INDEX

Explanations

names of things and their status

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

The neuron fires on Spanish (and closely related) words used as metadata labels or section headings (e.g. “Data,” “Enlaces,” “número,” “departamentos”).

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

non-English text, particularly Spanish and Italian.

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

non-English or foreign-language words, often with diacritics or from multiple languages.

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

New Auto-Interp

Configuration

google/gemma-scope-2-12b-pt/resid_post/layer_31_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

gete

0.54

Info

0.51

PIP

0.50

 animados

0.50

อิ

0.49

ומי

0.49

Assumption

0.49

 leves

0.49

 regulares

0.48

POSITIVE LOGITS

 каждого

0.88

 của

0.84

 tohoto

0.80

 основного

0.80

 файла

0.77

ของการ

0.77

នៃការ

0.76

 všech

0.76

នៃ

0.74

 των

0.74

Activations Density 0.033%

names of things and their status

The neuron fires on Spanish (and closely related) words used as metadata labels or section headings (e.g. “Data,” “Enlaces,” “número,” “departamentos”).

non-English text, particularly Spanish and Italian.

non-English or foreign-language words, often with diacritics or from multiple languages.

No Comments

No Known Activations

names of things and their status

The neuron fires on Spanish (and closely related) words used as metadata labels or section headings (e.g. “Data,” “Enlaces,” “número,” “departamentos”).

non-English text, particularly Spanish and Italian.

non-English or foreign-language words, often with diacritics or from multiple languages.

No Comments

No Known Activations