INDEX

Explanations

entities associated with criminal acts

oai_token-act-pair · gpt-3.5-turbo

the word "who" 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 @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ÂłÂłÂłÂł

-0.80

 Bound

-0.71

BACK

-0.68

 Delicious

-0.67

 Glob

-0.65

 Processing

-0.65

 Around

-0.65

Dos

-0.64

 Lions

-0.64

 Anything

-0.63

POSITIVE LOGITS

soever

1.11

oping

1.02

 accompanies

0.99

 preceded

0.96

oped

0.95

 attended

0.90

 oversaw

0.89

ever

0.89

 attends

0.89

 resided

0.88

Activations Density 0.168%

entities associated with criminal acts

the word "who" in various contexts

No Comments

No Known Activations