INDEX

Explanations

terms related to keeping a lookout or performing reconnaissance

oai_token-act-pair · gpt-3.5-turbo

terms related to surveillance and observation activities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

rys

-0.89

umat

-0.87

iev

-0.81

lock

-0.81

nik

-0.80

arak

-0.80

urity

-0.79

tering

-0.79

inous

-0.78

lez

-0.78

POSITIVE LOGITS

etheless

0.88

 Satellite

0.82

 Phant

0.75

 bulletin

0.73

 Grounds

0.72

 Tips

0.70

 sights

0.68

ally

0.66

 sidx

0.65

 Krug

0.65

Activations Density 0.045%

terms related to keeping a lookout or performing reconnaissance

terms related to surveillance and observation activities

No Comments

No Known Activations