AI user questions LLM training on core book models vs. web text

By PulseAugur Editorial · [1 sources] · 2026-06-01 16:45

A user on Mastodon is inquiring about the internal decision-making processes of large language models (LLMs). They are curious if LLMs continue to rely on a core training model derived from highly valued books, even as more web text is incorporated into successive versions. The user speculates that this foundational model might be the source of persistent problems if it contains the majority of the high-value weights. AI

IMPACT Understanding LLM training methodologies is crucial for developers and researchers aiming to improve model performance and address inherent limitations.

RANK_REASON The item is a user's question about AI training methodology, not a primary announcement or research finding.

Read on Mastodon — fosstodon.org →

LLM
Mastodon

other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

COVERAGE [1]

Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] · 2026-06-01 16:45

I’ve been wondering how these LLMs weigh their decision making when they train on many possible variations of patterns and I wonder whether they all still use a

I’ve been wondering how these LLMs weigh their decision making when they train on many possible variations of patterns and I wonder whether they all still use a core training model based on books considered high value that continues to be used in successive versions as they keep …

COVERAGE [1]

I’ve been wondering how these LLMs weigh their decision making when they train on many possible variations of patterns and I wonder whether they all still use a

RELATED ENTITIES

RELATED TOPICS