INDEX

Explanations

Pokémon card names and related terms

oai_token-act-pair · gpt-3.5-turbo

special characters or unique symbols that may represent specific game mechanics or items

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

 dope

-0.89

 peers

-0.87

 colleg

-0.81

 clinicians

-0.79

 kosher

-0.77

 betting

-0.76

 consensus

-0.76

 democratically

-0.75

 rigged

-0.75

 disg

-0.74

POSITIVE LOGITS

ãĥ«

2.26

ãĥ¼ãĥ

2.24

ãĤ«

2.23

ãĥł

2.21

ãĤ¹

2.19

ãĤ¿

2.17

ãĤ¯

2.17

ãĥ©

2.16

ãĥĪ

2.16

ãĥī

2.15

Activations Density 0.026%

Pokémon card names and related terms

special characters or unique symbols that may represent specific game mechanics or items

No Comments

No Known Activations