INDEX

Explanations

numerical quantities related to counts

oai_token-act-pair · gpt-3.5-turbo

references to the number 200 and its variations in different contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

•Embedding Space Digit Features

No Comments

Negative Logits

atto

-0.67

anos

-0.66

rh

-0.66

hew

-0.62

rift

-0.61

iosyncr

-0.60

OHN

-0.60

amy

-0.60

Reviewer

-0.60

Leary

-0.60

POSITIVE LOGITS

3.01

2.36

2.30

2.23

2.18

2.14

2.08

1.97

1.95

1.94

Activations Density 0.021%

numerical quantities related to counts

references to the number 200 and its variations in different contexts

No Comments

No Known Activations