INDEX

Explanations

words associated with warnings or caution

oai_token-act-pair · gpt-3.5-turbo

negative contractions, particularly forms of "not."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 fortun

-0.65

 obser

-0.61

 disadvant

-0.60

 pyramid

-0.58

 lawy

-0.58

 Folk

-0.57

 ende

-0.57

 Palestin

-0.57

 mathemat

-0.56

 Seym

-0.56

POSITIVE LOGITS

ï¸ı

1.33

Â£

0.79

âĻ

0.78

own

0.77

ï¸

0.76

Balt

0.75

âĤ¬

0.73

âĹ

0.71

_>

0.70

âĶĢâĶĢâĶĢâĶĢ

0.70

Activations Density 0.369%

words associated with warnings or caution

negative contractions, particularly forms of "not."

No Comments

No Known Activations