INDEX

Explanations

phrases indicating reasons or explanations

oai_token-act-pair · gpt-3.5-turbo

references to sections or parts of a larger text or context

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

mins

-0.68

marine

-0.65

 slightest

-0.61

 cannabin

-0.60

 clipboard

-0.59

Express

-0.58

lean

-0.58

 Commissioners

-0.57

 NEXT

-0.56

 Express

-0.56

POSITIVE LOGITS

 thanks

0.85

 because

0.80

due

0.76

 owing

0.70

because

0.70

atos

0.70

laughter

0.67

aughter

0.65

 attributable

0.63

icularly

0.63

Activations Density 0.041%

phrases indicating reasons or explanations

references to sections or parts of a larger text or context

No Comments

No Known Activations