INDEX

Explanations

mentions of physical structures or barriers like walls

oai_token-act-pair · gpt-3.5-turbo

references to a physical barrier or wall

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

Not in Any Lists

No Comments

Negative Logits

Ake

-0.83

Lear

-0.72

invoke

-0.71

registered

-0.71

 Lear

-0.68

akh

-0.66

acca

-0.65

served

-0.64

Topics

-0.63

iblings

-0.63

POSITIVE LOGITS

 wall

3.68

 walls

2.61

wall

2.24

Wall

2.03

 Wall

1.88

 Walls

1.85

 fence

1.67

 ceiling

1.66

 barrier

1.47

 firewall

1.30

Activations Density 0.009%

mentions of physical structures or barriers like walls

references to a physical barrier or wall

No Comments

No Known Activations