INDEX

Explanations

terms related to reports or journalism

oai_token-act-pair · gpt-3.5-turbo

the repeated occurrence of the word "report."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

cffff

-0.83

conservancy

-0.80

othing

-0.76

creen

-0.75

 Klux

-0.72

 stead

-0.69

etheless

-0.66

ategory

-0.66

inh

-0.66

 leisure

-0.65

POSITIVE LOGITS

report

1.13

 report

1.00

Report

1.00

books

0.89

 Report

0.88

 reports

0.88

reports

0.80

imony

0.78

 REPORT

0.75

Crash

0.74

Activations Density 0.039%

terms related to reports or journalism

the repeated occurrence of the word "report."

No Comments

No Known Activations