INDEX

Explanations

the word 't' followed by a verb

oai_token-act-pair · gpt-3.5-turbo

negations or contractions of 'will not'

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 Dickinson

-0.66

 Compass

-0.61

 Reborn

-0.59

cius

-0.58

 Vine

-0.57

Kings

-0.55

 modem

-0.53

Via

-0.53

 Economy

-0.53

ifter

-0.52

POSITIVE LOGITS

be

1.01

 hesitate

0.96

 necessarily

0.92

 bother

0.87

ogether

0.87

 discriminate

0.87

 tolerate

0.85

 interfere

0.85

aylor

0.85

 prevail

0.85

Activations Density 0.066%

the word 't' followed by a verb

negations or contractions of 'will not'

No Comments

No Known Activations