INDEX

Explanations

phrases related to editorial or historical context

oai_token-act-pair · gpt-3.5-turbo

structured data or references, often related to statistics or lists

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 casc

-0.89

 eleph

-0.87

 yours

-0.80

 crunch

-0.80

 urge

-0.78

 soak

-0.77

 downright

-0.76

 enriched

-0.76

 sucker

-0.76

 trump

-0.76

POSITIVE LOGITS

Trivia

1.99

Appearances

1.90

References

1.90

History

1.86

Contents

1.84

Notable

1.80

Gallery

1.77

Appearance

1.71

Notes

1.65

Background

1.64

Activations Density 0.145%

phrases related to editorial or historical context

structured data or references, often related to statistics or lists

No Comments

No Known Activations