INDEX

Explanations

terms related to academic or skill levels, specifically focusing on words indicating a higher level or rating

oai_token-act-pair · gpt-3.5-turbo

references to grading or quality assessments in 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

lit

-0.80

comb

-0.65

MET

-0.64

 registry

-0.63

ISC

-0.63

POST

-0.61

oos

-0.59

 hect

-0.58

combe

-0.58

doi

-0.57

POSITIVE LOGITS

grade

1.13

grad

0.96

ffiti

0.91

ionage

0.86

llular

0.84

xual

0.78

ctr

0.77

ipment

0.77

illet

0.76

ably

0.75

Activations Density 0.012%

terms related to academic or skill levels, specifically focusing on words indicating a higher level or rating

references to grading or quality assessments in various contexts

No Comments

No Known Activations