INDEX

Explanations

phrases related to editing and modifying content, such as narrowing down or cleaning up

oai_token-act-pair · gpt-3.5-turbo

actions related to decision-making and modifications

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

anasia

-0.78

onial

-0.66

anie

-0.62

oliberal

-0.62

riots

-0.61

cig

-0.60

liv

-0.60

illance

-0.60

onna

-0.59

drawn

-0.59

POSITIVE LOGITS

 things

0.97

it

0.93

 this

0.88

 everything

0.85

 them

0.81

 these

0.80

 those

0.79

itably

0.70

 ours

0.69

 matters

0.66

Activations Density 0.283%

phrases related to editing and modifying content, such as narrowing down or cleaning up

actions related to decision-making and modifications

No Comments

No Known Activations