INDEX

Explanations

words related to questioning or arguing

oai_token-act-pair · gpt-3.5-turbo

instances of the syllable "qu."

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 permitting

-0.70

HAEL

-0.69

 Prosper

-0.65

bul

-0.63

croft

-0.62

strap

-0.61

 overhead

-0.61

 ground

-0.61

 disparate

-0.61

 disappoint

-0.61

POSITIVE LOGITS

arre

1.27

idd

1.11

irk

1.10

ivering

1.10

arant

1.10

ibble

1.10

iets

1.09

asar

1.04

inoa

1.02

orum

1.01

Activations Density 0.012%

words related to questioning or arguing

instances of the syllable "qu."

No Comments

No Known Activations