INDEX

Explanations

words related to valuable objects or items

oai_token-act-pair · gpt-3.5-turbo

terms related to artifacts and relics

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 Miche

-0.78

NSW

-0.76

izon

-0.74

lip

-0.71

WAY

-0.69

 Parent

-0.69

 Bunny

-0.66

 Queensland

-0.66

abet

-0.66

 Junior

-0.65

POSITIVE LOGITS

 artifacts

3.85

 artifact

3.35

 arte

2.82

artifacts

2.76

ifacts

2.20

 Artifact

2.06

 relics

1.79

 relic

1.53

 archaeological

1.21

 archae

1.17

Activations Density 0.018%

words related to valuable objects or items

terms related to artifacts and relics

No Comments

No Known Activations