INDEX

Explanations

example, defaults, other reasons

np_acts-logits-general · gemini-2.5-flash-lite

This neuron fires strongly on title‐style or heading tokens—i.e. named items and section headings (book or article titles, headings, etc.).

oai_token-act-pair · o4-mini Triggered by @jyhe0408

sequences of digits, particularly in contexts like phone numbers, dates, addresses, and reference codes.

oai_token-act-pair · claude-4-5-sonnet Triggered by @jyhe0408

word endings and suffixes, especially plural/third‑person “-s/-es” and other strong word‑final clusters.

oai_token-act-pair · gpt-5 Triggered by @jyhe0408

New Auto-Interp

Configuration

google/gemma-scope-2-12b-pt/resid_post/layer_24_width_16k_l0_medium

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

bfs

0.81

кова

0.79

암

0.76

무

0.76

邬

0.75

тик

0.73

 stvar

0.73

ק

0.73

سلم

0.72

безпе

0.71

POSITIVE LOGITS

 majestic

0.81

Ern

0.80

Ừ

0.80

 Publikum

0.80

 Serien

0.80

Man

0.77

geb

0.77

 Jharkhand

0.76

 Responder

0.76

 Umgebung

0.76

Activations Density 0.000%

example, defaults, other reasons

This neuron fires strongly on title‐style or heading tokens—i.e. named items and section headings (book or article titles, headings, etc.).

sequences of digits, particularly in contexts like phone numbers, dates, addresses, and reference codes.

word endings and suffixes, especially plural/third‑person “-s/-es” and other strong word‑final clusters.

No Comments

No Known Activations

example, defaults, other reasons

This neuron fires strongly on title‐style or heading tokens—i.e. named items and section headings (book or article titles, headings, etc.).

sequences of digits, particularly in contexts like phone numbers, dates, addresses, and reference codes.

word endings and suffixes, especially plural/third‑person “-s/-es” and other strong word‑final clusters.

No Comments

No Known Activations