INDEX

Explanations

instances where actions or decisions are being justified or explained

oai_token-act-pair · gpt-3.5-turbo

conditional statements and expressions of intent or ability

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

 Cruel

-0.58

 Became

-0.56

 Definitions

-0.56

ģĸ

-0.55

 unbeat

-0.54

SI

-0.53

 Underworld

-0.52

 Scand

-0.50

worth

-0.50

abre

-0.50

POSITIVE LOGITS

 instead

2.01

Instead

1.78

 Instead

1.70

 merely

1.60

instead

1.53

 rather

1.50

Rather

1.49

 Rather

1.47

 simply

1.30

rather

1.30

Activations Density 0.628%

instances where actions or decisions are being justified or explained

conditional statements and expressions of intent or ability

No Comments

No Known Activations