INDEX

Explanations

phrases indicating a survey or research study

oai_token-act-pair · gpt-3.5-turbo

references to survey data and polling results

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 Clown

-0.68

der

-0.65

 insanity

-0.63

fun

-0.62

 supers

-0.62

 madness

-0.62

 continuity

-0.61

acity

-0.60

otion

-0.60

boxing

-0.60

POSITIVE LOGITS

 surveyed

3.79

 polled

2.76

 sampled

1.72

 interviewed

1.63

 survey

1.60

 surveys

1.60

 surve

1.59

 studied

1.55

 examined

1.53

 Surve

1.52

Activations Density 0.012%

phrases indicating a survey or research study

references to survey data and polling results

No Comments

No Known Activations