INDEX

Explanations

capitalized terms or acronyms likely relating to specific entities or concepts

oai_token-act-pair · gpt-3.5-turbo

references to the abbreviation "AL"

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

olicy

-0.77

ãĥ¼ãĥĨãĤ£

-0.74

æĪ¦

-0.71

 sterling

-0.70

ãģį

-0.70

 Chaser

-0.68

ledged

-0.65

ãģ¯

-0.64

NT

-0.62

åĤ

-0.62

POSITIVE LOGITS

manac

1.33

gorithm

1.31

gorith

1.26

gebra

1.22

ibaba

1.20

chemy

1.19

umni

1.17

abama

1.14

aska

1.12

tering

1.09

Activations Density 0.040%

capitalized terms or acronyms likely relating to specific entities or concepts

references to the abbreviation "AL"

No Comments

No Known Activations