INDEX

Explanations

places or locations

oai_token-act-pair · gpt-3.5-turbo

references to a central location or theme

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

opard

-0.68

RESULTS

-0.67

ods

-0.66

odcast

-0.66

BIT

-0.65

WN

-0.65

 Griffith

-0.64

 Smoking

-0.64

 Shotgun

-0.62

aceae

-0.62

POSITIVE LOGITS

pieces

1.12

piece

0.91

tenance

0.89

 centres

0.87

most

0.79

 fielder

0.76

 centre

0.76

stadt

0.74

 rotor

0.72

camp

0.72

Activations Density 0.010%

places or locations

references to a central location or theme

No Comments

No Known Activations