INDEX

Explanations

phrases prompting or advising against a certain action

oai_token-act-pair · gpt-3.5-turbo

imperatives and negative commands or suggestions

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

ELD

-0.76

Redd

-0.70

liner

-0.68

dimension

-0.63

prof

-0.63

ħĭ

-0.62

established

-0.61

upon

-0.61

milo

-0.61

ilage

-0.61

POSITIVE LOGITS

 underestimate

0.89

 hesitate

0.86

theless

0.86

ndum

0.82

heny

0.78

 kidding

0.76

ardless

0.74

 worry

0.74

omsday

0.74

ations

0.72

Activations Density 0.057%

phrases prompting or advising against a certain action

imperatives and negative commands or suggestions

No Comments

No Known Activations