INDEX

Explanations

phrases indicating agreement or concession

oai_token-act-pair · gpt-3.5-turbo

phrases that express acknowledgment or admission

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

cum

-0.73

cair

-0.70

que

-0.68

lets

-0.67

ENC

-0.65

mented

-0.65

QL

-0.63

ENCE

-0.62

¹

-0.62

rology

-0.62

POSITIVE LOGITS

imaru

0.78

terday

0.69

 admittedly

0.68

sorry

0.66

 forgiven

0.62

lihood

0.62

 disclaimer

0.61

argon

0.61

Obj

0.61

 Lerner

0.61

Activations Density 0.032%

phrases indicating agreement or concession

phrases that express acknowledgment or admission

No Comments

No Known Activations