INDEX

Explanations

phrases that imply a specific piece of information is being revealed or communicated

oai_token-act-pair · gpt-3.5-turbo

the word "indicate" and its variations, signifying evidence or suggestion

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

fare

-0.78

venge

-0.75

ÄŁ

-0.72

@#&

-0.71

ctors

-0.70

tal

-0.69

pping

-0.69

vas

-0.69

rit

-0.69

iling

-0.69

POSITIVE LOGITS

 indications

1.04

 indicates

0.90

 indication

0.90

 signs

0.86

 indicated

0.85

 hints

0.81

 indicating

0.80

 indicate

0.79

 intervals

0.78

 Signs

0.74

Activations Density 0.025%

phrases that imply a specific piece of information is being revealed or communicated

the word "indicate" and its variations, signifying evidence or suggestion

No Comments

No Known Activations