INDEX

Explanations

keywords related to conflict, such as battles

oai_token-act-pair · gpt-3.5-turbo

references to battles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

Property

-0.77

perm

-0.72

entric

-0.72

ħĭ

-0.70

ĻĤ

-0.70

attribute

-0.69

ollow

-0.69

eller

-0.68

 confir

-0.67

safety

-0.67

POSITIVE LOGITS

 battles

1.21

 battle

1.05

 fought

0.93

naire

0.91

cru

0.91

roy

0.91

axe

0.89

 duel

0.88

 waged

0.88

 battling

0.85

Activations Density 0.015%

keywords related to conflict, such as battles

references to battles

No Comments

No Known Activations