INDEX

Explanations

descriptions containing words related to physical appearance or conditions, especially negative ones

oai_token-act-pair · gpt-3.5-turbo

terms related to unflattering appearances or representations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 funds

-0.64

 stems

-0.63

 licence

-0.62

 solved

-0.61

 dies

-0.61

 Christ

-0.61

 reserve

-0.61

 Â©

-0.59

 license

-0.59

 fund

-0.58

POSITIVE LOGITS

attering

4.27

attered

2.15

atter

1.78

atters

1.69

ashing

1.19

attery

1.18

ithing

1.03

uttering

1.02

 scattering

0.99

owder

0.97

Activations Density 0.017%

descriptions containing words related to physical appearance or conditions, especially negative ones

terms related to unflattering appearances or representations

No Comments

No Known Activations