INDEX

Explanations

the word "berg" with varying levels of specificity

oai_token-act-pair · gpt-3.5-turbo

occurrences of the suffix "berg"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

FUL

-0.72

teenth

-0.71

BOOK

-0.67

fully

-0.65

 ease

-0.65

pee

-0.65

pent

-0.63

 referen

-0.63

perature

-0.62

lled

-0.61

POSITIVE LOGITS

enegger

1.11

hetti

1.06

sburgh

1.03

stein

1.00

ersen

0.97

rade

0.91

0.90

sf

0.88

dor

0.87

roth

0.86

Activations Density 0.054%

the word "berg" with varying levels of specificity

occurrences of the suffix "berg"

No Comments

No Known Activations