INDEX

Explanations

quotations or reported speech with the word "said"

oai_token-act-pair · gpt-3.5-turbo

statements attributed to individuals or sources

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

perate

-0.63

Â¶

-0.62

iscover

-0.59

cover

-0.58

eds

-0.58

escape

-0.57

 regress

-0.57

 âĨĴ

-0.55

ogenic

-0.55

prep

-0.54

POSITIVE LOGITS

 said

2.89

said

2.40

 Said

1.72

 stated

1.72

Said

1.72

 says

1.69

 told

1.62

 remarked

1.57

 explained

1.51

 joked

1.50

Activations Density 0.242%

quotations or reported speech with the word "said"

statements attributed to individuals or sources

No Comments

No Known Activations