INDEX

Explanations

phrases related to activities or tasks being completed

oai_token-act-pair · gpt-3.5-turbo

negative indicators or descriptions related to individuals or situations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Dickinson

-0.73

ulhu

-0.68

FY

-0.66

 incub

-0.64

 corros

-0.63

 floppy

-0.62

 modem

-0.62

 undergrad

-0.61

 repl

-0.60

 improv

-0.60

POSITIVE LOGITS

san

1.27

sama

1.27

based

1.24

style

1.11

Lago

1.10

themed

1.07

derived

1.06

winning

1.06

kun

1.06

sized

1.05

Activations Density 0.138%

phrases related to activities or tasks being completed

negative indicators or descriptions related to individuals or situations

No Comments

No Known Activations