INDEX

Explanations

words related to individuals, possibly with emphasis on authority figures

oai_token-act-pair · gpt-3.5-turbo

nouns and their variations, particularly related to people and roles

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

 Topic

-0.69

 duties

-0.60

+/-

-0.56

 lodge

-0.56

 prank

-0.56

 kernels

-0.56

irlf

-0.56

 timer

-0.56

 projecting

-0.54

 ages

-0.54

POSITIVE LOGITS

Reloaded

0.88

nah

0.87

aban

0.87

pha

0.86

omi

0.84

iq

0.84

aki

0.84

ea

0.82

hol

0.80

cia

0.80

Activations Density 0.118%

words related to individuals, possibly with emphasis on authority figures

nouns and their variations, particularly related to people and roles

No Comments

No Known Activations