INDEX

Explanations

words related to people's names, often including their titles or nicknames

oai_token-act-pair · gpt-3.5-turbo

quotations from dialogue or speech

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 reconcil

-0.76

 intersections

-0.74

 reconcile

-0.73

 overturn

-0.73

 coincide

-0.72

 inauguration

-0.72

 recount

-0.71

 forgiven

-0.69

 reorgan

-0.69

 orgasm

-0.67

POSITIVE LOGITS

Sem

1.05

Fat

1.02

Thor

0.96

Big

0.95

Hung

0.94

Fuck

0.92

Arm

0.91

Happy

0.91

Skip

0.91

vik

0.90

Activations Density 0.036%

words related to people's names, often including their titles or nicknames

quotations from dialogue or speech

No Comments

No Known Activations