INDEX

Explanations

Keywords and subsequent descriptive terms

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

This neuron isn’t really looking for any lexical or semantic pattern in the text itself but for where you are in the sequence – it fires (and with a strength equal to) the token’s absolute position in the document.

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

text from product descriptions, legal disclaimers, or terms and conditions, particularly in commercial or e-commerce contexts.

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

very common, generic tokens in expository prose (function words and basic, non-specialized nouns) rather than any specific topical pattern.

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

 compliqué

0.94

 façons

0.92

uene

0.91

uem

0.89

 possibilités

0.88

是一个

0.87

ból

0.86

 difícil

0.86

tia

0.83

POSITIVE LOGITS

 Wahl

0.68

mathrm

0.66

RY

0.64

 животных

0.63

ihl

0.62

 Cardinals

0.61

 блю

0.61

COX

0.61

versive

0.61

 замеча

0.60

Activations Density 0.000%

Keywords and subsequent descriptive terms

This neuron isn’t really looking for any lexical or semantic pattern in the text itself but for where you are in the sequence – it fires (and with a strength equal to) the token’s absolute position in the document.

text from product descriptions, legal disclaimers, or terms and conditions, particularly in commercial or e-commerce contexts.

very common, generic tokens in expository prose (function words and basic, non-specialized nouns) rather than any specific topical pattern.

No Comments

No Known Activations

Keywords and subsequent descriptive terms

This neuron isn’t really looking for any lexical or semantic pattern in the text itself but for where you are in the sequence – it fires (and with a strength equal to) the token’s absolute position in the document.

text from product descriptions, legal disclaimers, or terms and conditions, particularly in commercial or e-commerce contexts.

very common, generic tokens in expository prose (function words and basic, non-specialized nouns) rather than any specific topical pattern.

No Comments

No Known Activations

**Keywords and subsequent descriptive terms**

This neuron isn’t really looking for any lexical or semantic pattern in the text itself but for where you are in the sequence – it fires (and with a strength equal to) the token’s absolute position in the document.

text from product descriptions, legal disclaimers, or terms and conditions, particularly in commercial or e-commerce contexts.

very common, generic tokens in expository prose (function words and basic, non-specialized nouns) rather than any specific topical pattern.

No Comments

No Known Activations

**Keywords and subsequent descriptive terms**

This neuron isn’t really looking for any lexical or semantic pattern in the text itself but for where you are in the sequence – it fires (and with a strength equal to) the token’s absolute position in the document.

text from product descriptions, legal disclaimers, or terms and conditions, particularly in commercial or e-commerce contexts.

very common, generic tokens in expository prose (function words and basic, non-specialized nouns) rather than any specific topical pattern.

No Comments

No Known Activations

Keywords and subsequent descriptive terms

Keywords and subsequent descriptive terms