INDEX

Explanations

the word "so" at high activation levels

oai_token-act-pair · gpt-3.5-turbo

phrases indicating a comparison or similarity

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

ãĥ¯ãĥ³

-0.66

edom

-0.64

andise

-0.62

dayName

-0.61

erness

-0.58

break

-0.57

intend

-0.56

efe

-0.56

blocking

-0.54

look

-0.52

POSITIVE LOGITS

oner

1.19

apy

1.12

bered

1.05

oths

1.04

far

1.02

 forth

1.00

othe

0.98

othes

0.87

 does

0.86

ooo

0.83

Activations Density 0.059%

the word "so" at high activation levels

phrases indicating a comparison or similarity

No Comments

No Known Activations