INDEX

Explanations

proper nouns

oai_token-act-pair · gpt-3.5-turbo

references to prior events or actions

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

 ware

-0.85

 kits

-0.75

 wonders

-0.69

 uber

-0.64

 Elixir

-0.63

arty

-0.62

fab

-0.60

 fodder

-0.59

 Kamp

-0.59

 pearl

-0.59

POSITIVE LOGITS

Previous

4.04

 Previous

3.09

Prev

1.93

Previously

1.78

 previous

1.57

Current

1.56

 Previously

1.47

 Prev

1.37

prev

1.34

Recent

1.29

Activations Density 0.031%

proper nouns

references to prior events or actions

No Comments

No Known Activations