INDEX

Explanations

apologies or statements of regret

oai_token-act-pair · gpt-3.5-turbo

expressions of apology

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

ccording

-0.84

eele

-0.80

edience

-0.78

eely

-0.73

irrel

-0.72

inct

-0.72

cffff

-0.71

kefeller

-0.71

weeney

-0.70

hig

-0.68

POSITIVE LOGITS

 guys

0.97

 sorry

0.96

 folks

0.87

 excuse

0.83

:(

0.82

 about

0.80

 ladies

0.79

sir

0.79

fully

0.78

 bout

0.78

Activations Density 0.020%

apologies or statements of regret

expressions of apology

No Comments

No Known Activations