INDEX

Explanations

the word "rebel" at various instances

oai_token-act-pair · gpt-3.5-turbo

references to rebel groups and their activities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ographies

-0.85

thora

-0.82

gow

-0.79

orgetown

-0.78

 Robo

-0.77

Safety

-0.75

 Atmospheric

-0.75

Arthur

-0.75

ostics

-0.74

 Compton

-0.72

POSITIVE LOGITS

 rebels

1.08

 factions

1.06

 milit

1.05

 rebel

1.03

 militias

0.98

 shelling

0.95

 faction

0.94

 uprising

0.92

 army

0.90

 regime

0.89

Activations Density 0.017%

the word "rebel" at various instances

references to rebel groups and their activities

No Comments

No Known Activations