INDEX

Explanations

numerical values, specifically years or quantities

oai_token-act-pair · gpt-3.5-turbo

instances of the numbers 88 and 89, which likely relate to ratings or scores

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

gets

-0.86

uten

-0.82

arent

-0.79

folk

-0.73

thodox

-0.72

grass

-0.70

kson

-0.69

cha

-0.68

 hemisphere

-0.68

holders

-0.67

POSITIVE LOGITS

0.86

een

0.82

eers

0.80

uador

0.80

0.79

0.74

0.73

Activations Density 0.049%

numerical values, specifically years or quantities

instances of the numbers 88 and 89, which likely relate to ratings or scores

No Comments

No Known Activations