INDEX

Explanations

percentage comparisons in text

oai_token-act-pair · gpt-3.5-turbo

phrases indicating comparisons or relationships between different items or groups

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

gravity

-0.79

rog

-0.77

rak

-0.72

ger

-0.71

 Defenders

-0.70

azaar

-0.69

spe

-0.67

talk

-0.67

gery

-0.66

wordpress

-0.66

POSITIVE LOGITS

 eleph

0.83

 sexes

0.77

 sidx

0.77

 guiActiveUn

0.77

 weep

0.74

 proport

0.73

 nomine

0.71

 halves

0.69

nces

0.69

çĶ°

0.67

Activations Density 0.009%

percentage comparisons in text

phrases indicating comparisons or relationships between different items or groups

No Comments

No Known Activations