INDEX

Explanations

the word "preferred" in a sentence

oai_token-act-pair · gpt-3.5-turbo

references to choices or preferences

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

????????

-0.75

 Tours

-0.67

circle

-0.67

humans

-0.67

ences

-0.67

lifting

-0.66

iry

-0.65

 Revival

-0.63

 Unlock

-0.62

corruption

-0.61

POSITIVE LOGITS

 preferred

3.93

 favoured

2.15

 desired

1.95

 Preferred

1.95

 favored

1.94

 preference

1.93

 preferable

1.90

 prefers

1.77

 prefer

1.74

 disliked

1.64

Activations Density 0.008%

the word "preferred" in a sentence

references to choices or preferences

No Comments

No Known Activations