INDEX

Explanations

The neuron strongly activates on tokens containing non‐standard or accented characters (e.g. “purée,” “UNESCO,” “emotionally,” “cabezales”), so it’s looking for subword pieces with unusual or non-ASCII letter forms.

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

subword tokens that complete words in non-English languages or proper nouns.

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

adjectival or adverbial modifiers, especially past-participle descriptors used within sentences.

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

ﺔ

0.78

咏

0.76

CHEMICAL

0.74

цах

0.74

ления

0.73

 sprouted

0.72

秫

0.72

 disputa

0.72

нения

0.71

伸

0.71

POSITIVE LOGITS

şi

0.71

antial

0.71

 กว่า

0.70

isyon

0.70

 voire

0.69

 これは

0.68

pas

0.67

vector

0.66

rainer

0.66

Ι

0.65

Activations Density 0.001%

The neuron strongly activates on tokens containing non‐standard or accented characters (e.g. “purée,” “UNESCO,” “emotionally,” “cabezales”), so it’s looking for subword pieces with unusual or non-ASCII letter forms.

subword tokens that complete words in non-English languages or proper nouns.

adjectival or adverbial modifiers, especially past-participle descriptors used within sentences.

No Comments

No Known Activations

The neuron strongly activates on tokens containing non‐standard or accented characters (e.g. “purée,” “UNESCO,” “emotionally,” “cabezales”), so it’s looking for subword pieces with unusual or non-ASCII letter forms.

subword tokens that complete words in non-English languages or proper nouns.

adjectival or adverbial modifiers, especially past-participle descriptors used within sentences.

No Comments

No Known Activations

The neuron strongly activates on tokens containing non‐standard or accented characters (e.g. “purée,” “UNESCO,” “emotional­ly,” “cabez­ales”), so it’s looking for subword pieces with unusual or non-ASCII letter forms.

subword tokens that complete words in non-English languages or proper nouns.

adjectival or adverbial modifiers, especially past-participle descriptors used within sentences.

No Comments

No Known Activations

The neuron strongly activates on tokens containing non‐standard or accented characters (e.g. “purée,” “UNESCO,” “emotional­ly,” “cabez­ales”), so it’s looking for subword pieces with unusual or non-ASCII letter forms.

subword tokens that complete words in non-English languages or proper nouns.

adjectival or adverbial modifiers, especially past-participle descriptors used within sentences.

No Comments

No Known Activations

The neuron strongly activates on tokens containing non‐standard or accented characters (e.g. “purée,” “UNESCO,” “emotionally,” “cabezales”), so it’s looking for subword pieces with unusual or non-ASCII letter forms.

The neuron strongly activates on tokens containing non‐standard or accented characters (e.g. “purée,” “UNESCO,” “emotionally,” “cabezales”), so it’s looking for subword pieces with unusual or non-ASCII letter forms.