INDEX

Explanations

phrases related to participation or involvement in activities

oai_token-act-pair · gpt-3.5-turbo

instances of the word "entered."

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

 downt

-0.63

 debunk

-0.61

Ban

-0.59

 spaghetti

-0.59

rast

-0.56

PBS

-0.56

texture

-0.55

 criticism

-0.55

 Pistons

-0.55

 Plane

-0.54

POSITIVE LOGITS

 entered

3.56

 enters

2.11

 enter

1.99

 exited

1.81

 Entered

1.74

 entering

1.72

 reached

1.57

 arrived

1.53

 ENTER

1.43

 invaded

1.40

Activations Density 0.017%

phrases related to participation or involvement in activities

instances of the word "entered."

No Comments

No Known Activations