INDEX

Explanations

question words like "What" or "Who"

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "Wh"

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

 Grande

-0.67

WARE

-0.67

Duo

-0.65

 Awakening

-0.61

ULAR

-0.61

 Blazers

-0.61

 Strauss

-0.61

 Letter

-0.59

 Barton

-0.59

 sten

-0.57

POSITIVE LOGITS

istle

1.43

ilst

1.35

irlwind

1.26

olly

1.21

ispers

1.20

soever

1.17

isky

1.16

olen

1.11

irling

1.10

ichever

1.08

Activations Density 0.023%

question words like "What" or "Who"

occurrences of the word "Wh"

No Comments

No Known Activations