INDEX

Explanations

the

np_max-act-logits · gemini-2.0-flash

metadata or system-level instructions in text, particularly related to AI language models and configuration details.

oai_token-act-pair · claude-3-5-haiku-20241022 Triggered by @kdearsty

the

np_max-act · claude-3-5-haiku-20241022 Triggered by @kdearsty

definite article "the" followed by nouns, especially in question contexts.

oai_token-act-pair · claude-3-7-sonnet-20250219 Triggered by @jacktpayne51

New Auto-Interp

Configuration

andyrdt/saes-gpt-oss-20b/resid_post_layer_15/trainer_0

Dataset (Dashboard)

Various

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

hurt

-0.09

 nghe

-0.08

_it

-0.08

fällen

-0.08

(delta

-0.08

chè

-0.08

 này

-0.08

YO

-0.07

 την

-0.07

την

-0.07

POSITIVE LOGITS

 rationale

0.12

 bedoeling

0.11

 gist

0.11

 उद्देश्य

0.11

 verdict

0.10

 significance

0.10

 तरीका

0.10

 أبرز

0.10

 takeaway

0.10

 exactement

0.10

Activations Density 0.021%

the

metadata or system-level instructions in text, particularly related to AI language models and configuration details.

the

definite article "the" followed by nouns, especially in question contexts.

No Comments

No Known Activations