A Mathematical Framework for Transformer Circuits (2021)
Two-Layer Attention-Only Transformers Three Kinds of Composition Path Expansion of Logits Path Expansion of Attention Scores QK Circuit Analyzing a Two-Layer Model Induction Heads Term Importance Analysis Virtual Attention Heads Where Does This Leave Us? However, as these models scale, their open-endedness and high capacity creates an increasing scope for unexpected and sometimes harmful behaviors. Even years after a large model is trained, both creators and users routinely discover model capabilities – including problematic behaviors – they were previously unaware of.
- ▪Two-Layer Attention-Only Transformers Three Kinds of Composition Path Expansion of Logits Path Expansion of Attention Scores QK Circuit Analyzing a Two-Layer Model Induction Heads Term Importance Analysis Virtual Attention Heads Where Does
- ▪However, as these models scale, their open-endedness and high capacity creates an increasing scope for unexpected and sometimes harmful behaviors.
- ▪Even years after a large model is trained, both creators and users routinely discover model capabilities – including problematic behaviors – they were previously unaware of.
Hacker News (Front Page) files mainly under programming. We currently carry 1,521 of its stories. Top-voted stories on Hacker News.
Story provenance
Source · retrieval · rights · ranking — open for full record
inspect →
Story provenance
Attribution is not the same as permission. This drawer separates discovery metadata, excerpts, WeSearch-generated summaries, reuse status, and whether the publisher receives the visit. Nothing here claims a legal grant the publisher has not made.
Record
| Original publisher | Transformer-circuits |
| Canonical URL | https://transformer-circuits.pub/2021/framework/index.html |
| Publication time | Sat, 12 Sep 2026 13:56:58 +0000 |
| Retrieval time | 2026-09-12T16:45:11.138Z |
| Last seen | 2026-09-12T16:45:26.621Z |
| Headline source | Publisher (no WeSearch rewrite) |
| Excerpt source | publisher body |
| Excerpt method | First ~120 words (~800 chars) of extracted publisher body, fair-use limited. |
| Summary | WeSearch · cerebras-chat (WeSearch summarizer) |
| Summary source text | contentText |
| Citation coverage | Summary is a WeSearch-generated derivative; primary citation is the original publisher URL. |
| Cluster | UKxTlIH5yXAu · 1 stories |
| Cluster logic | Grouped by semantic title/content similarity across sources within a rolling window. Same-publisher template collisions are excluded from coverage comparison. |
| Ranking reason | Story pages are not engagement-ranked. Hub feeds use recency, with optional source-diversified chronological ordering (cap consecutive stories per source). No personalized ranking. |
| Publisher visit | Yes — open original |
| Substitutes article? | No — link-out required for full text |
Rights status (four layers)
WeSearch handling by dimension
| Indexing | May the item be indexed (stored, ranked, made findable)? | Allowed |
| Snippet | May a short excerpt of the publisher's text be shown? | Allowed |
| AI summary | May WeSearch generate its own short summary of the article? | Limited |
| Retrieval / RAG | May the content be exposed for third-party retrieval-augmented generation? | Not asserted |
| Model training | May the content be used to train AI models? | Not asserted |
| Commercial reuse | May the content be reused commercially? | Not permitted |
Basis: Derived from the published RSS/Atom feed. Contact: [email protected]. Reviewed: 2026-07-24.
Opening excerpt (first ~120 words) tap to expand
h1 a:hover, h2 a:hover, h3 a, h3 a:hover { text-decoration: none; border-bottom: none; } .comment { background-color: hsl(54, 78%, 96%); border-left: solid hsl(54, 33%, 67%) 1px; padding: 1em; color: hsla(0, 0%, 0%, 0.67); margin-top: 1em; } .comment h3 { font-size: 100%; font-weight: bold; text-transform: uppercase; margin-top: 0px; } .comment .commenter-description { font-style: italic; margin-bottom: 1em; margin-top: 0px; } .comment .commenter-description, .comment .commenter-description a { color: #777; font-size: 80%; line-height: 160%; } .comment div { margin-top: 1em; } .comment code { background: #FFF5; border-color: #EEEA; } .evidence-table { display: grid; grid-template-columns: auto 1fr 1fr 1fr; grid-gap: 4px; margin: 40px 0px 40px 0px; width: 100%; min-width: 700px;…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Transformer-circuits.
Discussion
0 commentsMore from Transformer-circuits
