INDEX

Explanations

specific percentage numbers in texts

oai_token-act-pair · gpt-3.5-turbo

quantitative data in percentages

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

HCR

-0.78

ãĤ¡

-0.77

ĸļ

-0.76

 Cassidy

-0.76

AVG

-0.75

ãĥ£

-0.71

ãĤ¯

-0.70

Osw

-0.69

ãĥ¼ãĥĨãĤ£

-0.69

earch

-0.69

POSITIVE LOGITS

sized

1.13

legged

1.06

eight

1.05

tailed

1.04

year

0.99

percent

0.98

six

0.96

nine

0.95

hour

0.93

seven

0.92

Activations Density 0.054%

specific percentage numbers in texts

quantitative data in percentages

No Comments

No Known Activations