INDEX

Explanations

references to a specific term spelled as 'To'

oai_token-act-pair · gpt-3.5-turbo

references to the word "To" indicating titles or names

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 intentions

-0.68

 exceptions

-0.66

 traces

-0.62

 deductions

-0.61

 intrusive

-0.60

 dece

-0.60

 vulnerabilities

-0.60

iHUD

-0.59

 regards

-0.59

reports

-0.58

POSITIVE LOGITS

ilet

1.65

pping

1.19

ilers

1.14

asted

1.11

pped

1.10

ppings

1.08

asting

1.07

asters

1.02

ggles

1.02

ffee

1.02

Activations Density 0.059%

references to a specific term spelled as 'To'

references to the word "To" indicating titles or names

No Comments

No Known Activations