INDEX

Explanations

references to physical injury or harm

oai_token-act-pair · gpt-3.5-turbo

phrases related to injury or harm

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

 coherent

-0.86

izons

-0.77

rawn

-0.72

 appropriated

-0.71

 collaborations

-0.70

 Progress

-0.70

 revised

-0.70

 unified

-0.70

 Innovation

-0.69

 unification

-0.68

POSITIVE LOGITS

 beware

1.22

ometimes

1.13

 ALWAYS

1.09

 NEVER

1.04

WARNING

1.02

 annoying

1.01

 VERY

1.01

danger

0.96

 usually

0.95

 advisable

0.94

Activations Density 0.975%

references to physical injury or harm

phrases related to injury or harm

No Comments

No Known Activations