An Altar for the Dead: The people who built your computer are disappearing from AI. I lit 85 candles.

Описание и применение

Автор создал веб‑альтар Still Spoken, чтобы показать, как модели забывают исторических деятелей информатики, и восстановил их память, обучив небольшую LoRA‑адаптер на Gemma 3 1B, что повысило точность до 74 из 77 записанных фигур, но снизило точность на незаписанных знаменитых людей.

Задача

Large language models forget historical contributors to computing, leading to inaccurate or fabricated biographical information.

Результат

Fine‑tuning a small LoRA adapter restored accurate recall for most written historical figures but introduced trade‑offs, such as reduced accuracy on unseen famous people and occasional misattributions.

Рабочий процесс

• Select 85 deceased computing contributors and verify facts via Wikipedia/Wikidata.
• Create a prompt set to test model recall of these figures.
• Fine‑tune a 52 MB LoRA adapter on Gemma 3 1B for 21 minutes on a 6 GB GPU.
• Evaluate recall before and after fine‑tuning using held‑out tests.
• Iterate training to correct misattributions (e.g., living programmers receiving death dates).

Доказательства

Asked about 85 people who built computing, Gemma 3 1B could place 14.

Источник

This repo measures that forgetting with held‑out tests, then writes 77 of those people back into the model's weights with a 52 MB LoRA adapter trained on a 6 GB laptop GPU.

Источник

Recall of the written people goes from 6 to 74 of 77. Recognition of famous people who were never written falls from 84% to 51%.

Источник

It only partly works, and the failures are the interesting part.

Источник