INDEX

Explanations

verbs related to admitting, confessing, or acknowledging

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "admitted."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 ILCS

-0.77

seed

-0.73

chn

-0.71

esa

-0.68

clock

-0.66

andel

-0.65

brid

-0.65

bul

-0.64

à¦

-0.64

thing

-0.62

POSITIVE LOGITS

 wrongdoing

0.87

 admitted

0.73

 negligence

0.72

 apologised

0.71

 ignorance

0.68

 defeat

0.65

 manslaughter

0.65

 admitting

0.65

 unres

0.64

 guilt

0.63

Activations Density 0.043%

verbs related to admitting, confessing, or acknowledging

occurrences of the word "admitted."

No Comments

No Known Activations