INDEX

Explanations

references to awards, nominations, and achievements

oai_token-act-pair · gpt-3.5-turbo

references to award categories and their respective nominations or wins

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

gypt

-0.76

 probing

-0.68

utherford

-0.67

ptin

-0.65

 coli

-0.63

 bypass

-0.63

 awake

-0.62

peror

-0.62

 Juda

-0.62

manent

-0.61

POSITIVE LOGITS

 Best

1.09

seller

1.06

 Worst

1.01

Best

1.00

sell

0.93

iary

0.91

worst

0.91

Winner

0.87

iaries

0.83

hest

0.81

Activations Density 0.009%

references to awards, nominations, and achievements

references to award categories and their respective nominations or wins

No Comments

No Known Activations