INDEX

Explanations

time-related expressions, durations, and percentages

oai_token-act-pair · gpt-3.5-turbo

numerical values, especially in the context of measurements or percentages

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

 Doodle

-0.75

NAS

-0.75

Joy

-0.72

lance

-0.70

Duo

-0.69

andi

-0.67

alogy

-0.66

 Probe

-0.65

 Parish

-0.65

Crime

-0.64

POSITIVE LOGITS

 times

0.97

 percent

0.96

 thousand

0.91

 dozen

0.88

 certs

0.88

 hours

0.87

 cycles

0.85

percent

0.84

ths

0.83

 million

0.83

Activations Density 0.099%

time-related expressions, durations, and percentages

numerical values, especially in the context of measurements or percentages

No Comments

No Known Activations