INDEX

Explanations

hashtag followed by topic

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

The neuron flags content-bearing “keyword” tokens—rare or topic-specific words (proper nouns, locations, domain terms) rather than common function words.

oai_token-act-pair · o4-mini

Triggered by @jyhe0408

words or tokens that are part of geographic locations, particularly cities and places.

oai_token-act-pair · claude-4-5-sonnet

Triggered by @jyhe0408

terms indicating locations or community/group contexts (places, directions, platforms, or audiences).

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

Negative Logits

 території

0.87

quando

0.86

ması

0.85

ﺌ

0.82



0.82

்களில்

0.82

avali

0.81

 cili

0.80

ienti

0.80

rils

0.79

POSITIVE LOGITS

м

0.96

م

0.85

 meet

0.81

 lotta

0.79

 grun

0.77

0.75

磨

0.70

row

0.68

쿱

0.67

 партне

0.66

Activations Density 0.000%

hashtag followed by topic

The neuron flags content-bearing “keyword” tokens—rare or topic-specific words (proper nouns, locations, domain terms) rather than common function words.

words or tokens that are part of geographic locations, particularly cities and places.

terms indicating locations or community/group contexts (places, directions, platforms, or audiences).

No Known Activations

hashtag followed by topic

The neuron flags content-bearing “keyword” tokens—rare or topic-specific words (proper nouns, locations, domain terms) rather than common function words.

words or tokens that are part of geographic locations, particularly cities and places.

terms indicating locations or community/group contexts (places, directions, platforms, or audiences).

No Known Activations