INDEX

Explanations

results or resultado

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

The neuron fires on tokens related to “result” (e.g. “result,” “results,” “resultado,” etc.).

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

mentions of the “outcome” lexical stem across contexts, including inflections, misspellings, other languages, and within code identifiers.

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

the word "result" or its variants in multiple languages (like "resultado" in Spanish/Portuguese).

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

New Auto-Interp

Configuration

google/gemma-scope-27b-pt-res/layer_10/width_131k

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 включает

-1.92

},

-1.89

、

-1.87

 Warum

-1.84

꒦

-1.82

 знали

-1.78

 jako

-1.78

消失

-1.75

jooq

-1.75

You

-1.73

POSITIVE LOGITS

時から

1.98

攤

1.86

ar

1.84

 montado

1.79

人們

1.78

 results

1.74

 stocker

1.73

 reinigen

1.72

イナ

1.70

 становятся