INDEX

Explanations

connecting words indicating a contrast or change in direction

oai_token-act-pair · gpt-3.5-turbo

instances of contrast or transition in statements

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

 segment

-0.78

 ceremony

-0.60

 mound

-0.58

sym

-0.58

stre

-0.57

tnc

-0.55

 hall

-0.55

ãģ®

-0.55

abilities

-0.54

 reef

-0.54

POSITIVE LOGITS

tons

1.67

 alas

1.12

tery

1.00

ts

0.98

chery

0.91

hey

0.87

LER

0.85

lers

0.83

chers

0.79

oner

0.76

Activations Density 0.122%

connecting words indicating a contrast or change in direction

instances of contrast or transition in statements

No Comments

No Known Activations