INDEX

Explanations

intentions or plans as articulated in text

oai_token-act-pair · gpt-3.5-turbo

mentions of intention across various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

cit

-0.71

apple

-0.68

 Tycoon

-0.68

java

-0.66

 Wolves

-0.63

Tier

-0.59

 Boss

-0.59

angs

-0.58

uf

-0.58

 Roses

-0.58

POSITIVE LOGITS

 intention

0.98

ality

0.97

 intent

0.89

edly

0.86

 guiActiveUn

0.86

terday

0.85

 intending

0.83

reprene

0.82

edIn

0.79

gomery

0.78

Activations Density 0.007%

intentions or plans as articulated in text

mentions of intention across various contexts

No Comments

No Known Activations