INDEX

Explanations

contractions and abbreviated forms containing 't

oai_token-act-pair · gpt-3.5-turbo

first-person singular pronouns and associated phrases indicating uncertainty or inquiry

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

 guiActiveUnfocused

-0.74

amara

-0.63

andon

-0.60

izational

-0.59

cedes

-0.58

rients

-0.58

come

-0.56

orld

-0.55

comes

-0.55

arget

-0.55

POSITIVE LOGITS

 presume

1.27

 assume

1.18

 doubt

1.15

 speculate

1.09

 dunno

1.09

 suspect

1.01

 guesses

0.98

 conjecture

0.98

 guessing

0.97

 cannot

0.95

Activations Density 0.255%

contractions and abbreviated forms containing 't

first-person singular pronouns and associated phrases indicating uncertainty or inquiry

No Comments

No Known Activations