INDEX

Explanations

acronyms or abbreviations with the characters "ud" and a numeric value, possibly related to military or governmental contexts

oai_token-act-pair · gpt-3.5-turbo

references to specific individuals or entities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

#$

-0.78

xon

-0.71

borough

-0.68

hazard

-0.60

 combust

-0.58

 Painter

-0.58

 knives

-0.58

 hypert

-0.57

 regress

-0.56

 diseng

-0.56

POSITIVE LOGITS

geon

1.45

geoning

1.27

imentary

1.24

apest

1.22

geons

1.09

elta

1.07

ges

1.05

iamond

1.04

ible

1.02

olph

1.01

Activations Density 0.022%

acronyms or abbreviations with the characters "ud" and a numeric value, possibly related to military or governmental contexts

references to specific individuals or entities

No Comments

No Known Activations