INDEX

Explanations

words related to reasons or explanations for certain events

oai_token-act-pair · gpt-3.5-turbo

references to causation or reasons for events

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 tick

-0.66

 matter

-0.65

 adapter

-0.62

 spread

-0.60

 index

-0.60

 shuff

-0.60

 dream

-0.60

sw

-0.59

 overl

-0.59

ho

-0.59

POSITIVE LOGITS

due

4.59

because

1.36

due

1.32

Due

1.31

Due

1.30

despite

1.17

given

1.16

since

1.15

thanks

1.14

cause

1.06

Activations Density 0.025%

words related to reasons or explanations for certain events

references to causation or reasons for events

No Comments

No Known Activations