INDEX

Explanations

the substring "ab" followed by a single-digit activation value

oai_token-act-pair · gpt-3.5-turbo

occurrences of the abbreviation "ab" in the text

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

 virtue

-0.80

 nomine

-0.74

 Perse

-0.69

 Coco

-0.65

bilt

-0.65

 Celest

-0.63

backer

-0.61

 Uriel

-0.60

 Izan

-0.60

 Patriot

-0.59

POSITIVE LOGITS

stract

1.34

bing

1.11

yrinth

1.10

urger

1.09

raham

1.04

road

1.03

dullah

1.03

bed

1.00

ecause

0.98

riel

0.96

Activations Density 0.038%

the substring "ab" followed by a single-digit activation value

occurrences of the abbreviation "ab" in the text

No Comments

No Known Activations