INDEX

Explanations

references to violent acts

oai_token-act-pair · gpt-3.5-turbo

references to assassination

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

 Avalon

-0.77

eer

-0.76

ORN

-0.75

RET

-0.75

 Canary

-0.68

 issuance

-0.65

 Reserve

-0.64

 largeDownload

-0.63

irl

-0.63

 Sunshine

-0.63

POSITIVE LOGITS

hing

0.82

hetically

0.79

aukee

0.78

 attm

0.77

 dstg

0.76

atography

0.75

hemat

0.74

agonists

0.71

ochemistry

0.70

heses

0.69

Activations Density 0.000%

references to violent acts

references to assassination

No Comments

No Known Activations