INDEX

Explanations

words related to following instructions or actions

oai_token-act-pair · gpt-3.5-turbo

instances of the word "follow"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

pite

-0.73

orc

-0.69

ukemia

-0.67

 wounding

-0.67

aucas

-0.67

inese

-0.66

ldom

-0.65

intendent

-0.64

rimination

-0.64

Extreme

-0.63

POSITIVE LOGITS

follow

0.91

 follow

0.85

 Follow

0.83

 follows

0.82

LLOW

0.78

ĸļ

0.76

 suit

0.72

 faithfully

0.71

SHIP

0.70

ansen

0.68

Activations Density 0.025%

words related to following instructions or actions

instances of the word "follow"

No Comments

No Known Activations