INDEX

Explanations

code-related terms, particularly those involving numbers

oai_token-act-pair · gemini-2.0-flash

self-referential pronouns

np_token-act-pair-logits · gpt-4o-mini

say "="

np_max-act-logits · gemini-2.0-flash

New Auto-Interp

Configuration

google/gemma-scope-2b-pt-transcoders/layer_21/width_16k/average_l0_13

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.21.ln2.hook_normalized

Architecture

jumprelu_transcoder

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

add

-0.54

med

-0.54

alt

-0.53

zz

-0.52

ch

-0.51

aux

-0.51

dat

-0.50

nier

-0.50

POSITIVE LOGITS

 itſelf

0.96

 myſelf

0.95

AndEndTag

0.94

 raiſ

0.85

rrggbb

0.83

 Monfieur

0.83

amaño

0.82

 ſever

0.82

 juſt

0.81

 faſt

0.81

Activations Density 3.562%

code-related terms, particularly those involving numbers

self-referential pronouns

say "="

No Comments

No Known Activations