INDEX

Explanations

phrases related to specific individuals or titles

oai_token-act-pair · gpt-3.5-turbo

mentions of Pope Benedict and related figures

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

EAR

-0.82

MAG

-0.81

ODUCT

-0.79

perse

-0.74

ãĥ£

-0.73

PLAY

-0.73

Agent

-0.72

claimer

-0.72

wash

-0.69

place

-0.68

POSITIVE LOGITS

 Cumber

1.35

 Benedict

1.28

XVI

1.22

itial

0.94

icum

0.94

itude

0.82

umen

0.80

shire

0.79

ric

0.79

icts

0.76

Activations Density 0.016%

phrases related to specific individuals or titles

mentions of Pope Benedict and related figures

No Comments

No Known Activations