INDEX

Explanations

instances of confessions or admissions of guilt

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

confessing or admitting something

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

DockStyle

-0.60

AnimationsModule

-0.59

 Ducks

-0.54

 Waterman

-0.54

Shea

-0.53

GraphicsUnit

-0.53

SRM

-0.52

榄

-0.51

RectangleBorder

-0.51

Lucas

-0.50

POSITIVE LOGITS

 confesses

0.97

 confess

0.96

 confession

0.93

 Confession

0.92

 confessed

0.88

 confessions

0.86

 Confessions

0.81

 confes

0.79

告白

0.63

 CONF

0.52

Activations Density 0.004%

instances of confessions or admissions of guilt

confessing or admitting something

No Comments

No Known Activations