INDEX

Explanations

phrases or words related to authority figures

oai_token-act-pair · gpt-3.5-turbo

mentions of legislative or legal terminology related to rules or structures

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 Drawn

-0.76

 Mirage

-0.75

 mounts

-0.72

 passions

-0.71

 avenues

-0.71

 Sammy

-0.71

 intu

-0.71

 counsel

-0.70

 roots

-0.70

 sources

-0.69

POSITIVE LOGITS

uthor

1.13

reci

1.05

vernment

1.04

Â£

1.01

amily

1.00

very

0.97

little

0.95

actual

0.95

significant

0.95

swer

0.94

Activations Density 0.081%

phrases or words related to authority figures

mentions of legislative or legal terminology related to rules or structures

No Comments

No Known Activations