INDEX

Explanations

phrases related to making tasks easier or more efficient

oai_token-act-pair · gpt-3.5-turbo

phrases indicating ease of use or improvement in efficiency

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 irony

-0.67

 wonders

-0.66

 aside

-0.66

 treason

-0.64

 hypocrisy

-0.62

 dearly

-0.61

secret

-0.60

 faint

-0.59

warm

-0.59

 straw

-0.59

POSITIVE LOGITS

 than

0.98

than

0.87

usable

0.76

etitive

0.74

iaz

0.70

 allocate

0.70

catentry

0.70

 Faster

0.70

 Than

0.70

 communicate

0.69

Activations Density 0.330%

phrases related to making tasks easier or more efficient

phrases indicating ease of use or improvement in efficiency

No Comments

No Known Activations