INDEX

Explanations

quantitative measurements, specifically focusing on points scored per game

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

hoe

-0.69

enegger

-0.68

Redditor

-0.64

 Halls

-0.63

eworks

-0.62

spir

-0.61

 Benefits

-0.61

MQ

-0.60

opol

-0.60

tops

-0.60

POSITIVE LOGITS

 outing

0.96

 game

0.85

 capita

0.79

 season

0.78

 contest

0.77

 attempt

0.77

 stride

0.74

 league

0.73

 inning

0.73

 minute

0.73

Activations Density 0.045%

quantitative measurements, specifically focusing on points scored per game

No Comments

No Known Activations