INDEX

Explanations

phrases related to actions or instructions involving the reader

oai_token-act-pair · gpt-3.5-turbo

phrases that directly address or involve the reader

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

course

-0.68

 Sunset

-0.66

 Meadow

-0.65

İĭ

-0.65

stadt

-0.63

 Reconstruction

-0.63

duty

-0.62

odor

-0.61

=~

-0.59

Mission

-0.58

POSITIVE LOGITS

're

1.51

've

1.25

 hear

0.98

 guys

0.93

 realize

0.90

'll

0.90

 compare

0.88

 realise

0.87

'd

0.86

ask

0.85

Activations Density 0.093%

phrases related to actions or instructions involving the reader

phrases that directly address or involve the reader

No Comments

No Known Activations