INDEX

Explanations

phrases related to expressing opinions or adding commentary in a discussion

oai_token-act-pair · gpt-3.5-turbo

phrases indicating statements or speech

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

Tud

-0.88

 Titanic

-0.85

idine

-0.84

 Tunis

-0.83

adan

-0.81

isons

-0.80

istani

-0.80

aban

-0.80

adi

-0.80

pak

-0.77

POSITIVE LOGITS

Gr

2.35

Gr

2.29

gr

2.01

gr

1.96

GR

1.85

 Griff

1.69

 Griffin

1.65

GR

1.64

Gro

1.56

Gro

1.52

Activations Density 0.339%

phrases related to expressing opinions or adding commentary in a discussion

phrases indicating statements or speech

No Comments

No Known Activations