INDEX

Explanations

phrases indicating giving instructions or sharing information

oai_token-act-pair · gpt-3.5-turbo

phrases related to rules, instructions, and cautionary advice

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

uga

-0.81

rane

-0.75

ourning

-0.72

Rot

-0.72

Sword

-0.71

ordable

-0.70

imentary

-0.70

obar

-0.70

acers

-0.68

hap

-0.67

POSITIVE LOGITS

 mentality

1.13

 attitude

1.04

 initiative

1.04

 clause

1.04

 mantra

1.02

 approach

0.99

 syndrome

0.97

 campaign

0.96

 distinction

0.94

 ratio

0.93

Activations Density 0.412%

phrases indicating giving instructions or sharing information

phrases related to rules, instructions, and cautionary advice

No Comments

No Known Activations