INDEX

Explanations

negative statements or warnings

oai_token-act-pair · gpt-3.5-turbo

phrases indicating the distinction between capability and moral or practical obligation

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

ahime

-0.78

 throats

-0.71

ebted

-0.70

 bottled

-0.63

 energies

-0.61

 matured

-0.60

 administered

-0.60

hel

-0.59

ãĥ¼ãĤ¯

-0.58

 chairs

-0.58

POSITIVE LOGITS

soType

0.89

cause

0.79

 disqual

0.77

¿½

0.76

advertising

0.76

 necessarily

0.75

Cause

0.75

ECA

0.74

 automatically

0.74

etheless

0.73

Activations Density 0.154%

negative statements or warnings

phrases indicating the distinction between capability and moral or practical obligation

No Comments

No Known Activations