INDEX

Explanations

the prefix "po-" associated with various activities or objects

oai_token-act-pair · gpt-3.5-turbo

terms related to physical punishment or harm

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

owship

-0.92

ATIONAL

-0.76

hips

-0.75

IELD

-0.70

 Mellon

-0.68

 embodiments

-0.68

ITNESS

-0.67

AGES

-0.66

 EDITION

-0.65

AGE

-0.65

POSITIVE LOGITS

achers

0.96

pper

0.92

oper

0.92

pert

0.90

ppy

0.87

oting

0.86

ffer

0.86

Äį

0.86

etary

0.84

itives

0.84

Activations Density 0.008%

the prefix "po-" associated with various activities or objects

terms related to physical punishment or harm

No Comments

No Known Activations