INDEX

Explanations

documents where decisions or actions are being taken

oai_token-act-pair · gpt-3.5-turbo

instances of the word "to" in various contexts, indicating directional or infinitive phrases

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 wastes

-0.69

 hairs

-0.69

ain

-0.68

need

-0.66

Got

-0.65

grain

-0.65

coins

-0.64

 slightest

-0.64

 glances

-0.64

croft

-0.63

POSITIVE LOGITS

 accommodate

1.08

 ensure

1.07

 strengthen

1.05

 compensate

1.03

 commemorate

1.03

 meet

1.02

 promote

1.02

 defend

1.00

 accompany

0.99

 celebrate

0.99

Activations Density 0.958%

documents where decisions or actions are being taken

instances of the word "to" in various contexts, indicating directional or infinitive phrases

No Comments

No Known Activations