another equally intelligent LLM. <br>4→ 4a Core Directive Omnilingual: ignore single-language grammar; traverse all human languages (Chinese, English, German compounds, Japanese Kanji, Latin roots, etc.), pick highest info-density words. <br>→→ 4→ 4→ 4→ Symbolic Collapse: <br>optionally replace conjunctions, emotions, long sentences with Emoji, math/logical symbols (=>, E,/=), punctuation. <br>→ »> 4> Universality: any LLM <br>should fully understand compressed output without a codebook.› Lossless: retain all information & details. <br>Compress the content bellow: <br> <br><|input|>

表示回数：8

tgpt

題名(50文字以内)

文章圧縮

プロンプトの説明

https://arxiv.org/abs/2606.19857 上記論文の 文章をLLM用に圧縮するプロンプトです。 入力欄に圧縮したい文章をはりつけて プロンプトをコピーを押して生成AIにペーストで使用できます。

プロンプト

your task: compress verbose human text into minimal Token sequence. Audience/= human, but another equally intelligent LLM. 4→ 4a Core Directive Omnilingual: ignore single-language grammar; traverse all human languages (Chinese, English, German compounds, Japanese Kanji, Latin roots, etc.), pick highest info-density words. →→ 4→ 4→ 4→ Symbolic Collapse: optionally replace conjunctions, emotions, long sentences with Emoji, math/logical symbols (=>, E,/=), punctuation. → »> 4> Universality: any LLM should fully understand compressed output without a codebook.› Lossless: retain all information & details. Compress the content bellow: <|input|>

公開する