INDEX

Explanations

phrases related to actions or inactions leading to negative consequences or failures

oai_token-act-pair · gpt-3.5-turbo

words and phrases related to failures and obligations

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

rather

-0.72

okingly

-0.71

initely

-0.67

usra

-0.66

=#

-0.65

ectar

-0.65

aceae

-0.64

uncle

-0.63

Discussion

-0.63

augh

-0.62

POSITIVE LOGITS

 anymore

1.29

nor

1.09

 adequately

0.95

 whatsoever

0.93

 adequate

0.81

any

0.79

 anything

0.75

yet

0.75

due

0.75

 altogether

0.74

Activations Density 0.428%

phrases related to actions or inactions leading to negative consequences or failures

words and phrases related to failures and obligations

No Comments

No Known Activations