INDEX

Explanations

words related to capabilities or functionalities

oai_token-act-pair · gpt-3.5-turbo

modal verbs that indicate potential or capability

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 hating

-0.78

forcing

-0.72

 Lauder

-0.68

 Niet

-0.67

soType

-0.66

 Watching

-0.63

 waning

-0.63

 looming

-0.63

pants

-0.62

 wanting

-0.61

POSITIVE LOGITS

 withstand

1.31

't

1.27

 accommodate

1.13

berra

1.13

 operate

1.12

isters

1.11

 penetrate

1.11

 optionally

1.10

be

1.10

 tolerate

1.07

Activations Density 0.132%

words related to capabilities or functionalities

modal verbs that indicate potential or capability

No Comments

No Known Activations