INDEX

Explanations

numeric values in the form of quantities

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "two" and numbers indicating quantities or counts

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

LER

-0.70

eah

-0.67

UGE

-0.61

indust

-0.61

asta

-0.60

 ours

-0.59

grim

-0.59

 Krish

-0.58

adish

-0.58

ampunk

-0.58

POSITIVE LOGITS

peat

1.14

 consecutive

1.08

teenth

1.01

eenth

1.01

 apiece

0.96

 successive

0.92

 assists

0.90

 interceptions

0.89

aciously

0.86

straight

0.83

Activations Density 0.175%

numeric values in the form of quantities

occurrences of the word "two" and numbers indicating quantities or counts

No Comments

No Known Activations