INDEX

Explanations

specific numerical values

oai_token-act-pair · gpt-3.5-turbo

sequences of characters or patterns resembling "x" followed by numbers

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Hab

-0.75

 breath

-0.62

 deserted

-0.61

Nun

-0.60

 washed

-0.60

 balancing

-0.59

 relied

-0.59

 cultured

-0.59

 forgiven

-0.58

 Labrador

-0.58

POSITIVE LOGITS

3.85

xes

2.85

xc

2.29

xb

2.25

xa

2.23

xd

2.23

xf

2.23

xff

2.15

xe

2.15

xs

2.11

Activations Density 0.017%

specific numerical values

sequences of characters or patterns resembling "x" followed by numbers

No Comments

No Known Activations