INDEX

Explanations

references to the television show "Law & Order"

oai_token-act-pair · gpt-3.5-turbo

conjunctions and phrases indicating association or lists

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

auga

-0.91

 Cosponsors

-0.69

medium

-0.67

unts

-0.67

iliary

-0.66

bley

-0.66

lihood

-0.64

mint

-0.64

:[

-0.63

voy

-0.61

POSITIVE LOGITS

order

0.73

 order

0.70

 racket

0.66

 reconciliation

0.66

 economics

0.65

 medicine

0.65

riage

0.63

 looph

0.63

enforcement

0.62

 jurisdiction

0.61

Activations Density 0.100%

references to the television show "Law & Order"

conjunctions and phrases indicating association or lists

No Comments

No Known Activations