INDEX

Explanations

phrases related to rankings or lists, especially superlative terms like "Top"

oai_token-act-pair · gpt-3.5-turbo

references to rankings or lists

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

merce

-0.84

issance

-0.71

 hanged

-0.69

pci

-0.69

 peacefully

-0.67

ually

-0.65

Â¯Â¯

-0.64

URA

-0.64

éĹ

-0.64

urion

-0.64

POSITIVE LOGITS

Top

3.57

Top

2.69

TOP

2.15

top

2.13

top

2.08

 Bottom

1.97

TOP

1.78

Bottom

1.45

tops

1.43

 tops

1.42

Activations Density 0.013%

phrases related to rankings or lists, especially superlative terms like "Top"

references to rankings or lists

No Comments

No Known Activations