INDEX

Explanations

instances of the word "texts."

oai_token-act-pair · gpt-3.5-turbo

references to texts and scripture

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

edo

-0.67

iler

-0.66

inion

-0.65

IR

-0.65

buck

-0.64

 cushion

-0.64

irk

-0.63

ï¸

-0.63

hing

-0.63

cone

-0.62

POSITIVE LOGITS

 texts

3.79

 text

1.90

 manuscripts

1.77

 textbooks

1.67

 poems

1.60

 Text

1.58

 scriptures

1.56

 texted

1.49

 messages

1.44

 writings

1.44

Activations Density 0.012%

instances of the word "texts."

references to texts and scripture

No Comments

No Known Activations