INDEX

Explanations

mentions of poking or poking-related actions

oai_token-act-pair · gpt-3.5-turbo

instances of the word "poke" and related terms, indicating playful or humorous contexts

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

 estranged

-0.73

wealth

-0.65

Sem

-0.65

 Status

-0.63

 semin

-0.63

Zur

-0.63

 Journal

-0.62

 eminent

-0.62

 Font

-0.61

sem

-0.60

POSITIVE LOGITS

poke

5.11

peak

1.81

 Poke

1.46

 poke

1.35

Pokemon

1.10

okÃ©

1.09

 pokemon

1.07

twitch

1.06

oleon

1.05

 poked

1.03

Activations Density 0.039%

mentions of poking or poking-related actions

instances of the word "poke" and related terms, indicating playful or humorous contexts

No Comments

No Known Activations