INDEX

Explanations

strong, negative actions or criticisms

oai_token-act-pair · gpt-3.5-turbo

instances of verbal aggression or confrontation in text

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Notting

-0.74

 orderly

-0.70

Transfer

-0.68

 Genius

-0.67

 Transform

-0.65

stable

-0.62

sterdam

-0.59

 Alive

-0.59

 safest

-0.58

kj

-0.58

POSITIVE LOGITS

 accusing

1.00

 against

0.91

 leveled

0.84

jab

0.81

 accuses

0.81

 criticizing

0.81

 critiques

0.80

 critics

0.79

 insults

0.79

against

0.78

Activations Density 0.203%

strong, negative actions or criticisms

instances of verbal aggression or confrontation in text

No Comments

No Known Activations