INDEX

Explanations

references to completing a task or reaching a goal

oai_token-act-pair · gpt-3.5-turbo

the word "finish" and its variations, which indicates a focus on completion and endings

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

amount

-0.76

reported

-0.75

elled

-0.73

models

-0.73

orb

-0.72

eros

-0.71

edia

-0.71

ria

-0.68

add

-0.68

bour

-0.67

POSITIVE LOGITS

 finish

1.14

 Finish

1.09

 finishes

0.96

 finishing

0.89

smanship

0.76

 unfinished

0.75

Finish

0.75

 Finished

0.74

isans

0.72

 elim

0.71

Activations Density 0.008%

references to completing a task or reaching a goal

the word "finish" and its variations, which indicates a focus on completion and endings

No Comments

No Known Activations