INDEX

Explanations

percentages or numerical figures

oai_token-act-pair · gpt-3.5-turbo

references to statistical percentages

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Honour

-0.76

 Whilst

-0.69

 organising

-0.69

 neighbours

-0.69

↵Âł

-0.65

 emphas

-0.63

 whilst

-0.63

£ı

-0.62

 rumours

-0.61

oeuv

-0.60

POSITIVE LOGITS

 percent

3.19

percent

2.58

 Percent

2.30

2.15

%.

1.94

Percent

1.93

 percentage

1.90

%,

1.90

%"

1.90

1.87

Activations Density 0.037%

percentages or numerical figures

references to statistical percentages

No Comments

No Known Activations