INDEX

Explanations

references to heroic figures or characters

oai_token-act-pair · gpt-3.5-turbo

references to heroes in various contexts

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

aeda

-0.90

orie

-0.85

ntil

-0.84

ateur

-0.82

Ã±o

-0.78

rupt

-0.76

aton

-0.76

igree

-0.73

imentary

-0.73

vant

-0.72

POSITIVE LOGITS

 heroes

0.93

 heroine

0.83

ku

0.77

 Reborn

0.75

 hero

0.74

ically

0.74

acters

0.69

 Saur

0.65

rities

0.65

 Pengu

0.64

Activations Density 0.012%

references to heroic figures or characters

references to heroes in various contexts

No Comments

No Known Activations