INDEX

Explanations

phrases indicative of hidden information or processes

oai_token-act-pair · gpt-3.5-turbo

references to inner experiences or thoughts

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

ensen

-0.85

essors

-0.81

eday

-0.74

ensable

-0.73

ares

-0.72

orthy

-0.72

etting

-0.71

atoes

-0.71

llah

-0.69

ILLE

-0.67

POSITIVE LOGITS

 workings

1.27

most

1.20

 circle

0.95

 sanct

0.88

 turmoil

0.85

 Mongolia

0.83

circle

0.80

 combustion

0.80

ranean

0.79

 thigh

0.77

Activations Density 0.021%

phrases indicative of hidden information or processes

references to inner experiences or thoughts

No Comments

No Known Activations