INDEX

Explanations

surprised reactions and discoveries

oai_token-act-pair · gpt-3.5-turbo

expressions related to surprise and discovery

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

•layer 8 - surprise

No Comments

Negative Logits

 vend

-0.72

hung

-0.69

 condu

-0.67

ulton

-0.67

aim

-0.65

burgh

-0.63

bye

-0.62

leaf

-0.61

interrupted

-0.61

 stake

-0.60

POSITIVE LOGITS

 surprise

0.83

icably

0.69

how

0.67

 surprises

0.65

 unexpected

0.64

how

0.64

 reactions

0.64

cus

0.63

 encountering

0.62

 surprised

0.61

Activations Density 0.161%

surprised reactions and discoveries

expressions related to surprise and discovery

No Comments

No Known Activations