INDEX

Explanations

well-formed syntactic structures or functional blocks in written content

oai_token-act-pair · gpt-4o-mini Triggered by @bot

Tokens followed by non-alphanumeric characters

np_acts-logits-general · gemini-2.0-flash

reflexive pronouns and titles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 20-gemmascope-res-16k

Configuration

google/gemma-scope-2b-pt-res/layer_20/width_16k/average_l0_71

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

im

-0.72

ta

-0.69

int

-0.68

sa

-0.65

du

-0.63

<eos>

-0.63

-0.59

ma

-0.59

 call

-0.58

ch

-0.58

POSITIVE LOGITS

 itſelf

1.16

 myſelf

1.14

 Monfieur

1.04

 themſelves

1.02

1.01

$)$

0.99

 himſelf

0.96

UnusedPrivate

0.96

 ་་

0.93

 uſed

0.91

Activations Density 0.039%

well-formed syntactic structures or functional blocks in written content

Tokens followed by non-alphanumeric characters

reflexive pronouns and titles

No Comments

No Known Activations