INDEX

Explanations

numerical comparisons, especially when one value is significantly larger than another

oai_token-act-pair · gpt-3.5-turbo

quantitative comparisons and expressions of magnitude

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

RAW

-0.88

PLA

-0.75

 =================================================================

-0.72

////////

-0.72

AMA

-0.68

ARA

-0.66

 Zionism

-0.64

SPONSORED

-0.64

 Adin

-0.63

Sty

-0.62

POSITIVE LOGITS

 average

0.93

 usual

0.90

manac

0.82

ietal

0.80

 busiest

0.80

average

0.80

 fastest

0.76

umat

0.75

oples

0.73

usual

0.70

Activations Density 0.125%

numerical comparisons, especially when one value is significantly larger than another

quantitative comparisons and expressions of magnitude

No Comments

No Known Activations