INDEX

Explanations

phrases related to actions and outcomes

oai_token-act-pair · gpt-3.5-turbo

conjunctions that indicate a series of actions or consequences

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

Redditor

-0.72

Tes

-0.71

agon

-0.70

asper

-0.68

ibble

-0.66

rb

-0.64

Bas

-0.64

fal

-0.64

bub

-0.63

blems

-0.63

POSITIVE LOGITS

 assorted

1.08

 consequently

0.96

 possibly

0.96

 finally

0.95

 downright

0.95

 ultimately

0.94

 frankly

0.86

 perhaps

0.84

 then

0.82

yes

0.79

Activations Density 0.157%

phrases related to actions and outcomes

conjunctions that indicate a series of actions or consequences

No Comments

No Known Activations