INDEX

Explanations

phrases indicating the start of a list or section in a document

oai_token-act-pair · gpt-3.5-turbo

references to structured formats or sections within a document

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

 work

-0.63

 thinking

-0.62

 culture

-0.61

 unit

-0.61

 commitment

-0.61

 doctrine

-0.61

 tant

-0.60

 party

-0.59

 capacity

-0.59

 stop

-0.58

POSITIVE LOGITS

Below

3.56

 Below

2.41

Above

1.96

below

1.82

 BELOW

1.77

Here

1.54

 Above

1.47

 below

1.40

above

1.33

Beyond

1.25

Activations Density 0.011%

phrases indicating the start of a list or section in a document

references to structured formats or sections within a document

No Comments

No Known Activations