INDEX

Explanations

expressions related to achieving victory or winning a game

oai_token-act-pair · gpt-3.5-turbo

instances of the word "win" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 encyclopedia

-0.67

 intestine

-0.67

senal

-0.63

 notor

-0.62

 copying

-0.61

 cens

-0.60

 incorpor

-0.60

 bowel

-0.60

 labour

-0.57

omorphic

-0.57

POSITIVE LOGITS

nings

1.43

liness

0.97

now

0.96

iem

0.95

throp

0.93

 streak

0.92

stre

0.88

ces

0.83

fig

0.80

athon

0.80

Activations Density 0.039%

expressions related to achieving victory or winning a game

instances of the word "win" in various contexts

No Comments

No Known Activations