INDEX

Explanations

phrases related to opinions or statements, potentially with emphatic language like "no more" or "hope"

oai_token-act-pair · gpt-3.5-turbo

notations indicating the absence of conflict or moderation

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

igent

-0.87

Enlarge

-0.85

ashington

-0.83

estine

-0.80

typically

-0.78

obook

-0.78

occup

-0.77

associated

-0.77

Supported

-0.76

edIn

-0.75

POSITIVE LOGITS

 surprises

1.16

 booze

1.11

 explosions

1.10

 hugs

1.10

 fireworks

1.10

 roses

1.08

 spoilers

1.08

 applause

1.07

 goodies

1.05

 excuses

1.03

Activations Density 0.563%

phrases related to opinions or statements, potentially with emphatic language like "no more" or "hope"

notations indicating the absence of conflict or moderation

No Comments

No Known Activations