INDEX

Explanations

programming code-related tokens or markers

oai_token-act-pair · gpt-3.5-turbo

unique formatting symbols or special characters

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

BBC

-0.73

Sob

-0.73

 Tribal

-0.71

 Holden

-0.66

 Victory

-0.62

TV

-0.62

 Registration

-0.61

 Bild

-0.61

 Bened

-0.61

 Cornwall

-0.61

POSITIVE LOGITS

âĹ¼

1.30

âĸł

0.98

âĢ¢

0.91

Â·

0.87

âĹı

0.87

have

0.87

âĻ¦

0.86

âĤ¬

0.85

depending

0.84

Interstitial

0.83

Activations Density 0.498%

programming code-related tokens or markers

unique formatting symbols or special characters

No Comments

No Known Activations