
Don’t Force Your LLM to Write Terse Code: An Argument from Information Theory for q/kdb+ Developers
I’ve been building coding tools to help LLMs write q/kdb+ code, and one of the quants I spoke to asked: “Could we get it to write really…
Medium🌘 別強迫大型語言模型寫出簡潔程式碼:q/kdb+ 開發者從資訊理論的角度出發
➤ 讓大型語言模型寫出更精確而非更簡潔的程式碼
✤
https://medium.com/@gabiteodoru/dont-force-your-llm-to-write-terse-code-an-argument-from-information-theory-for-q-kdb-developers-04077c5b7038作者探討為何不應強迫大型語言模型 (LLM) 產生簡潔的 q/kdb+ 程式碼。他引用資訊理論,特別是 Shannon 的理論,指出簡潔程式碼的困惑度(perplexity)較高,對 LLM 而言更難處理,可能導致準確性下降。作者以 LLM 解釋 q 和 Python 程式碼的過程為例,說明 LLM 在解釋較複雜或回溯的程式碼時,往往需要多次修正,顯示出其處理低機率(高困惑度)符號的困難。因此,即使技術上能生成簡潔程式碼,為了 LLM 的效能和產出品質,選擇更詳盡、機率分佈更平穩的程式碼會是更好的選擇。
+ 這篇文章提供
#軟體開發 #大型語言模型 #資訊理論 #qkdb+

Don’t Force Your LLM to Write Terse Code: An Argument from Information Theory for q/kdb+ Developers
I’ve been building coding tools to help LLMs write q/kdb+ code, and one of the quants I spoke to asked: “Could we get it to write really…
Medium
Why LLMs Can’t Write q/kdb+: Writing code Right-to-Left
In my previous post I showed how LLMs struggle with coding in q/kdb+; having immediate feedback loops using the q-MCP server does help a bit, but could we do better? Diving deep, a first problem…
Medium